[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[GNUnet-SVN] [gnunet] Main GNUnet code base. annotated tag initial-impor
From: |
gitolite |
Subject: |
[GNUnet-SVN] [gnunet] Main GNUnet code base. annotated tag initial-import-from-subversion-38251 created. initial-import-from-subversion-38251 |
Date: |
Tue, 8 Nov 2016 13:49:04 +0100 (CET) |
The annotated tag, initial-import-from-subversion-38251 has been created
at c676f2c6c74febcd2256b2866ada43d9ed146c7d (tag)
tagging 5b9aeb0a9e957b084809679ca0fc83c24fde88d2 (commit)
tagged by Christian Grothoff
on Tue Nov 8 13:48:48 2016 +0100
- Log -----------------------------------------------------------------
Initial import.
-----BEGIN PGP SIGNATURE-----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=5kEI
-----END PGP SIGNATURE-----
Alejandra Morales (2):
Cryogenic stuff
GNUNET_TIME_UNIT_FOREVER_REL added to GNUNET_SCHEDULER_add_write_file call
Andreas Fuchs (3):
Initial conversation (experimental) commit
Added missing header file for conversation
Added test file for conversation
Andrew Cann (1):
Minor bigfixes.
Arthur Dewarumez (9):
Adding files for whanau.
Adding some function in the handler
Finish the create of the handler for every message.
Add some comments in the handler functions.
Start implementation of some functions.
Add delete_trail function.
Adding the handle_dht_p2p_trail_destroy function.
Adding function GDS_NEIGHBOURS_send_get_result()
Adding a function pick_random_friend ()
Bart Polot (3204):
Bart's public key added
fix compile error if MHD is missing
Added first data structures for the MESH service
Added more data structures and functions to build the skeleton of mesh
Resolved some weird conflicts
Fixed doxygen complaints
Changed message declaration location, added api messages
Changes to integrate mesh application_type re-definition
Added local mesh messages
Update gauger to allow high precision logging
Make mesh service compile
New mesh API
Work in progress
Adapt all API calls to the new style
Adaped to new API, documented
Work in progress, fix compile error
Simplified notification messages
Work in progress
Work in progress (handle_new_client)
Fixed compiler warnings
Finished an attempt on handle_new_client, changed message types to new API
WiP (create tunnel)
Unnecessary message eliminated, network byte conversion added
WiP
Fixed compile error
WiP
WiP
WiP
WiP
WiP
WiP
WiP
WiP
Fixed compiling error
WiP
WiP
WiP
WiP
Separated structs used in network protocol into new file, as recommended
in HACKING file
Fixed typo
Prepared for new API client implementation
WiP
WiP
Added mesh mockup test
WiP
Fixed wrong perception about how doubly linked list work
WiP
Added test mockup
Added mesh testing code and configuration files
Added logging, code formatting, documentation and style changes
WiP
Finished basic connect/disconnect test
WiP
WiP
WiP
Added mapping of client-service tunnel numbering
WiP
Updated gauger binding
Fixed gauger executable name
Fixed type in initializer
Use better condition
WiP (service rewrite)
WiP
Changed PeerInfo / Path structure
WiP (peer route creation)
WiP (adding paths)
WiP
CHanges in doxygen
WiP (handle of create_path P2P message)
WiP (add path to peer algorithm added)
Added search function for PEER_id given a PeerIdentity
WiP
WiP (mesh data handling and connection ACKs)
WiP
WiP (in-tunnel data forwarding)
Multicast and back-to-origin P2P traffic handling
WiP (deliver packets to client)
Fix last commit
Do not weight the average
Adapted nse-profiler for empty mesh multiple-peer test
fixes
Added function to send data to specific peer
Fix some FIXMEs
Added real ACK for path acknowledgement
fixes
Changed how messages are sent to clients to use Notif Context
Fixed message retransmission
Added handler for Path ACK messages
Fixes
Changed API<->Service messages
Change client <-> service traffic
Added API level handling of data and peer notification
Greetings from Philadelphia. Also, fixes.
Greetings from San Francisco. Connect_by_type registration and search
Local data delivery Fixes Tests
Various changes, can't get indent to work
Fix compilation of custom indent with texi2html v5.0
Fixed API test case
Fix tests, peer commands api <-> service
Indent
Add struct to indent
Indent
indent
fixes, documentation, refactoring
Header for new connect call
DIsabled the use of UNIX sockets, to allow running on Win platform
oops, one more unix path removed
Added dns section to try make it work on Windows
WiP
Added peer add timeout handling, tunnel connect/disconnect notifications
Added timeout & notification code
Changed add_peer to drop timeout, updated documentation, minor fixes
Updated header for change in service (drop timeout from add_peer),
updated doc
Changed variables and functions to meaningul names, changed debug
logging, fixed bug in priority queueing.
WiP
Implemented reconnect, creation of incoming tunnels, refactoring
Added tunnel destroy, changed message types in client <-> service protocol
Changes in API, adapted code and testfiles, improved client reconnect,
new documentation, small fixes
Fixed bad merge
Completed incoming tunnel handling, refactored destroy tunnel
WiP
Merged changes to destroy_tunnel, removed data pointer in th struct,
cancel pending traffic on tunnel destroy
Added cancellation of least priority packet on notify transmit ready
Fixed bugs Added exponential backoff in reconnection Fixed all FIXMEs and
TODOs
Added new test for multiple clients on one peer Fixed dumb obvious bug in
api
Changed documentation, type names and code formatting Extended test_local
Fixed compiler whining about test
More test code
Simplified own ID handling in mesh service
Extended test case, fixed bugs, refactored code and connection packet
format. DHT still does not return results...
Changed mesh service's storage to multihashmap, fixed bugs
Fixed doxygen warnings
Fixed test local to pass at least when DHT is not used
Added testcase for explicit case of DHT not finding apptype
Fixed broken assertions on abort
Added demnostration of PUT after GET not working
WiP
Changed behavior not to crash on external data
Fixed compiler warnings
Fixed typo
Added code for handling a case when a PUT is received with a key for
which there is an active GET query.
Fixed compiler warning in tests
Change datacache_get_iterator to use stack instead of heap
Added check of size of results
Fixed data size given to dht_get_iterator
WiP
Changed path management - Peers now have a set of individual paths to
reach them - Tunnels have a tree of the paths in the tunnel to reach all peers
in the tunnel in an efficent way, allowing to add peers efficiently taking in
account existing paths and having each tunnel reach a peer on a different path.
- Adapted helper functions, periodic announcements, etc to new structure Added
initial support for link disconnection notification and readjustment
Deleted old path to peer in tunnel tree before adding new one, conserving
the peer node
Fixed a use after free
Fixed adding path to tunnel, deleting old path and replugging node under
new one
Added notification of client that peers disconnected from tunnel
Fixed send data to origin to work with the new data structures
Ran indent script
Trying to make gnu indent's braindead indentation of comments *somehow*
readable
Fixed a use after free that occured when doing DHT_Disconnect while the
client was trying to reconnect to the service.
Delayed end of test to expose problems on slow machines (like a Core 2
Duo 9400 @ 1.86GHz...)
Added code to create siglne paths from trees Fixed create path packet
creation Fixed memory leakage on tunnel destroy Added more debugging output to
help locate problems Tried to change comments so gnu indent doesn't break the
code SO badly
Fixed reconnect Refactored connect packet sending
Fixed creation and destrucion of trees with parent tunnel
Added debug code, fixed minor bug
Changed own id handling
Fixed use before initialization
Fixed many bugs, test local 2 now passes, but with memory leaks
Fixed memory leak
Refactored MeshTunnel Trees and Paths in a separate file to allow
testing, since it's non-trivial code by itself and to allow future separation
in a different service.
Added path testcase
Cleaned and fixed refactoring to improve separation, only 3 structs are
now shared
Added testcase code, fixed minor bugs
Updated testcase
Renamed functions, make valgrind stop complaining about memory leaks by
explicitly freeing stuff before exit
Fixed use of uninitialized memory from realloc leading to segfaults and
bus errors
Fixed memory leaks
Fixed double chaining on note reattachment
Added debug print of tree structure
Added notification when peers go offline during a path change
Changed tree children magement from array to DLL, fixed bugs, extended
testcase.
Fixed bugs and completed API
Fixed double destroy
Added new api in the tree module to handle core or remote disconnections
Added reconnecting mechanism to find new paths in case of peer disconnection
Minor fix, comments
Adapted service API handling to new tree API
Fixed updating status of peer on ACK
Fixed compiler warning, otherwise WiP with multicast retransmission
WiP
Added multicast route keeplive
WiP
Fixed multicast retransmission, fixed memory managemenet at path creation
Added send path create on new peer with direct connection, added resource
liberation on neighbor disconnection which queued messages
Commented out DHT code, removed DHT-dependent test from run set.
Debug statements
Fixed a bug in create path handling, extended debug info.
Fixed size request from core on create path and path ack, reformatted
call to make more readable
Tweaked test configuration and log messages
Fixed core quota, added logging of error messages at connection setup
Fixed a double DLL_insert of an incoming tunnel
Optimized timing, fixed formatting, fixed callback bug
Added local tunnel service numeration handling, added incoming tunnel
client notification, fixed peer_id rc bug on core peer connect
Enabled DHT use again
Enabled DHT testcase, added unicast testcase
Enable DHT in testcase
Adapted path generation to new DHT_RECORD_ROUTE behavior
Moved dht get handles for types into tunnel handlers, added tasks to
restart dht gets periodically
Added periodic restart to all DHT get queries
Removed DHT_get restarting, set DEMULTIPLEX_EVERYWHERE on all DHT
operations
Added testcase to check proper working of insertion of own ID into a
2dtorus dht
Simplfied base multipeer testcase, removed tests from experimental
Tweaked configuration files and testcase
Renamed functions, make valgrind stop complaining about memory leaks by
explicitly freeing stuff before exit
Revert 17260
Stop all DHT get requests before disconnecting
WiP
Added debug info in service
Added new testcase for connecting two peers on opposite parts of a torus
Updated doxygen
Refactored connection to peers, cancelation of transmissions
Refactored sending a CREATE PATH message
Use new functions for CREATE PATH instead of explicit calling core
Fixed bugs, refactored path tree
WiP
Added mangaement of tunnels and paths by intermadiate nodes
Fixed order of PUT and GET paths from service to client
Fixes in path creation
Fixed bugs, added debug messages, extended unicast testcase
Fixed coverity 10777
Fixed coverity 10178
Revert 17243: not needed, coverity 10178 and 10179 false positives
Fixed coverity 10176
Fixed coverity 10118
Fixes to various coverity errors
Extended testcase, fixed bugs in client -> service data traffic handling
Various fixes in data traffic, expansion of unicast testcase
Expanded test to include traffic to origin, fixed bugs, separated unicast
traffic handling from traffic to origin
Additional extra check for coverity 10211
Removed unnecessary hashmap actions
WiP
Fixed bug with cancellation of pendint transmisions on peer disconnect,
added timeout for pending multicast data.
Refactored multicast code to allow sending any message as multicast,
added multicast testcase copied from unicast
Renamed
Updated TODO, docuementation, testcase
Added debug info for multicast
Fix
Made multicast work for single peer tunnels
Added tunnel timeout on inactivity
Change keepalive packets to explicit message type number
WiP
Added new message types
Added peer in tunnel deletion, changes to tree management
Small fixes
Updated doxygen
Added code to manage deletion of individual paths
Do not execute test under valgrind as default
Ignore test scripts
Fixed bugs in tunnel destruction chain
Added tunnel destruction to the testcase
Changed testcases structure
Added testcase for opposite peers in a small 2d grid
Changed testcase to include real multicast tests
Wait more to complete tests on slower machines
Solved Mantis #1834
Refactored to use functions instead of duplicated code for starting DHT
GET for ID
Integrated all 2dtorus test in one and added some more targets
Fixed compile issue, changed logging & dht_get_id working
Changed testcase to not die on unknown peer, log warning instead
Don't die on unknown peer
Fixed logging
Added connection broken p2p notifications
Refactored 'type' DHT response processing
New message type
Changed assert to break, not critical failure
Changed log, fixed crash in tree, caysed #1852 instead
Fixed #1853
Exclude multicast test from running by default
Fixed double free, formatting
Transformed assert into warning
Fixed bug to keep looking though list of clients on disconnect
Changed debug define for API
Fixed #1857
Eliminated full hash logging
Fixed a bug when adding a path to a peer_info, modified logging
Allowed to destroy NULL paths
Fixed #1856, added extra check for tunnel local numbering
Attempt to fix assertion failures on buildbot related to RC of client
(keep/drop)
WiP
WiP
WiP
Fixed testcase and small bugs in tree
Fix testcase
Simplified and extended testcase
Added transition code for new MESH API
Fixed coverity #10227, added doxygen to avoid bug again.
Fixed coverity #10228
Fixed coverity #10198
Fixed a memory leak
Ignore return
Fixed a bug when adding a path to a peer who is already a relay in the
tunnel
Fixed a memory leak when receiving a second create path for the same
tunnel
Extended logging
Dont add whole path to peer, just the relevant part
Fixed tree behavior for relay peers, adapted testcase to look for it.
Don't try to send traffic to a peer that is not connected to us
Direct paths can only come from CORE, ignore other sources about paths
with l <= 2
Fix coverity #10232
Added test to check why some buildbots fail connections and tests
Trying to debug buildbots
Fixed id counters' cycling
Fixed mantis #1870, stopped dht_get_type request twice on client
disconnect/tunnel destroy
Reverted change to conf file
Fixed leak
Improved legibility, formatting
Fix for matins #1871
Make tree library independent from tunnels, add cls to callbacks
Changed tree library to be completely detached from mesh, adapted
testcases
Delete ps tests
Added missing file
Initialize before usage
Disable DHT debug of MESH separately
Fix mantis #1874
Fixed error handling
Added check to catch #1871
Renamed functions
Fix for #1871
Additional fix for #1871
Fixed coverity #10250
Fixed coverity #10249
Reenable multicast testcase
Reduced duplication
Added an empty 2d torus test (connect and to nothing)
Added testcase, fixed naming
Added testcase for 2dtorus scenario from dht and mesh
Small fix, improved comment
More time to connect peergroup
Added comments, fixed debug code, added doxygen
Attempt to fix error on peergroup start on buildbots
Added logging, fixed doxygen
Fixed compilation error with new GNUNET_HELLO_Address
Added debug info, small fixes, doxygen.
Improved testcase parameters
Improved debug logging
Revert changes from r18076
More sensible values for testcase timeouts
Added TTL and packet ID to multicast packets, to avoid eternal
retransmission and packet duplication on trees with loops/mutliple paths
respectively.
Set appropiate error_level
Fixed debug stuff
Changed tree library: now assumes own short ID == 1, refactored code in
library and mesh service, adapted and extended library testcase
Fixed mantis #1896: check whether incoming tunnel is new before notifying
clients
Fixed issue related to mantis #1896, when a path ack of an already
connected peer would cause client notification about new peer
Adapted timeout for slower buildbots
Adapted timeout for slower buildbots
Fixed mantis #1900
Fix for missing me node causing assert and segfault errors
Temporal fix: Don't fail when peers reconnect on a better path
Real implementation for getting cost of a path compared to an existing
tunnel tree (additional hops)
Add info about peer's paths on ACK of a path
Added valgrind suppressions needed for Arch Linux and valgrind 3.7.0 and
glibc 2.14.1
Fix coverity #10263
Fix coverity #10264
Changed logging for testcase
Added code to avoid creashing on #1904 and identify the origin of the
missing path
Removed deprecated testcase
Fixed empty tree deletion problem (assert on test_mesh_api)
Dirty fix for #1920
Additional fix for #1920
- Fixed nse profiler handling of config files
- Fixed simplified variance code
Removed variance debug, initialized variance in WEST method
- Updated doxygen
- Same result, cleaner code
- Added explicit cast, just in case
- Fixed testcase evaluation condition
Use _weighted_ average of last estimations to calculate network size
- More changes in po files
-Fixed coverity #10265
-Fixed coverity #10264
- Changes to run 4k peers nse networks
- Do not reserve ports for unused services
- Removed misleading code
Mesh adds info about paths to peers as early as possible to avoid
unoptimal routes and artifacts caused by parallel path creation of less optimal
paths (loops, path corruption, etc)
- Off by 1
- allow slower buildbots to finish tests
- Use #defined constant, not direct numbers
Changed quota configuration from core to ats section, renamed from
TOTAL_QUOTA_* to WAN_QUOTA_*
- Even longer timeout to allow SPARC to finish testcase
- Fixed an insufficient allocation, probably causing OS X crashes
Make reeeeeally sure SPARC has time to do its stuff
Removed legacy code
- Add periodic statistic readings to output
Made sure that NSE never returns invalid standard deviation values
- revert 18495
- Added periodic transport & nse stats data logging
- Added logging of peers to NSE service
- Fixed memory leak / possible crash of freeing an invalid pointer
- Revert 18513, invalid fix
- Fixed message accounting, file creation flags, added logging in verbose
mode
- Gather more stats from nse runs
- Added more stats gathering code
- Wrong file commited, fixed.
- Fixed doxygen to reflect real API params
- Removed duplicate statistics, minor stylistic changes
- Fixed division by 0 on small values
Added statistic reading to NSE, fixed some minor bugs
New DHT-monitor functionality
- Use correct message types
- Remove unnecessary include
-Fixed coverity #10274
-Fixed coverity #10275
- Assert callback as in other calls
- Fixed some bugs in monitor code
Added testcase for DHT monitoring.
- Improved test to count only relevant messages, completed info
transmitted from service
Fixed queue bug
Temporl workaround for #2070
Temporal workaround for #2070
Fixed #2070 and simplified data transmission unicast/multicast handling
Added debug info for #2071
Fixed a bug when a multicast packet is delivered exclusively to local
clients
Allowed client to send again
Implemented workaround for #2071
Fixed client shutdown case, various minor fixes
Improved debug message
Fixed an assert error when a client disconnects with open tunnels and
without doing MESH_disconnect
Fixed a peer_rc bug.
Fixed client disconnect bug, delimited debug messages.
Fixed a memory leak on disconnect, double peer_rc decrease on tunnel
destroy, adjusted backoff, completed doxygen
Fixed #2087, wrong local tunnel number sent when multiple clients are
subscribed to one type message on same peer and one clientis owner of the
tunnel and the other is target
Fixed #2088, don't call receive_done on traffic not generated by client.
Improved debug output.
Added TTL and MID initialization to tunnel refresh packets.
Added more debug info
Fixed bug with remote tunnel traffic reception
Don't call cleaner on tunnels explicity destroyed. Updated and improved
documentation.
Reflected changes in r19282.
- fixed broken 19282 commit
- Attempted to fix #2095 (add incoming tunnel to incomings), don't let
destination destroy tunnel, disconnect instead
- Additional fix to #2095 (local clients can do ToOrigin too)
- Added debug info
Changed incoming tunnel notification to delay until relvant traffic is
received from remote peer. Allowed several remote clients for each tunnel.
- Fixed #2090, removed ugly debug lines
- Don't crash on bogus data from client
- Use a consisten (and correct) way to detect data packets
- minor fix
- Added initial speed test, fixed minor bugs
- Adjustment of error condition
- Added summary info for speed test
- Potential fix for #2104.
- When a client destroys an incoming tunnel, dont send him notifications
about the tunnel anymore
- Fixed coverity #10084
Workaround for #2104, initialize local tid when local type destination
connects after tunnel connect request by origin
- Fixed #2105
- Implemented a full explicit list-of-ignored-clients per tunnel to fix
#2107
Let the compiler not include debug strings in binary when make is not
configured with verbose
- Define new debug statements
- Improved readibility of code by removing a ton of #if DEBUG ... #endif
- Fixed compilation error, restored 2D torus as testing topology
- Fixed MESH API logging behavior and code readability
- Adapted testcase to working blacklisting, reduced number of peers,
topology line
- Fixed #1877
- Fix compile error on buildbots
- Fix cleanup on timeout
- Change DHT testing into common code, to allow adding further topologies
- Updated testcase to test for line as well as 2d torus with common code
Attemp to fix ctrl-c crashes - disconnect from services before calling
daemons_stop, since the cfg is free'd in the latter.
- Adjusted debug level
- Doxygen
- Doxygen, reconnect and let caller take care of CTRL-C handling
- Made cppcheck happy(er)
- fix testcase (use after free)
- fix coverity
-fix coverity
- fix for access of freed meter after ctrl-c
- mark meters as freed, documentation as missing
- attempt to narrow down os_priority bug
- Cancel task to prevent acessing freed memory
- Adjusted the monitor test to new topology
- Added missing parameter
- Removed old param name
- Fixed doxygen
- Fixed doxygen
- Fixed test_monitor testcase crsh
- Added documentation, clarified output text
- Fixed assertion in testcase
- Destroy tunnel to avoid api complaints, initialize return value
explicitly
- Removed MESH: prefix in debug messages
- Fixed reconnect procedure
- Avoid calling shutdown twice
- Fixed crash on DHT_put continuation
- Kill mesh on Ctrl-C
- Minor doxygen update
- Remove MESH prefix from debug messages, include missing testcase files
in svn ignore
- Without these lines MESH does NOT find the destination peer in the DHT
- Fixed bugs in notification of destruction of incoming tunnels
- Removed prefixes of DEBUG string, marked testcase output INFO, fixed
several bugs regarding shutdown
- Changes for long churn (test with 10 peers)
- Gather more stats
- Remove duplicate transport section
- Cancel tasks when freeing memory
- Fix crashing testcases
- Fixed crashes when ctrl-c testcases
- Attemp to avoid testing hang on ctrl-c during bootstrap
- Avoid calling daemons_stop with NULL, minor doxygen fixes
- Config file to reproduce #2170
- Clearer testing results
- Give more time for GET to suceed, minor cleanups
- Adjusted logging level
- Adjusted logging level, testing parameters
- LRN: Use RENAME() instead of 'mv'
- Fix for #2177: don't wait forever to reconnect to the service during
shutdown. In fact, don't wait at all.
- Fix for #2183: own position in tree only checked on new nodes caused
assertion error when own node was reinserted
- Attempt to fix #2181
- The next_message in the previous round should have the now
current_timestamp. - Check if it's better than our estimate before using.
- Fix comparison direction
- Fix contidion for next round
- Extend checking for activity to transmit handle
- Improved doxygen, added debug code
- Added more debug info, cleaned code, tried to fix boot time part of
#2202
- Fix related bug to 2202: 'flood messages not generated (lack of
history)'
- Fixed 2202, removed debug stuff
- fixed small bugs, added FIXME, added monitor client (to be tested)
- Fixed bugs of incoming data for tunnels without handlers, passing along
free'd memory.
- Fixed bug where ignoring incoming tunnel by owner caused tunnel
destruction
- Fix doxygen
- Add local DHT request to the monitoring results
- New callbacks for DHT monitoring
- Rewritten DHT monitoring
- Missing header file
- Adapted command line client to new monitor API
- Fixed service shutdown, workaround no longer needed
- Added key info on monitor tool output
- Make monitoring test compile again
- Fixed compile errors
- Wrong message type
- Wrong order of function calling, caused notifing of destruction when
ignoring.
- Fixed killing of random processes on shutdown
Added stop operation for dht monitoring
- Make test use monitoring_stop
- Fix #2312
- Added FIXMEs
- Clarify doxygen
- sane profiler config
- Document statistics api and fix testing's usage of it
- proper type
- minor cast change
Renamed GNUNET_OS_process_close to GNUNET_OS_process_destroy
- fixed memory leaks
- adpated code to styleguide, fixed more memory leaks
- fixed more memory leaks, doc, refactor
- Minor style changes, doc
- Added new mesh service (WiP)
- Makefile for mesh_new service
- WiP new mesh service
- cancel queue on disconnect
- make one global core queue
- fix
- fix dumb > vs < error, add debug messages
- fix speed test to finish
- remove old queue code
- add per-peer core queues
- add basic data structures for complex queues, refactor tunnel creation
- keep track of queue occupancy
Removed core queue requirements from mesh (backporting mesh_new)
- add basic REGEX API calls to mesh
- add peer blacklist API
- debug messages
- add mesh API to remove peer from blacklist
- clarify doxygen
- new mesh local messages
- document blacklist messages
- add api side implementation of blacklisting
- Added basic blacklist support
- blacklist api calls, service side
- doxygen
- added dummy mesh block plugin
- added mesh block header
- use proper block types
- move header to include
- use blocks to store type info
- use correct types
- complete block plugin, use block by put id
- use blocks for checking by type results
- use correct hashmap
- fix #2315
- api part of connect by string
- send packet
- add api for regex announce
- add service skeleton for regex connections
- more regex in mesh
- more mesh/regex integration: basic announce completed, lookup started,
wip
- wip mesh/regex
- initialize return variable
- initialize return variable
- implement mesh regex block plugin checks
- move mesh regex block code to a lib
- fix comments
- doxygen
- Reorder iterator arguments
- use hashmaps for active gets in by_strig connect
- finish implementation of mesh search by regex, testing pending
- added mesh regex testcase, fixed bugs
- fix shutdown crash
- more debug messages, tweaked parameters for testing
- mark old API calls used for previous VPN implementation as deprecated
- fixed bug for offset calculation, added debug statements
- improved debug messages, added break_op for malformed block
- simplified checks for success
- run test_mesh_regex by default
- remove d^Cug
- dont run regex test until mesh_new is made default
- reuse old results for looping states in the dregex
- use all functions, declare edge iterator
- temporarily avoid infinite recursion
- ignore kdevelop project/metadata files
- fix recursion in case of * states
- doxygen
- use different contexts for differents search stages/branches
- cancel search only after successful connect/ack
- keep track of all results, try others peers on timeout
- refactoring
-extend regex testcase, add debug info to service
- fix tests in offline situations
- full version of testcase, included data traffic to trigger individual
incoming tunnel detection
- enable all 3 strings/regexes to be evaluated against each other
- fix for a periodic accepting state, some refactoring
- improve logging, add debug message for GNUNET_break
- more debug messages
- typo in printf
- fix regex search timeout crash
- added debug messages, fixed regex timeout
Added REGEX announce and discovery to MESH
- fix #2498
- fixed use after free #2499
- give slow machines more time to complete the test
- add canceling of untransmitted requests upon tunnel destroy, to avoid
free'd memory access
- let tunnel empty queue before destroying
- fix memleak
- don't run valgrind
- fix memleaks in regex
- check for NULL
- use free_non_null
- try to fix compilation on Windows
- partial change, broken build system on laptop
- new API calls
- oops forgot includes
- make po files forget about mesh_new
- add packet ID to both multicast and unicast
- fixes, new message
- use configuration file, use statistics, bugfixes, doxygen
- use statistics
- collect more statistics
- fix coverity 10087
- fix coverity 10086
- listen to client again
- message format for flow control info
- unused msg numbers
- ack function
- doxygen, explicit raw packet queueing
- collect children information for ack
- more flow control
- handle data acks
- fix typo, adjust testing values
- add children info update on unicast
- doxygen
- don't allow unexpected ACKs
- fixes, doc
- more fixmes
- add tunnel and queue accounting
Eliminated mesh API buffering
Eliminate mesh priority option
- adapt callback to ack
- initialize pointer
- little refactoring
- fix sending, refactoring
- more debug logging
- static
- add a test for traffic between two clients of a single peer
- ignore executable
- don't clear th immediately, in case notify_cb calls notify_tmt_rdy
- do clear th ALWAYS after calling notify_cb
- mid -> pid
- debug
- wip flow control service<-> client
- dont send duplicate acks, allow null client
- more debug msgs
- not a debug message
- fix crash on typeless clients
- wip
- allow loopback
- fix client<->service flow control
- gauger test name
- doxygen
- timeout fix by harsha
- timeout fix by harsha
- fix client_tmt_rdy call, refactoring
- fix overflow condition
- improved testcase, feedback, debug messages
- more debug, client side ACK
- make sure to always have GNUNET_break immediately before
client_receive_done with SYSERR
- dont reject local ack from client
- wip client->service ack
- fix segfault
- wip
- wip
- use min/max functions to cover overflow cases
- refactoring, documentation
- fixes, more refactoring
-m fix client FWD ACK
- try to fix windows compilation error
- make executable, buildbots fail otherwise
- fix installscript for sudo when root
- wip
- initialize and consider ACK values fr tmt rdy
- fixes, refactoring
- changed client / children bck ack handling, wip
- initialize ack values on start
- towards local packet numbering for flow control
- typo
- wip
- wip bck ack
- fix PID sequence
- validate client generated PIDs
- validate client generated PID
- prepare for different local tests
- add 3 test in one to local traffic testcase source file
- add debug functions, refactor, add common code file
- refactoring
- use strings
- better debug
- fix FWD testcase
- fix error check
- ignore new binaries
- duplicate
- blindness cured
- send on correct tunnel
- use explicit tunnel
- avoid buildbots going crazy forever
- autocomplete fail
- debug
- more debug, fixes
- doxygen, refactor
- refactoring
- cleanup flow control code
- Clean debug output
- Fix local ACK
- more debug
- explicit bck counter
- more debug
- fix message send on reconnect
- refactoring, more debug
- adjust initial ack value for clients
- initialize recv pid, reset on reconncet
- use deltas in bck ack
- fix
- make api use same window size
- use window size client<->server, x4 faster
- fixes, eliminate malformed data assert
- clients sends ACK only on threshold, ~1.7x speedup on loopback
- oops, remove debug code
- added timing debug macros, improved local root->leaf performance ~20%
- don't show on empty data
- more info on error
- don't break build on osx, win
- fix peer fwd ack gathering
- fix unsolicited message control
- more debug
- fix SIGSEG
- refactor wip
- multicast flow control
- more debug, fix ack supression
- fixes, refectoring
- too much effort for buildbots
- not an error
- help analyze multipeer mesh logs
- send only one ack per packet, send before callback (callback can issue
destroy)
- separate multicast payload from control messages
- per-child buffer control
- prepare cinfo on queue_send
- explicitely drop oldest message
- update peer buffer on send
- leaks
- debug
- only manage buffer on payload
- Coverity 10269
- run p2p stress test by default
- Remove deprecated code, coverity 10236
- fix crash on disconnect
- debug buildbot logs
- Fix memory posioning on OS X (MSIZE is pre-defined to 256, doesnt
HAVE_MALLOC_USABLE_MEMORY -> compile error)
- more info on multicast fc
- try find where the extra ACKs come from
- be verbose on buildbot logs
- use proper next implementation
- looking for the stall bug
- make sure all int64_t are treated as such
- make sure -1LL is consistently used everywhere, document
- initialized ttl
- coverity 10210
- coverity 10222
- more debug info
- simplify multicast refcount
- simplify multicast client flow control, base all on ACKs
- correct update of tunnel queue for multicast messages
- debug
Added explicit keepalive to mesh
- More documentation
- Update Mesh Client <-> Service API documentation
- code format
- dont start vpn on mesh testcase
- dont start vpn on mesh testcases
- adapt mesh to strided regex implementation
- doxygen
- return void
- fix strided regex usage
- no return on void
- log errors only when they are errors
- fix ghost regex bug
- add min and nobuf options to tests
- svnignore new test
- timeout tunnel faster
- log timeout, try to fix ACK for min_speed
- reorder for case of shutdown
- doc
- delete deprected parameters, update doc
- fixed keepalive
- Avoid 4G loop, code style
- code style, refactor, documentation
- still needed
- show on buildbots
- dont destroy a tunnel right when a client leaves, send pending messages
first
- dont calculate hash if not necessary
- free info when destroying tunnel_destroy message
- Fix compilation errors
- timeout sooner on stall
- on queue destroy, eliminate the entry in the peer's children info
circular buffer
- dont send uninitialized memory
- initialize local variable
- fix uninitialized local variable use
- cancel tmt_rdy on peer disconnect
- doc
- dont use valgrind by default
- Big regex testcase
- fix
- run on computers with gnunet_prefix != /usr/bin
- fixes
Passing -v or -h options to a command no longer returns error code
- Exit on -v, -h
- Don't return error on -v, -h
- fix
- WiP
- Set correct path on remote systems
- no end coma
- ack off-by-one
- tunnel destoy is explicit now
- error messages more translator friendly
- disable debug by default
- fix mesh fc when buffering is off
- dont crash on cleanup
- notify clients when a tunnel times out
- Fixed memory leak
- Unsigned vlaue, can't be < 0
- New mesh mesasage type
- New mesh message
- WiP timers for ACK polling
- fix #2580
- Makefile changes to compile new tests for leaf to root speed
- temp. deactivate polling (causes crash)
- in speed testcase use a flexible number of packets
- change errorlevel
- add testcases for heavy traffic leaf to root
- fix crashes on backwards tests, doxygen, cleanup
- adjust log level
- refactor testcase
- initialize timer in all perf cases
- Mark queue entry as deleted before destroying This is to avoid
queue_destroy accessing that data structure and moving the entries around,
which causes later entries to be avoided (jumped over) and causes crashes on
core disconnect as queue_destroy accesses the then-invalid tunnel structure
- arm doesn't have time to finish in 30s
- adjust mesh clients to number of peers
- more debug
- check for to_origin duplicates
- initialize pid for case of error
- update bck pid on all traffic, not only local
- account for lost messages
- avoid crashing on shutdown in the middle of a test (timeout)
- open in second monitor if possible
- fix bck testcase, fix crash on warning, add FIXME
- fixed to_origin copy detection in multicast cases
- run more tests on buildbots
- more verbose debug message
- fix counter used for bck ACKs, style
- fix bck ACK value
- adapt to longer execution time while timing out soon on stall
- allow slow botnets enough time
- debug, style
- ignore
- debug backwards
- debug
- fixes, varbosity
- correct expected result
- off by one target for bck traffic
- give peer more time to find each other in the DHT
- correct debug
- allow n peers, typos, style
- enforce topology
- show info on buildbots
- fix bck numbering (fixes multicast but breaks loopback, I know)
- add compression length to mesh api announce_regex
- implementation of configurable regex compression
- fix testcase
- fixed regex offset, doc
- add new service to rewrite flow control
- new service
- unify client/child data structure
- refactoring
- refactoring, cinfo -> fcinfo
- refactoring, clinfo -> fcinfo
- refactoring, oneliners
- refactoring
- refactoring, make compile
- fix old mesh
- sync
- remove test code
- fix coverity 10365
- simplify bck, avoid sending duplicate acks
- backport fix to mesh_new
- doc, remove unused parameters
- fix
- clean config files
dont cancel mesh pending traffic on tunnel destroy
- todo
- allow force ACK to enable polling
- todo
- initialize all fcinfo fields
- fix stream
- fix polling
- don't poll so eagerly
- Keep calm and carry on
- cleanup
- dist fix
- dist fix
- dist fix
- dist fix
- Keep calm and carry on
- fix multicast keepalive
MESH uses DHT xquery to limit the results returned during a regex string
search
- doc
- add cause (line) of termination call
- more info on errors
- don't reject cached blocks
- use correct type
- test for mesh block cache
- dont record route for regex_accept blocks, avoid any kind of
duplication for datacache
- more mesh regex stats
- more mesh regex stats
- fix debug
- more mesh regex stats
- adjust default dht_replication_level
Added GNUNET_BLOCK_EVALUATION_RESULT_IRRELEVANT
- better debug, minor code cleanup
- debug
- save xquery for neighbors
- running valgrind by default makes buildbots unhappy
- fix typo
- exponential backoff for ACK polling
- TODO
- fix use after free
Added mesh CLI with basic tunnel listing
- try to fix buildbots
- oops
- apparently full paths break windows buildbot
- print whole ID
- Add whole-tree iterator, for debugging / monitoring / visualization
purposes.
- add testcase for whole-tree iterator
- add new message types
- wrong copy&paste
- TODO
- doxygen
- monitor tunnel inital implementation, WIP
- API additions for monitor tunnel
- Clarify doc
- changed mesh 'monitor' API
- changed mesh api for monitoring
- try to fix #2660
- typo
- fix compilation when HAVE_LIBGLPK is false
- debug info for stream / hostkey bugs
- fixed #2675
- wip
- fix default conf warning
- handle errors more gracefully, be more verbose for debug
- improve mesh handling of unexpected traffic
- silence!
- destroy tunnels that have no clients or child nodes after a timeout
- add children count api call
- use child count, add debug info
- die verbosely
- fix
- break regex into multiple messages
mesh now accepts REGEXes longer than 64KiB
Run testbed-helper in a sh interactive session to make sure the env is
set up properly
- use correct size for message
- feedback about abort cause
- signal both-ways tunnel destruction notification
- don't notify about tunnel that is dying
- avoid double-scheduling: causes crash on shutdown
- added regex performance profiler
- example data for perf-regex
- use new format for perf
- new perf
- typo
- missing header
- compress test data
- move regex_from_file and combine to separate lib file
- free memory
- more debug
- check return
- added mesh testing library
- try to use new lib. fail.
- not needed
- make distclean && bootstrap fixes
- allow NULL arrays of handlers and stypes
- adapt first test to mesh test lib
- c++ guards
- c++ guards
- extra includes to try fix w32
-do not permit queueing of messages that are too big
-prevent looping of mesh messages in the DHT
- use prefix
- make fixed prefix 20 bytes long -> 16 initial states (24 byte regex
prefix)
- fix makefile a/la
- regex test adapted to new mesh test lib
- port regex testcase to new lib
- give regex testcase more time for slower buildbots
- fix double free
- use API free
- remove unused variables
- coverity 10400
- coverity 10398
- coverity 10397
- port traffic testcase to test lib
-new conf file
- use consistent numbering
- fixes, consistency, test name
Remove testing_old
- add new test type, fix disconnect
- add new build for signal test
- refactor makefile
- fix
- fix
- not necessarilly a bug
- doxygen, alignment
- fix
- delay destruction
- initialize message for both directions
- unvalgrind
- notify owner of disconnecting peers
- dont destroy self
- account for disconnection
- unvalgrind
- ignore
- make sure GNUNET_TESTBED_operation_done is not called twice
- dont destroy tunnels with local clients
- revert 25599, fix tree destruction segfault
- fix memleak
- cache node parent to retransmit tunnel_destroy in case of pure relay
node
- enable signal testing by default
- not a bug
- fix
- show on buildbots
Move regex DHT integration from mesh to regex
- missing file
- doxygen
- cpp guards
- ignore log files
- debug messages
- alert about error
- fix
- debug messages, fixed assertions
- move code to regex_dht
- add debug for case of no xquery
- code from mesh
Moved regex profiler and profiler daemon from mesh to regex, adaped to
regex dht lib
- copyright notice
- cosmetic changes
- fix compile error
- pass local ID to announce
- missing argument
- fixed crash in case of DHT_disconnect called from a DHT_get iterator
- fixes
- slow down the searching of strings to avoid collapsing the testbed
- remove hack
- spread DHT PUTs over the announce period
- make a copy of the regex for announcing
- add randomness to the announce period: 10-20 min
- indentation
- doc, fixes
- doc
- more feedback: count and show connection errors
- profanity--;
- use ecc for regex profiler daemon
- fix
- use ECC in mesh
- change RSA for ECC in NSE
- read proper keyfile
- read proper keyfile, error messages
- read proper keyfile
- use correct config name
- fixes
- start the announcing daemon AFTER the testbed connect phase
- remove regex daemon dependency to peers' public keys: allow to use HEAD
regex with 0.9.5a core and others, allowing also committing regex changes to
svn HEAD
- fix crash
- not needed size of t->peers should suffice
- cancel dht get for peers when done, style
- fix segfault
- debug
- fix
- fix 2699
- verboser log, faster start
- Remove non-distributed seach: buggy and not useful - Do search with a
maximum parallel factor, starting a new one when an old one is done or times out
- keep track of the search parallelization
- start announce immediately (profiler takes care of load balance),
reannounce immediately for more avaliability
- not needed
- not error
- Don't silence WARNING
- fix off-by-one, imporve log
- Unify log format
- verbose++
- verbose++
- keep regex to HEAD
- keep track of fails/successes
- fix in monitoring, documentation, style
- fix in monitoring, style
- path log documentation, style
- logging, cleanup
- add INFO level meaningful logging, documentation, style
- keep dht up to date
- fixes, debug
Remove CORE corking from DHT traffic
- less log pollution
- dont offer hellos if not going to use them
- dont even start a FIND_PEER if TRY_CONNECT is disabled
- fix message
- new debug separation for dht-clients
- dont #define debug output, use getenv
fix
- more debug info
- more verbosity in DEBUG
- debug
- debug xquery
- allow GNUNET_BLOCK_evaluate on PUT requests for regex blocks
BLOCK_EVALUATE regex PUT messages - reorder debug info for easier
following
- Add show parent key for every edge (grep friendlyness++)
- Remove printf, use GNUNET_log INFO
- debug leftover
- debug leftover
- adapt to new ARM, debug
- debug info
- fix arm_api crash
- doc
- fix
- correct type
- use constants for delays
- avoid direct printf when possible
- doxygen
- doxygen
- change regex daemon to announce every rand(0, t_max) period
- fix compile
- Change profiler to start search with a delay realtive to the
announcement
- simplify start
- fix use of search_index on start
- fix
- fix index
- fix app types allocation
- doxygen
- sync wait period with deamon
- 80 char align
- fix
- debug
- dont print the whole regex on debug
- exit after N rounds
- use GNUNET_TESTBED_peer_manage_service instad of manually connecting to
ARM
- fix appearance of '%s' in error messages
Dont restart services/daemons that have exited with exit code 0.
- remove old internal testbed code
- fixes, refactor
- move initializtion to run()
- fix
- bratao's fix to compile on w32
Add filename getopt operator
- change profiler cmd interface
- wip
- fixes, debug
- remove extra debug
Allow callback to stop a DISK_scan with no error, similar to other er APIs
- doc
- fix
- add disk scan
- cleanup
- fix test timeout task: do collect stats on timeout
- add feedback for testbed setup
- allow more time for testbed start
- fix
- remove warnings
- update service new to ecc
- move parallelism to config file
- don't require hosts file anymore (testbed accepts NULL)
- don't require hosts file anymore (testbed accepts NULL), now for real
- speed up shutdown
- simplify parameters
- unify testbed operation handlers, avoid pointer arithmetic
- more verbose debug
Initial mesh2 api import
- remove unnecessary code
- remove more unnecessary code
- make api compile
- remove regex integration from mesh2
- wip
- wip
- make mesh2 compile
- dont compile old tests anymore
- remove regex from mesh
- remove regex client-side
- remove remaining regex code
- remove application-related option
- Fixes
- simplify adding path to tunnel
- fix accidental removal
- proper memory usage
- clean up
- remove trees from tunnels
- remove one indirection level (no multicast ref counter)
- remove the use of ntermediate MeshPathInfo struct, all can be passed in
tunnel
- adapt path keepalive to unicast
- adapt path keepalive to unicast (retransmission)
- change internal API to use short IDs
- implement new fwd/bck flow control mechanism
- fix poll for lost ack
- cancel poll when receiving big enough ack
- owner does not timeout tunnel
- avoid mixing path_refresh and tunnel_timeout
- Change hash to speed up 16/32 bit lookups
- re-added APP_ANNOUNCE_TIME, still needed for mesh v1
- revert not-ready changes
- remove capability of multiple clients per tunnel
- use faster hash for tunnel hashmaps
- refactor fc init
- doxygen
- adapt queue selction to new fc
- remove transmission descriptor, simple cls is enough
- fix ack value sent on data
- style
- fixes
- Added status flag for tunnels
- Remove old tree lib
- Fix
- fix core error handling
- dont use per-hop packet ID
- make sure all flow control values are updated correctly
- Added tunnel port parameter
- use port for tunnel creation / notification
- missing api changes
- notice
- stop using message types of payload on service side
- use 3-way handshake CREATE-PATH_ACK-ACK to delay traffic on new tunnels
- fix
- add mesh2 local testcase
- config files for test, ignore
- debug
- always set prev & next hop to allow reuse of p2p code for local-only
tunnels
- fix memory leak, path assignment, peer_rc updates
- doh
- wipe in case of local variable usage
- dont run PREFIX by default
- start listening to clients only after CORE connection
- avoid double free
- use lib
- doc
- explicitly log requests dropped due to TTL exhaustion
- use dht-traffic class for service-client as well as p2p
- remove old debug
- wait for dht connections
- undebug
- automatically adapt links to peer
- use testbed SETUP_TIME + 1 min margin for aborting
- typo
- auto set FORCE_NSE
- fix hosts file command line option
- fix setup timeout
- show more info on debug
- verbose error
Fix regexes with accepting common prefix, optimize regex prefix lengths
- expand regex perf to show resulting DFA: evaluate simplification from
r27315
- valgrind fix
- collect traffic info by default
- disable debug by default
- adjust timeout when no more strings are left to be found
- use 40 conn/peer
- patch to reduce transport workload in testbed setups
- fix timeout handling
- fix compile
refactor, improve debug output
- improve debug output
- how did this disappear from here???????
- fixes
- remove fixed window size from service, fix crash
- unify flow control with service, adapt to client-controlled flow control
- remove flow control constants
- fix crash, fix success conditions
- don't crash on errors
- fix shutdown crash
- fix shutdown segfault
- again, how did this get here?
- avoid infinite loop when destroying local loop tunnel
- fix memleak
- initialize to proper new values
- fix memleak
- obsolete break
- mystery #3: why was this deleted?
- relay PATH_ACK to client
- refactor, check stack buffer size
- tmp hack
- fix testcase
- remove unfounded ERROR message
- add framework for mesh2 mutipeer tests
- fixes
- notify success
- dont wait so long
- Adapt new testing lib to mesh2
- fix compilation errors of new multi-testcase
- more work towards generic mesh2 test
- fix remaining warnings
changes to MESH API: add receive_done, document end handlers
- adapt mesh api to explicit client flow control
- change service to support client explicit ACK - simplify tunnel
destruction code - typos, documentation
- use explicit ACK in local testcase
- use explicit ACK in small testcase, remove multicast leftover, fix
- fix
- fixes, debug
- disable try_connect to force topology
- use ports
- fix
- fix 2897: data handler SYSERR closes tunnel, not whole handler
- adress bug #2896: call end handler on requested tunnel destroy
- document bug #2896: call end handler on requested tunnel destroy
- document bug #2896: domunet tunnel_destroy behavior
Change in mesh API: data callback no longer provides sender, user must
remember tunnel<->sender if needed (fix bug #2894)
- use new data callback with no sender info
- combine payload encapsulation structs
- drop to_orig messages on 0 ttl
- fix memleak
- compiler warning
- typo on debug
- fix unicast/to_origin oversimplification
- fix peer_id rc
- make sure not to leak rc on shutdown
- avoid double free
- warning
- fixed use after free on shutdown
- avoid valgrind uninitialized memory warning
- fix 2906
- make sure to free all peer_rc on error
- fix unrelated bug in 2906 comments
- add client connection debug info
- add paths to self from core, no need for dht lookup
- separate key generation from path managemente
- avoid double adding of tunnel to peer_info for local tunnels
- doxygen
- style
- use clearer debug numbers
- doc
- fixed 2906
- update to mesh2
- just update fc->queue for payload messages
- ignore
- when creating a tunnel, pre-accept first incoming packet
- fix success condition
- use more universal names for test discrimination
- add equivalent of old tests to mesh2
- update ignores
- fix nobuffer behavior
- force timeout during tests
- fix path refresh
Replace mesh with new version
- fix braindead win32 build process
- freebsd doesn't have log2
- note to self: when debugging, use correct executable
- remove dead code
- doxygen
- no longer used
- add drop option for future testcases
- update messages used by mesh
- prepare for relaible mesh
- ignores
Added mesh reliable tunnels
- fix unreliable traffic
- adapt testcases to new api
- update message_to_string
- fix local infinite retransmission loop
- debug, indentation
- fix assumption that owner and destination client must be different,
(bug #2924)
- fix unreliable traffic bck
- fix keepalive loop
- test order
- add reliablity testcase, fix nobuf API use
- new config file for reliability (enable drop)
- ignores
- indentation
- fix relay data retransmission of repeated messages
- send data_ack from the same ack function as hop-by-hop acks
- fix segfault on retransmission
- changes not intended to be commited yet
- don't shcdule retransmissions of unreliable packets
- don't overwrite logs
- fix unreliable expected traffic
- don't reject future traffic, related to bug #2934
- fix script
- use multihashmap32 instead of own implementation
- use deltas for ack polling
- don't use deltas, send last PID in POLL message instead
- fix
- simplified MESH client<->service API as per suggestions in bug #2934
- rollback
- client can be owner AND destination
- use proper reliability metadata
- when did this disappear????
- change retransmit time to #define, debug
- fix decoding of tunnel options
- nbo
- use different DATA ACK for both directions
- updated message numbers
- missing message types
- data_ack ack's all previous messages, change hashmap to FIFO DLL
- don't let retransmissions mess with per-hop ACK
- doxygen
- avoid POLL messing with reliability counter
- avoid POLL messing with reliability counter
- drop newest message when a retransmission comes in
- queue_destroy does all the work
- fix memleak
- more mess from retransmission of data
Finishing mesh reliable: - per-tunnel retransmission, always retransmit
oldest not-ACK'd message - adaptive retransmission delay -- start with 1s --
use per-message timing to update expected delay on each ACK as d = avg (d, new)
-- update pending retransmissions with new values
- fixed weird 'feature'
- fix use after free
- adapt delay after each ACK
- fix segfault on error
- use API change from r27944
- use correct ACK for nonreliable tunnels
- calculate round trip time of real message, not original
- don't retransmit a message if it's stil in the outgoing queue
- limit retransmission buffer size
- use message ID for end-to-end ACKs
Add context to a client immediately after connection. Otherwise calling
SERVER_get_client_context on a client that never sent an initial message (and
didn't have a chance to receive something via SERVER_set_client_context) causes
an assertion failure in server.c:355. This happens when the client can connect
to the service on multiple protocols, one of the connections will be
immediately closed (see bug for ghost client disconnects)
- connect_notify is also called with NULL on server shutdown
- fix disconnecting of ghost clients
- fix unreliable unicast
- undo mix of pid and mid
- fixes
- when injecting a retransmitted message in the queue, update its pid to
the lowest from the queue
- debug
- off by one
- separate hop and end ACK at destination
- filter by type
- don't retransmit so eagerly
- tweak timing, improve precessing of received messages
- use enum, add out of order option
- initital implementation of bitfield ACK-ing of future messages
- update status of client protocol
- bitfield does not loke NBO
- use one tunnelnotification struct
fixes. lunch break
fixes
dont skip when mid > traget
- doc
- fix end-to-end stalling, update todo
- ignore
- debig
Change get_info API
- use fitting data type
- fix polling in the presence of lost data
- don't allow more than 64 messages in the input buffer of the owner
- Poll whenever the queue is not empty
- don't discriminate against unreliable tunnels
- fixed
Refactored code to eliminate duplicates depending on FWD/BCK traffic
- debug
- change MID to 32bits, allow overflow (fixes lost packet 0)
- off by one
- debug
- unchoke client after 64 messages hace been ACK'd
- use same options everywhere
- disconnect tunnels before destorying mesh handle
Remove stream
- don't do valgrind by default
- destroy tunnel at the end of the function
- uninitialized memory on the network
debug tunnel destroy
-cancel polling on tunnel destruction/empty queue
- make send_client_destroy generic and same-client-friendly
- update cleaner in testcase
-use new API
- start poll on queue_add
- log event, not an error
- use proper ID
- return on error
- refactor
- cleanup
- TODO
- gns depends on vpn
- doc, warnings
- debug, cleanup
- debug
- fix ACK for TO_ORIG traffic
- don't zero send flag after a cancelled transmission
- fix
- fix condition to omit hop-by-hop ACK
- major refactoring, implement bi-directional keepalives, retransmit
initial messages when lost
- missing message types
- log tunnel state evolution
- fix cov10603
- Fix coverity 10602
- allow same owner && client
- make sure fc->t is initialized
- fix
- move tunnel number verification to function
- log wrong TID
- fix mantis 2954
- try to work if both RELIABLE and NOBUFFER options are selected
- unregister ports on client disconnect
- cancel retransmissions on destroy
- copy and d'oh
- inverted comparison operators
- if the next peer has destroyed tunnel, don't poll, destroy any leftover
traffic
- not an error
- add port number to debug
- not an error, messages could have crossed on the wire
- cancel polling on cancel_queues
- remove assertion: can happen that if both own and cli send a destroy
tunnel, funcion will be called twice, this is not an error
- don't try to do stuff on an error callback
- send ack could call queue_add and set the handler
- don't leak memory on repeated PATH_CREATE
- doxygen
- don't use peer_info as a data strcut name
- doxygen
- add separate encryption enabled mesh to avoid breakage during
developement
- dead code
- copy mesh protocol for encrypted version
- add .h files for new mesh api
Change MESH API to use Tunnel to use Channel
- missing file
- wip
- don't build new mesh for now, too broken
- new data structures
- change function declarations to new data structures
- doc
- New message names
- connection creation function
- more renamed message types
- new P2P messages
- connection messages
- new connection messages
- connection message construction
- fix
- calcualte path cost including overlap
- Changed tunnel_use_path to use connection
- remove duplciate enum
- compilation fixes
- typo
- ack is independent from tunnel
- wip
- connections dont have options
- wip
- wip
- wip
- wip
- fix
- fixes
- channel management
- channel messages do not include tunnel info
- more channel modifications
- refactor
- encapsulate channel traffic in tunnel message
- mesh encapsulate message types
- channel destroy notification
- channel message types
- new channel message types
- fix
- dont send p2p traffic for local clients
- connection, tunnel destroy
- tunnel, connection destroy
- fixes
- channel destroy iterator
- connection timeout management
- queue_destroy, queue_send
- fix
- queue_add, priorities
- handle connection_create
- dead code
- fixes
- connection ack
- fixes
- handle connection destroy
- doxygen order
- channel data manipulation
- channel data ack
- wip
- doxygen
- fixes
- add const status to getter params
- generic handler for encrypted data
- change hop-by-hop ACK to be per-connection not per-peer
- handle new ack/poll connection-aware messages
- handle connection keep-alives
- new client handles
- apparenty inline comments are bad
- handle local create channel, add channel id for global, root, dest
- demultiplex encrypted payload
- use struct fc instaed of pointer
- handle P2P channel messages
- wip
- fixes
- fixes
- fixes
- major refactorization
- fixes
- add exerimental compilation of -enc
- use new mesh for test_mesh_local
- aux function
- missing header...
- fix
- fixes
- rollback
- use after free
- fixes
- tunnel queue
- doh
- fixes
- update debug info
- fixes, removed dead code
- fix queue destroy order
- fix core cancel on shutdown
- fix memleak
- not always an error
- refactor queue_add, simplify cleaning up
- fix scan defects
- refecator
- doh
- rewrite flow control towards clients
- avoid double pid increment
- fix loopback channel creation (service already knows channel)
- add -enc (again??)
- add generic tid/cid filling, ack processing
- add channel_ack
- add channel ack type
- mark deprecated, use generic data
- use generic data_ack
- add generic channel buffering
- change id handling as discussed on friday
- fix debug message
- fix
- sync
- fixes
- fixes
- refactor client flow control
- fixes
- use -enc for test by default
- sync
- better debug
- indentation
- fix ack direction
- more refactoring client flow control
- make a generic ack function
- separate channel from connection ack counting
- sync
- avoid loopback ack nightmare: send local traffic directly to
destination client
- fix shutdown when root and destination are the same client
- remove channels from client hashmap on disconnect
- fix memleaks in loopback case
- fix
- check hashmap for initialization
- initialize all fields
- fixes, document weirdeness
- sync
- fix PID accounting for congestion control
- debug, minor fix
- fixes, debug
- fix 3way handshake
- don't crash on NULL tunnels
- dead code
- fix neighbor disconnection notifications
- send connection handshake before data
- fix 'synack' processing
- debug
- record channel number correctly
- dont insert channel in tunnel before search for free ID
- dont cancel poll if ack is not enough for future traffic
- fix queue unlocking
- use queue, not ack, for buffer size calculation
- cancel old ACK/POLL when issuing a new one
- fix ack'ing - resend TUNNEL_DESTROY - explicitly log about faulty
routes instead of letting core_api complain
- fix crash
- send data directly if tunnel is already online (or local)
- multiple fixes for connection-less loopback traffic
- avoid double free on shutdown
- fix double destroy of half-open loopback channels on shutdown
- dont prevent destroy in case of fully initialized channel
- use GNUNET_log_config_invalid
- cancel core_tmt_ready on core disconnect
- doc
- tweaks for hex experiments
- dont require xquery for PUTs
- fix block validation for DHT PUT requests
Add REGEX_INTERNAL_iterate_reachable_edges which only reveals edges
reachable from states with proof longer or equal to REGEX_INITIAL_BYTES
- use REGEX_INTERNAL_iterate_reachable_edges
- fix
- fix
- decrease amount of PUT rounds to finish all before the GETs start
- indentation, style
- more log, style
- more log
- more feedback regarding shutdown
- setup time feedback
-fix
- undo r28718
- normalize with profiler
- install daemon in libexec
- extra debug
- extra wait time
- debug
- patch to ats performance
- debug info
- too much uncommited work
- update messages
-loglevel
- fix compilation
- more refactoring
- sync
- sync
- fix channel static functions
- finished coarse refactoring
- MOve peer2s to peer section
- fix main mesh file
- add new modules to Makefile
- move all core/PeerIdentity interacting code to _peer, remove
GNUNET_PeerIdentity from _connection
- export core handlers from connection: all p2p messages are connection
related but core is connected to peers
- use per-module logging
- fix LOG
- move connection initialization to GMC_new - fix core references
- use public channel handlers
- add peer's tunnel setter
- share one statistics handle across all subsystems
- fixed _tunnel - solved data struct visibility
- cleanup
- sync
- sync
- move get channel to local
- fixing channel functions
- local channel id better accounted for per-client
- queue is internal: channel -> tunnel -> connection -> peer, no clean
way to access for now => remove optimization
- one catch-all ack handles renedered obsolete
- add callback when we give a message to CORE
Calculate outgoing connection performance
- typo
- connection fixing
- move queue canceling to peer.c
- use different enum to prevent collisions
- fixes
- order
- remove old include
- move path manipulation to API
- share ID
- fixes
- fixes
- fix connection.c
- use const for peer IDs
- expose appropriate functions in API
- add polling API to connection
- change queueing API, adapt connections
- move connection message accounting
- use proper const qualifier
- fix queueing
- fix peer
- fix dht
- fixes in local
- sync
- migrate handlers to channel.c - clients shouldn't know about tunnels
- sync
- sync
- fix compile errors in local
- fixes in DHT handling
- use new API functions
- fix "broken connection" notifications
- fix encryption/decryption visisbility
- fix wild replace
- simplify acks
- fix channel data/ack
- use proper channel_ack/data_ack
- backport
- mesh builds again
- fix build
- debug
- fix shutdown
- address FIXMEs
- avoid infinite ACK loops when creating loopback channels
- send data ACK only after receiving the channel ACK
- use server<->client ACKs, not reliable
- avoid double ACK'ing to clients on loopback create - fix ACK id being
sent - debug
- always update client's rel info on ACK
- use send_client_ack
- typo
- move debug log
- wrong flag...
- GMCH_send can try to free saved message, save first
- debug
- uninitialized memory
- fix new connection handler
- multiple fixes
- unify payload, fixes
- debug
- fix conection message accounting
- fix segfault
- refactor
- add separate encryption enabled mesh to avoid breakage during
developement
- undo accidental commit, previous comment should be: accounting for
connection ACK messages
- flow control changes and fixes
- fix ACK direction
- fix flow control
- debug
- doing stuff after destroying a connection? not good...
- fixes
- doxygen
- doxygen
- doxygen
- doxygen
- doxygen
- doxygen
- doxygen
- doxygen
- cancel all DHT GET request while shutting down
- debug
- fix use after free
- doxygen
- doxygen
- stop recursion on destroy
- debug
- dont send connection broken on terminal peers
- fix reentrant looping
- fix use after free
- document possible 'hidden' frees
- doxygen
- log
- avoid double-adding a channel into the tunnel list (causes
use-after-free on shutdown)
- fix leaks - refactor to simplify APIs
-debug
- fix double free
- remove debug
- fix connection destroy accounting
- encryption stuff
- missing header
- kx
- missing headers
- new message for mesh KX
- change key exchange messages to own encapsulation
- cleanup
- refactoring, extra checks
- document, alpha renaming
- fixes, doc
- header
- doc
- fixes
- doxygen
- refactoring - fixes - doxygen
- fixed state overwriting
- don't handshake on loopback tunnels
- loopback traffic redirection
- use explicit kx send function, avoid many ugly hacks
- fix mixed-direction connections on tunnel
- clean up API
- make sure to free all memory in possible luibgcrypt leak
- internal api change: add channel direction autodetection in
non-loopback channels
- fix loopback channel create
- remove all possible ambiguous 'fwd' parameters in Tunnel API
- new message to fix #3091
- local message tyope for channel NACK
- fix #3091
- test for #3091
- new message type
- fix create handler when port is closed
- log
- doc
- doc
- move connection set code to connection.c
- doc
- doc, fixes
- account for out-of-order kx messages
- don't allow ACKs to change KX state
- ditto
- debug
- fix kx
- allow test to run offline
- dont confuse doxygen
- fix polling
- fix
- allow to cancel connection messages - change poll/ack mechanism
- fix NPE
- don't allow stalled reliable channels
- fix double free
- don't die on transmission error
- fixes
- timing fixes
- log
- fix channel flow control in corner cases - adapt connection polling to
traffic
- fix NPE
- remove warning
- account timing info in "backlog" free
- fix memleak
- log
- free data copy in case of error
Use encrypted MESH by default
- rename tunnel->channel
- send ping on rekey timeout
- remove superfluous tunnel state
- avoid timeout in test_reliable_* if KX message is dropped
- use shorter timeout for connection establishment (CREATE/ACK) to avoid
timeout to test_reliable_* if a connection management message is dropped
- fix memleak on KX drop
- log
- doxygen
- use correct binary
- reorder shutdown
- log
- when did this dissapear????
- dont allow a client to send more data when output buffer is full
- fix coverity 10859
- mark fall-through explicitly
- fix memleak on error (cov 10812)
- fix coverity 10814
- explicitly ignore return value (cov 10819)
- ignore return values
- non-null check always performed before call: not needed in the function
(cov 10802)
- fix NACK handling (related to cov 10832)
- fix cov 10808
- ignore return value (cov 10870)
- show WARNING if elements were not in the hashmaps
- dont use a cancel handle if cont is NULL, fix potential memleak
- add tunnel callback for retransmissions
- dead code
- dont use uninitialized memory
- initiate retransmit timer only after message has been sent
- doc - use tunnel queues on channel retransmit
- fix use after free on shutdown (#3112)
- dont double free in case of error
- cancel queued message on connection destroy
- log
- missing return
- doc, notes
- fix handling of path's on duplicate PATH_CREATE messages
- fixme
- destroy channels when tunnels is being destroyed (but complain:
channels should have been destroyed first). Probably wrong shutdown sequence.
- don't reschedule a POLL if connectio is being destroyed, hopefully fix
#3111
- delete tunnel from peermap at shutdown
- api fix
- remove warning
- fix shutdown order
- missing header
- more shutdown fixes
- don't crash on channel not found
- destroy all connections on shutdown, so core_disconnect doesn't cause
errors
- log
- log
- fix #3111
- testcase for #3114
- fix #3114: add feedback from _local to _channel on who is requesting
the channel destruction, notify the other part
- fix 2966: destroy all channels on service crash/disconnect
- fix crash on GMC_create (3131, 3123)
- allow for destruction of channel inside inbound handler
- fix port mangling
- log
- dont resend useless channel_destroy messages
- dont send ACK on dying connections
- log
- dont try to send data on connections that are going to be destroyed
- log
- tunnels need separate connectivity state and encryption/handshake states
- fix case of reconnection during valid handshake
- dont update tunnel c-state on data
- fix consensus profiler test (multiple channels reusing same
tunnel/session)
- log
- remember caller source
- dont send anything to clients shutting down
- log
- fix line for client audio retransmission
- plug leak
- supress compilation warnings
- remove old dead code / message types
- use only once KEEPALIVE message
- remove old message types
- update message types
- revamp mesh_common code, minor refactoring
- log
- typos, indent
- send conncetion ack before kex exchange -> cleaner state maintenance
- use new keepalive message type
- warning--
- remove deprecated tests
- use callback awareness to cancel unneeded traffic
- use force flag to avoid connections dropping channel management data -
multiple bugfixes with channel shutdown
- doc
- process channel destroy notifications properly
- do not count channel management traffic as payload
- initialize timing info with real round trip time if available - fix
channel queues on unnecessary retransmissions
- wip
- only complain if mid is BIGGER than expected, because sender sent
unauthorized traffic. if its smaller it is just an unnecessary retransmission
- fix messed scheduler management in data ack handling
- fix super-fast (< 20ms) retransmission behavior when the RTT is smaller
than the core delay
- doc
- code cleanup
- retransmit channel create based on separate timeout tasks
- initialize timers
- legibility
- refactor
- refactor mesg sent
- use queue/task mechanism for SYNACKs
- fix segfault
- log
- loopback traffic should not touch tunnel code. Mark as a warning,
replace with assert in the future.
- cleanup, get rid of goto
- cleanup tunnel api - doxygen
- don't mark duplicate channel_confirm messages as error
- use cleaner code
- dont authorize a client to send more data in case a duplicate
CHANNEL_ACK is received while the connection buffer is full
- avoid warning on successful shutdowns
- log
- timing
- improve dealing with duplicate channel create (in case of replay attack
/ missed channel ACK)
- dont use millisecond restransmit delays for "synack" retransmission
- avoid dereferencing copy after channel (and therefore copy) was
destroyed
- treat data as ACK on the destination end of a channel
- fix procession of channel create in loopback channels
- fix compilation issues
- nonexistent channel does not necessarily means an error, could be
cross-notification of channel_destroy
- change mesh channel creation/notification API
-doc
- add peer cmp
- fix channel numbering
-wip
- fixed compile warning
- wip
-wip
- wip
- wip
-create channel
- basic connect/listen
gnunet-mesh has netcat capabilities
- wip
-fixed 3174 part 1
- doxygen
- log
- fixed 3174: check reliable gap before allowing client to give more data
- remove dead code
- don't flag as ERROR if channel is loopback (no retransmissions
necessary in that case)
- fix #3172
- dont generate errors on async channel destruction, use stats instead
- dont show as error (break takes care of that)
- dont allow NULL connections in a tunnel
- improve handling of duplicate connection_create
- log condition prior to abort
- connection_cancel_queues can destroy connection, make sure to add
retransmission to queue before calling
- conditional break already called freom register_neighbors
- relay error to peer, indentation
- fix handle for tunnel queueing during KX
- crazyness (see doc)
- fix extra iterator step
- simplify the crazyness of sending tunnel-queued data: avoid going back
to channel, no more channel parameter necessary
- cancel queued transmission on channel destroy
- dont send connection create for dying connections, avoid segfaults by
the way
- dont check destroy, opaque pointer
- wip
- refine assert condition
- close statistics handler on shutdown, fix #3194
- don't confirm channels scheduled for destruction, client handles are
NULL'd
- don't send ack on channels to be destroyed
- new plugin API functions
- log
- doc
- dont retransmit connection_destroy to self
- simplify tunnel delayed queue
- react to connection destrutcion in exisitng tunnel
- fix warnings
- cancel transmission altogether on error
- wrong condition
- dont GNUNET_break on non-validable paths, DHT seems to routinely be
aware of core connections faster than MESH
- return NULL when continuation is NULL, avoid channel:860 break
- only complain if channel is not being destroyed (client has forgotten
about it already)
- don't return on found connection, more checks to go
- if a kx fails due to reconnect state, don't complain (it'll get
retransmitted eventually)
- fix delete loop
- fix 3163#c7868
- various minor fixes and breaks
- hunting 3214
- signal shutdown, don't create new connections, fix 3214
- not an error, always happens on relay nodes
- don't loop on the same invalid path
- fix assertions
- avoid double free / use after free
- small refactor
- fixes
- simplify error handling
- log
- fix
- debug
- not a break
- debug info
- stop kx on empty tunnels
- loglevel
- remove connection from tunnel when no longer usable
- fix prev commit
- don't complain about async PONG messages
- fix memleak on tunnel destruction during kx
- duplicate PONG can happen if PING was re-transmitted but first one came
through: second PONG will be "unexpected"
- only treat KX as confirmation if connection is not in destroy state
- debug
- fix timing out of connections
- fix
- don't keep alive connections that are scheduled to be destroyed
- debug
- don't show WARNING/ERROR on wrong connection create, debug+statistics
instead
- log - don't process ACKs on dying connections
- debug info
- not a protocol break, plausible asyc msg
- not a break, can happen if only one path is known and a connection is
already using it
- logs
- missing message type (accidental deletion?)
- don't always cancel channel's task on confirm: might be a duplicate
confirm and task is then a copy retransmit
- refine break condition
- register keepalive for destruction
- debug info
- refactor and improve path creation
- fix log msg
- always allow sending KX messages
- restore break
- fix connection ack direction, revert assert to break
- dont count destroyed connections
-call GMT_sent when canceling tunneled buffered messages before the KX
completes
- use tunnel encryption state to select decryption key
- fix state selection
- include all message types in switch, complain if unknown
- respond to data on unknown channels with destroy (avoids
retransmissions)
- dying connections can drop traffic
- log malformed pings
- log fixes, limit kx to live tunnels
- dont destroy a path right away, broken paths can cause long loops with
outdated DHT data
- fix core_disconnect handler to avoid reusing the direct path being
destroyed
- send notification immediately if a connection fails to be created -
allow to queue messages with no connection
- show WARNINGs for bad handshakes
- rmeove break
- log
- log traffic size
- debug info on API side
- use proper signedness
- use write to allow binary data as well
- debug
- fix compilation issue
- use STATIC buffer, remove debug
- take care of PEER RCs
- Don't crash, _break instead
- Add some of the tunnel introscpecion back
- missing header changes
- timing log
- use explicit GNUNET_NO instead of 0
Add echo functionality for mesh CLI
- fix echo flag
- ping every second
- new message types for debug
- add -i function back to CLI
- missing cli<->serv messages
- copied and adapted from testbed (previous failed during 30c3 with IPv6
VPN setup)
- refactor channel nack
- more tunnels info for CLI
- missing headers, removed message types
- remove extra \n
- delay tunnel destruction 1 min to avoid rekeying
- do exponential backoff for connection creation
-revert r31823, it broke gnunet-ecc CLI tool
- mark task as completed
- only re-send keepalive/recreate once the previous one has left the peer
Add NSE proof of work standalone tool and benchmark
- don't resend create messages for channels being destroyed
- use proper log function
- fix error messages
- add user feedback
- various CLI fixes
- fix compilation
- attempt 2
- complete tunnel info gathering
- ignore binary
- add explicit shutdown state - don't try to reconnect when destroying
last connection during tunnel_destroy - don't create new connections for a
tunnel in shutdown
-when tunnel is empty -- don't end connections -- don't stop encryption
handshake
- add warning if check fails
- dont schedule keepalive on connection creation, wait for first creation
message to be sent
- dont start keepalive on intermediate peers
- wait for conncetion_ACK to start keepalive timer
- Test if we are shutting down (on Ctrl-C, for instance)
- don't try to rekey loopback tunnels (#3254)
- ignore gnunet-ats-sim
- fix tunnel info message type
- changed option to get all tunnels from -i to -T
- small CLI fixes
- add peers info to mesh CLI and API
- use correct data type
- verbose log
- don't show output on stdout in case of error - terminate after response
- fix show tunnel
- use response struct size
- copy peer id from request
- use variables to cound ch, conns - use ntohs for tunnel state
- change tunnel API to pass connection and channel payload
- show connections and channels on stdout
- document tunnel internal API - change iterator parameter order
- service-side implementation of peer queries
- tunnel is a boolean (int)
- client-side implementation of peer queries
- expand peer info API callback
- adapt mesh CLI to API changes
- fix nbo
- log
- doc
- use a memory-efficient, cleaner info API (limiting concurrent info
requests to 1)
- fix SEGFAULT on ctrl+c in testcase
- script for overnight checking that: -- check succeeds -- no coredump is
generated -- no assertions are getting triggered
- debug strings for new message types
Allow gnunet-ecc to list all public keys in a file
Use proper core priority for each message type and message origin.
- add hello messages to peers
- add hello subsystem to get hellos for known peers
- doc
- TODO
- merge hellos in case old hello is still valid
- doxygen, return on NULL
- dont return expired hellos
Use stored hellos to try to establish direct connections to peers with
tunnels
- refactor logging to allow for shorter log files
- use hach directly instead of _2s function
- copy hellos since parameter is const
- use correct function return type
- fix segfault at get size of null hello
- add hello lib
- do'h
- wrong assertion
- store own hello
- simplify timing out of old hellos
- ignore NULL hellos
- don't call set on NULL hellos
- doxygen, add get_mine
Use DHT_HELLO for MESH peer discovery, benefit from wider replication
- remove MESH own block code (MESH uses DHT_HELLO now)
- don't run test on shutdown
- log
- fix PUT, use proper key
Fixed GNUNET_HELLO_get_last_expiration, as it always returned min (EPOCH,
expiration times), so: EPOCH
- fix crash on NULL hello with valid ID
- log FULL keys on mismatch
- wipe all previous data in *key so there is no potentially uninilialized
memory
- log levels
- change log info of LOG_TRAFFIC, use R5N to distinguish from XDHT
- fix search, use target ID, not own
- whitespace
- use HELLOs from DHT to try_connect on TRANSPORT level
- log
- remove _break
- Added HTML log viewer
- changes to web log viewer
- use AJAX to avoid loading DEBUG messages (breaks firefox by multi-MB
log files)
- cleanup log messages
- extract peer information for filtering
- cleanup level code
- only hide body rows, leave header
- add peer selection
- select only level from tr classes
- show only peer rows of active level
- fix all and none buttons
- consider peers on level change
-add class and peer_id to ajax-loaded rows
-add color
- fixes, fonts
- colors
- use new service startup message
- simplify message names
- reduce debug messages amount of hello operations
- reduce button size
- fix log messages
- use fixed-width message type names
- add missing message type names
- limit message name width
- revamp logging
- use <pre> for messages
- standarize log messages
- do not mark <pre> background
- log
-log
-add filter by component
- sort peers and components to ensure consistent order
- filter peers and components aware of each other
- fix level selection
- remove debug
- log payload ID for channel traffic
- refactor fwd/bck connection keepalive
Move mesh connection keepalive messages inside tunnel encryption to
reduce information leakage.
- register keepalive as tunnel payload
- remove dead handler
- remove unneccessary reserved fields
- remove reserved filed code
- add hmac to tunnel messages
- fix expected KX size
- cleanup
- add keepalive stats
- update config files for tests
- add keepalive testcase, statistics gathering
- evaluate keepalive counters
- add makefile for new testcase
- fix scheduler clearance
- debug, indentation
- run statistics service by default
- don't use global peer counter, cls instead
- finally fixed missing keepalive messages
- warning on direct KEEPALIVE sending
- svn:ignore testcase
- fix warning
- encrypt-then-MAC
- Define MeshHash
- use MeshHash
- replace all HashCodes with short MeshHashes in Mesh messages
- refactor
- add const
- generate smaller hashes for Conn IDs
- use an array of 256 hashes on connection info
- use MeshHash in connection code
- use 256 hashs for hmac
- inital mesh_profiler commit
- use testbeds stats
- add a getter for peer->unique_id
- add const
- use testbed's get_index to identify peer in array
- minor fixes
- prepare profiler: start peer, get IDs, log at the end
- change test to profiler
- fix compilation issues, add to Makefile
- connect in clique
- change cls to pointers instead of indexes
-add info about traffic
- use pointers in struct itself
- fix warnings, remove explicit keepalive count
- use global hashmap for IDs
- remove unused constant: test based on time, not messages
- rename mesh profiler
- name inside file
- use proper Hash
- use ping-pong to avoid collisions
- refactor
- dont ping after timeout
- don't send duplicate pings when a pong is delayed
- nse-style round cycling
- check incoming channel for NULL
- fix infinite loop
- log, debug
- gitignore
- fix cleaner
- mark peer as down
- cancel pings on peer shutdown
- update peer counter
- check task (could be in the middle of transmission loop and be 0)
- don't shutdown twice
-debug
- fix operation handling
- NPE
- fix run flag
- fix disconnect task scheduling
- disconnect all mesh handles
- only ping from select peers, do not shut them down
- dont disconnect from mesh service
- doc
- doc
- use explicit ping/pong messages with counter and timestamp
- debug log
- set searching state before trying to reconnect, otherwise waiting is
overwritten
- log
-log
- add connection count check, doc
-remove dead code
- establish connections on new paths
- add DISABLE_TRY_CONNECT option
- set disable_try_connect for tests/profiler
- remove explicit keepalive rest
- add message pop from transmission queue
- log
- requeue messages on connection broken
- dht debug path
- test delay
- check channel creation
- copy & fail
- create own config file
- adjust transport delay
- changes to round system, stats collection
- add stddev
- read peers form argv
- wrap profiler in script to alter config file
- hide template
- hide temp config file, show config template
- ignore
- not abort condition
- allocate memory, not array anymore
- log output
- move pings to cli
- check early
-move round time to cli, fix config file
-log
- avoid excessive keepalives
- clean stdout output, log debug still
- add warmup phase
- fix path optimization bug
- fix false break() on retransmisson of connection_broken messages
- use full path debug
- dont show warning on duplicate broken connections
- connections no longer have native 512 hashes, translation is used
- expire data with hello
- adjust retry time to expiration
- adjust nse on the fly
- optimize
- simplify, improve to 15000 lines/s
- start test only after warmup
- fix
-fix own position
- log
- check path length AFTER shortening
- fix infinite resend loop
- use proper type
- eliminate broken paths
- elimiate all paths
- invalidate instead of destroying, to avoid reusing right away (from
stale DHT info, for instance)
- don't delay tests
- fix peer to notify
- send ACK on duplicate channel_create
- set first channel number properly
- fix mask
- don-t re-scheduler, destroy immediately (comes from scheduler)
- cancel SYNACK retry task when forced to SYNACK by an incoming SYN
- fix
- pass warmup parameter
- debug
- dont assert
- profiler changes
- fix handling of duplicate incoming channel create with respect to
queued retransmissions
- start test if warmup "fails"
- add more links
- optimizations for runtime
- fix
- fix
- remove unnecessary break
- move mesh hash to public api
- implement h2s
- use small mesh hashes for info api
- add peer option, move listen to free '-p'
- implement client-side peer_id info request
- don't dereference NULL tunnels
- peer info api
- stop dht route search on tunnel destroy
- use specific GMP_ function to stop dht
- fix queue priority
- remove explicit keepalive from queueing
- log
- shutdown dht last to allow tunnels to clean searches up
- doc
Cancel CORE transmit ready request when cancelling last transmittable
message
Do not account canceled messages as sent regarding flow control
- check handles on send_echo, fix #3349
- log checksum error (for #3333)
- fix uninitialized variable
- remove ERROR log
- rename 256 bit mesh hash type
- use new hash name
- log
log
- fix coverity 12989
- fix dead code covertiy 12876
- refactor, check for existing copy
- fix count check
- use only one core, to fix logging race
- fixme
- fix #3364, establish a new connection to a peer when a core connection
becomes available and a tunnel is present
- missing merge
- don't complain, from tunnel it's not possible to check
- show error context
- dont send keepalives if payload traffic is already queued
- ignore logs
- ignore
- remove old tests
- fix cli arg parse
- log
- always check if channel was last, then mark tunnel for destruction
- comments
- send buffered data to client before destroying channel
- destroy a channel if some traffic is missing and the channel has been
remotely destroyed
- fix compilation
- just create one connection, from smaller to bigger ID
- reuse test config file
- make sure handle is NULL
- assert order
- rename test file
- ignore
- log
- remove drop
- fix names
- add context in path to link connection
- ignore
- start reconnecting before running out of connections
- log
- don't schedule tunnel destruction on shutdown - remove any scheduled
delayed_destruction AFTER destroying all channels - fixes #3366, #3372
- fix path string representation
- keep track of connection pids on queueing/sending
- fix use after free
- log
- don't cancel a connection's destroy message when destroying a
connection and canceling all its traffic. since the connection will be freed,
remove the connection pointer instead
- if data comes from an unknown channel (we might have just destroyed it
and the service doesnt know), ignore and warn, dont crash
- log
- skip all metadata update on destroyed connections
- don't use NULL connections on callback
- less verbose debug
- don't free connections in the middle of being build. - fixes #3373,
#3361
- use constant instead of hardcoded value
- verbose error
- de-duplicate ifs
- KX are also keepalives
- log
- add debug function to inspect queues
- fix log/indentation
- show debug info on each queue_add
- log
- debug queue on send
- log on INFO the data MID received
- more info in log about payload
- log
- comply with doc
- allow NULL peer id for broken connection
- doc
- use full peer IDs
Notify neighbors if received messages about unknown connections,
implements #3378.
- const
- log
- add PID getter
- Eliminate next_pid
- fix buffer fill for payload messages
- debug info
Rename mesh->cadet
- fix makefile
Rename files
Renamed directory
- update Makefile for new cadet dirname
- rename supp file for some valgring distributions
- update gitignore with new mesh/cadet filenames
- remove mesh references
- update config file
- more renaming
- change contents
- move include file
- update makefile
- update .am files
- pot update
Update all am files
- rename man file
- update fs
- rename also contents of config files, update config files
- update renaming to new version of scalarproduct
- update config file contents
- rename in conf.in files, update conf.in file still referencing mesh
- when a malformed PING is received, re-send KX and own PING
- send KX even when a tunnel is being destroyed, remote peer might be
trying to revive it
- resend KX on malformed PONG
- fix doc
- revert logging changes back to debug level
- change GM (Gnunet Mesh) function prefixes for GC (Gnunet Cadet)
- stop using CadetTunnel3 and go back to CadetTunnel
- storage for old _encryption_ keys
- doc
- refactor key generation, cache peers ephemeral key to create new keys
immediately
- use old key for encryption if key exchange was recent enough
- move HMAC verification to decryption function
- don't inifinite loop on error
- refactor decryption
- don't destroy the KX context right away, wait 1 minute for possible old
traffic
- use cadet for log script
- log
- don't use the old key during a KX
- revert log
- change connection density for cadet profiler
- speed up ID announce to force different PUT paths in the DHT
- use increasing size data messages
- perform tests with only 2 peers for basic P2P test and with 5 peer line
for testing of relay peers
- add default size, doc
- fix id check on data callback
- simplify names, update Makefile
- ignore tests
- fixes
- remove hardcoded target peer numbering
- fix testbed statistics collection
- fix stats collection
- rename tests
- fix name detection
- fix test tmt_rdy size request
- code review notes
- update default values, eliminate obsolete ones
- intialize a new kx_ctx on receipt of a new ephemeral key if none exists
- debug info, refactor
- expire old keys adaptively
- kx debug
- reduce rekey period for tests, ensure even shorter keepalive
- verbose tunnel debug in error cases
- test and log for bug #0003423
- Allow two GC_2s calls in one line (limited reentrant code)
- limit queued messages on tunnel level when tunnel is not ready
- only create keys on new ephemeral info
- add rekey state
Allow to send traffic with the new key when the rekey period is over. In
case of a late PONG, very high latency or partial traffic censorship, this will
allow the receiver to still receive the traffic if the ephemeral key did in
fact get through.
IMPORTANT FIX: use return value for integrity check
- check message size to avoid infinite loop
- send 1000 messages to force re-key testing
- rename
- add 2s for new state, reduce buffer
- refactor key selection, hmac calculation
- fixes, renaming
- refactor kx ctx creation, add debug
- fix KX for very short rekey periods (seconds)
- rename
- fix crash on unexpected malformed PING
- make sure the keepalive test generates more than 2 keepalives
- ignore
- ignore
Add a debug "dump to stderr" feature to cadet
- fix NULL connection on queue_send
- fix ERROR log
- add a peer debug function
- add debug function about a peer for #3405
- fix params
- add last contact, DHT handle to peer debug info, for #3405
- use setter
- more debug for #3405
- fix peer debug null deref
- refactor peer debug
- attempt to fix #3405
- use setter, imporve logs
- doxygen
- use same loglevel as connect
- more info API
- ignore
- don't try update info on NULL connections, avoid segfault for #3453
- check return values from GCC_send_prebuilt_message
- remove debug info used to fix #3405
- more verbose warning message
- don't use payload_type on non-payload messages
- update m2s for non-payload
- debug info for #3425
- adjust test's log levels
- log
- ignore
- debug
- revert to standard Crockford base32
- fix log
- don't update ACK based on queued and yet undelivered messages - send
ACK upon delivery of queued messages to client
- don't warn when connection is being established
- fix error condition
- fix #3479
- refactor
- destroy kx_ctx created on receipt of an ephemeral key if key is not new
- dont destroy in-progress kx
- trigger an immediate rekey if a peer's ephemeral key changes
Added a cleaned-up vanity key generator to gnunet-ecc
Since we started using ECC, peer identities ARE public keys, not their
hashes. This commit removes the miss-documented, redundant option and the
duplicate code associated to its handling.
- Revert "Fix transport blacklist tests" - This reverts commit
3f7a1601abadaa0a8036a39e5227ae4c156dd50e/r33895: we went back to standart
Crockford base32.
- Convince 'make -j 10' to work nicely.
- debug info for buildbot reports
- fix connection pending counter
- dont use pointer after free
- verbose debug
- fix #3483
- document code, small sanity checks
- include order upsets kdevelop
- fix variable log level, add connection debug function
- include NULL guards in debug
- allow to run specific checks in a loop
- when a connection is broken, remove it from tunnel, allow tunnel to
update cstate and try to create new connections
- warn when test goes wrong
- test for coredumps and assertions on loopcheck
- cancel any rekey tasks when calling it directly
- check that the connection belongs to a tunnel when keeping it alive
(otherwise, it might be BROKEN)
- simplify loopcheck
- save old logfiles
- don't re-encrypt data queued in a broken connection
- show queued messages when allowing client to send more
- more debug info on received ACK, doxygen
- don't dereference c in connection pop, c might be free'd if previous
call popped last message
- as connecton_pop destroys many message type wihtout return, count is
not applicable
- doxygen, code formatting
- notify about deleted messages when popping messages from a connection,
allowing to keep a count and make sure the connection is destroyed
- On a new EPHM, do a immediate rekey
- debug INFO for missing keepalive in BCK
- schedule next keepalive even if tunnel is still in SEARCHING
- debug info for tunnel status mismatch
- fix tunnel state management after connection failure<
- fix
Force sending of KX messages even when connection queue is full
- more debug info on buildbots
- persuade social to at least compile
- convince gcc to compile
- add connection state BROKEN for connections destroyed on behalf of
intermediate nodes
- Keepalives should NOT be transmitted in the clear
- only an error if counter was wrong
- make sure connection is destroyed and no use-after-free happens
- fix logging
- clear ready flag on send
- make sure head of pending messages is > expected message, fix integer
underflow
- correct data feedback
- log payload of dropped data
- send past messages to client, as they are ordered below ACK
- count a MIMD as received only after sending it to client
- send ack AFTER removing message from queue, otherwise ack triggers an
assert for sending and ACK of still-existing data
- revamp almost all cadet tests, keep track of messages, sizes, acks
- in ACK tests, the initializer message is part of the data exchange
- restore long tests
- fix compile
- don't send keepalives if traffic is queued
- debug info on error
- inform about dropped messages
- if an "old_key" is not valid (for instance, at tunnel establishment),
clear the context ASAP
- fix handling of SYNACK corner case
- fix error condition
- set tunnel cstate to SEARCH when 0 connections are present
- use WAITING if some connections are in progress
Resend traffic queued on connections that time-out, if the peer is an
endpoint.
- dont complain about tunnels in search of a path
- fix trim-after-free
- missing cstate2s
- don't revive dead tunnels
- dump keys to STRERR on debug, if flag is on
- more info on buildbot errors
- mark connections as to be destroyed by the _sent continuation
- don't break on old errors, check for ERROR level log messages
- better feedback to users about duplicated messages
- don't show error messages if tunnel's state is WAITING
- don't try to salvage traffic if no payload is queued
- show tunnel info on error
- remove unencrypted keepalive sending
- debug info on queued messages when popping
- log
- code review
- add address@hidden
- remove duplicate/incorrect API call
- remove dead code, coverity #13008
- free memory on error, coverity #13058
- check path length, simplify error conditions
- indentation, fix old(er) gcc compilation warnings "may be uninitialized"
- remove dead code
- fix faux compiler warnings
- fix false assert trigger when message #0 is lost
Use optimizaiton of paths on store from DHT as well as on receipt from
CONN_CREATE
- off by one: add own ID to ID array when creating path from DHT
- signal when dht path is not correct, style
- wrong condition, random paths removed instead of direct ones, fix #3531
- ignore for git
- ignores
- test all log files (except .log)
- log
- fix assertion during single-connection timeout, update docs
- fix connection.c:972 assertion, mark tunnel as WAITING when an incoming
connection is received
- only try to pop messages if there are any, otherwise directly destory
the connection
- fix logging of key material
- debug key clearing
Import RPS
- fix ext leftovers
- disable NSE proof of work in the test
Add maximum message size
- proper CADET overhead
- avoid usage of direct stdio sprintf
- use proper STDIO
- log
- fix resetting of key cleaner task
- doxygen
- doxygen
- refactor enc_state changer
Fixed key context manipulation in case of duplicate EPHM messages
- fix minor t->estate issues for better debug traceability
- log
- fix sending of queued tunnel data on KX end
- log EPHM on receipt
Fix conservation of old keys in case of simultaneous KX: when creating a
new kx_ctx, copy old keys based on tunnel estate, not finish_task
- more debug for received traffic
Fix message resend on connection timeout: mark connection as destroyed to
avoid re-queueing in self.
- debug for missing paths on incoming messages
- debug for KX
- refactor resend_messages_and_destroy for popping and sending messages
of a dead connection
- remove duplicate logging
Fix resending of connection traffic * avoid SEGFAULT on connections with
no tunnel * don't leak memory
Never invalidate a direct path. (Like after getting a CONNECTION_BROKEN
on a direct connection)
Fix memory corruption: new connection always makes copy of path.
- fix segfault
- move script to proper dir
Make process_log more generic
- don't break if connection is empty (no messages to dequeue/call
callback)
Fix connection count to count only ESTABLISHED connections, use DEFINE
instead of hardcoded number
- raise loglevel of timeout events
- more debug info
Send all queued messages that fit into the core buffer - Add timing info
for debug
- leave space for any number
Fuse PING into KX_EPHEMERAL
- send queued data the first time
- make connection always log with ---, reserve === for upper levels
- calculate the correct amount of messages to allow per connection
- use proper wait time for core timing
- Added correct core request timing info
- refactor buffer filling
- don't try to ack on non-ready tunnels
- clear core handle to avoid shutdown crash
- fix PID management on send
- having allowance too spare is not an error
- check KX context at PONG
- avoid duplicate (same path) connections
- more debug, eliminate bogus assertions
- refactorize path checking code, look both ways
- use path equivalency
- destroy old connection in case of duplicate connections when ownID >
peerID
Fix channel buffer retrieval, add debug
- ignore
- Always send traffic coming from channel, even overfilling connection
buffer. Clients won't get an ACK anyway.
- always allow an ACK
- refactor kx_ctx management
- fix tunnel encryption state on time triggered rekey
- signal wrong tunnel encryption state
Only allow one EPHM/PONG in the queue.
- fix cadet's overhead
Fix empty tunnel detection
- update test config
- test message packing
- avoid noise in the testcase logs
- add space for second kx
Added keyfile with peers named AAAA, BBBB, ..., ZZZZ (to improve
testcase/profiler legibility)
- check minimum size, param order
- fix size handling of payload data
- add size argument to allow to verify message size
- more size checks
- more message size checks
- simpify api size handling code
- minor changes: tabs, logs, reorder
- fix test: provide real message size
- test connection_create size, add size to log
- add statistics to DHT module
- refactor test, always gather stats
- minor log modifications, docs
Accept Out-of-Order connection messages
- allow more time to 5 peer testcase, change stats loglevel
- remove debug statements
- include message size in log
Fixed inverting of same-priority bundled messages
- fix crashes, zombie controller on ^C
- bump message amount
- fix failed git merge
- added path_cmp
- use path_cmp
- Check path for NULL, fix #3583
Added a delayed search function to avoid saturating the DHT with requests
on boot.
- log
- changes to handling of own identity
- alignment, style
Documentation update - fix contradiction
- more documentation
-remove prefixes
- get size direcly from sampler
- never resize to 0
- doc
- debug script to all gdb with the correct core file and executable path
- fix memory management of sampler tasks
- fix memory mangement when calling cadet to detect liveness
- RPS sampler destroy also accesses peer_map
- don't free static memory
- avoid chain-calling the destroy handler to destroy the channel
currently being destroyed
- avoid NPE
- memleaks
- fixes for NPE
- doc
- added test for EC-DH with Peer IDs
- added asymmetric crytp perf
- fix sig size
- ignore
- remove .gitignores
- ignores
- added basic axolotl support
- reduce number of messages for snappier test
- refactor to use same function to decrypt based on tunnel's enc method
- data struct for axolotl internal state
- refactor to check messages from both enc systems
- fix const, log
- count traffic as connection confirmation
- NULL connections means error
- if C is NULL, can't use it for getting the ID
- const
- refactor kx handling
- document axolotl fields
- doc
- ax kdf
- cast own key
- doxygen
- remove dead code
- memleak
- generate a new ephemeral ratchet key on tunnel creation
- add alice/bob order to kx
- axolotl encryption
- doxygen
- typo
- use proper connection to determine direction
- don't send receivers ID, not needed
- add kx ephemeral key to ax context
- add send kx for axolotl
- don't check own ephemeral
- use axolotl by default
- fix
- add cid field to ax_kx
- register ax traffic
- refactor kx sending, unify under send_kx
- fix alice-bob identification
- ignore unknown connections
- debug
- debug message name
- debug info
- fix wrong state modifier
- implement single peer info
- due to incompatibilities in the representation of the ECC public key
for ECDH and EdDSA, peers must provide en explicit ECDH key and sign it with
theis EdDSA ID.
- add ECDH key caching and verifying
- added a signature purpose for signing of axolotl kx messages
- refactor otr - use signed axolotl ECDH - doxygen - log - minor fixes
- log
- don't do otr-rekey if a tunnel is Ax type
- doxygen
- set tunel's e_state from axolotl
- split hmac_derive in two: hmac and hmac_kdf
- implement ratchet advance
- fix pointer
- don't count all messages as pid when checking, only payload
- log format
- refactor axolotl encryption
- add AX type to send processors, refactor cid setter
- store peer's permanent key in ax
- add getter for peer's permanent ECDH key
- axolotl debug - show only relevant tunnel key info on debug
- log
- handle AX messages
- adjust checks for different encryption overheads
- use correct overhead size, log wrong sizes verbosely
- shorter names for more info in 80 cols
implemented AX dectyption and racheting
- remove dead code
- fix size output, don't include headers
- register AX for unqueueing
- allow just message headers (keepalive uses this)
- remove debug
- proper name for continuation
- use ax as payload
- keep track of all message types
- refactor
- fix log params
- fix Nr counter
- initialize reserved
- properly read otr/ax fields
- don't ttl drop ax
- activate ratchet on receive
- ttl is deprecated, don't warn
- fix ttl
only ratchet after a certain amount of messages or time has passed
- register otr-downgrades
- don't queue more than one ax_kx at a time
- resend ax kx when lost
- only modify to key sent before the key derivation
- skip checking all messages that are not there, no matter what the
bitfield says
- doc
- refactor storing AX keys
- critical error
- refactor buffer calucations
- remove debug
- don't queue more than 3 messages
- fix debug msgs, mark tripped break
fix traffic stall in case the is more messages queued in the tunnel
buffer than a connection would accept at once
fix excessive queueing, don't force regular traffic
- improve POLL logging, debug info
- add n_recv to debug info
- fix log
- more messages to force errors
Fixed channel recv buffer space accounting error when receiving a
duplicate message
- fix debug messages
note to self
- fix poll log
- start poll after send
- move break to documentation
Always allow a connection to receive data in a tunnel, even if all
buffers are full: new channels and keepalives must be allowed.
- fix returned buffer amount, to make sure channel bugs don't affect
connections
- refactor connection timeout into one implementation - always send
CONNECTION_BROKEN to alert of the timeout
- compiler fixes
- make sure tunnel queues are consistent
- log
- fix shutdown memleak
- roll back duplication of peer->connections
- don't invalidate direct connections unless peer disconnects on core
level
- make sure there are no duplicates in the hashmap
- since the connection->id reduction hashmap puts use a helper function
with local static buffer, workaround
- fix memleak
- destroy peers on shutdown
- free hello when destroying peer
- cleanup
- add core timing to peer debug
- only log time if core is requested
- always free connection path, fix memory leak
- free path after unregister
- free skipped keys on tunnel Axolotl destroy
- fix compilation
- check connections on peer destroy
- complain about excessive core waiting times
- fix
- fix error message
- if this happens the code is badly broken, bail out
- documentation, statistics about failures
- If the path was cleared before, connection is not registered with
neighbors. Should fix #3845, #3846
- destroy queued messages only after making sure no channels have handles
to them
- don't try to retrive a peer during a shutdown task
- on shutdown it will happen that core will return 0/NULL, this is not a
warning
- Fix #3846. When receiving a create request for an invalid path singal
that register failed to avoid trying to unregister the connection on destroy.
Merge branch 'crypto'
- don't torture buildbots with 5000 messages
- ignores
- ignores
- explicitly track shutdown status
- try to retransmit in case of core error
- don't log old mesh versions as errors
- fix log format
- first log, then free
- if a path is NULL is not equivalent to any other
- GNUNET_CADET_disconnect will destroy pending channels and call the
cleanup callback on them, and cleanup_channel does access peer_map. If peer_map
is destroyed before CADET_disconnect, a use-after-free happens.
- use 32 byte cast
- improved logging of payload traffic
- fix end of kx condition to reception of valid payload only
- don't send KEEPALIVE if the tunnel has traffic queued
- let one keepalive be queued at shutdown
- send payload on AX key receive
- doc
- debug info
Detailed statistics
- log warning
- better log
- code formatting, logging
- cancel retry task if all pending messages are freed by bitfield ACK
- log
- remove hello logging
- ignore AX KX when we already know the peer's key, not when the state is
KEY_OK. This allows the peer to issue a re-key.
- fix uninitialized mem
- warn about leftover CHANNEL_DESTROY in tunnel, do not break
(connectivity error, not implementation)
- refactor GCP_get_peer, fix get peer info
- fix CLI netcat
- remove unnecesary include
- only change from uninitialized to sent, to avoid having the tunnel as
not ready on re-key
- more descriptive tunnel statuses
- cancel pending AX_KX when forced to send a KX by peer, fix #3927
- log
- log
- resolve duplicate connection handling
- mark cancelled task as NULL
- move var declaration to innermost scope
- fix #3928: make sure accessed variables are below size threshold
- use proper printf type, avoid accessing memory out of scope
- reset message counters on rekey
- register conn_destroy on sent_callback
- assert on create
- use channel debug
- debug framing
- use hex channel numbers
- log
- use transmit handles in cadet CLI
- use NBO
- simplify loglevel
- log
- save Header Key for skipped messages
- cancel transmission on shutdown
- minor fixes
- refactor client handling
- unregister from client on destroy
- dump client info on request
- handle case if ports == NULL
- return NULL when handle is NULL
Fix #3929: cancel poll towards disconnected neighbor
- make sure own_id is a correct NULL-terminated string
- format
- fix building (missing source)
- don't call core if there is no traffic
- typo
- unify logging format
- indicate retransmitted packets on conn log
- don't stop looking for connections as long as the tunnel exists
- style, change rounds
- use correct hash type
- compiler warning
- reduce log level
- use proper connection states for a tunnel
- fix logging
- don't leave freed memory on the peer's path list, signal error and
remove anyway
- connect if there are less connections than the minimum, not more
- don't try to connect to peers that are neighbors already
- use better variable names
- never invalidate direct paths, show warning instead
- additional test
- fix log
- fix processing of messages with no payload (keepalives)
- don't connect twice on core_connect, as it may cause problems since
connections hashmap is not yet created
- cancel transmit requests before destroying channels, warn API users if
not
Fixed usage of known broken paths. If a connection was destroyed while
handling a neighbor disconnection, cadet tried to reconnect to peers using
paths containing the disconnecting peer.
- removed false break
- log shutdown
- cancel all transmissions so core_api doesn't complain
- use correct message number in the logs
- typo in variable name
- escape line to avoid false matches
- added hexdump-like function to write binary data to log
Fix try_old_keys, in order to use it for data with the same Header key
but out-of-order message number
Don't reset axolotl state on receipt of a pseudo-duplicate AX key
exchange.
- find the correct stored key, not always the first one
- adjust logging
- remove debug
- only break for cadet errors
- don't show WARNINGs on duplicate traffic
- Revamp traffic logging, small bugfixes, complete doxygen
- refactor test dropping, reduce test load
- remove unused variable
- ignore confirmation of connections that are being destroyed (tunnel
shutting down, for instance)
- fix script to work with barebone versions of sh
- fix use of uninitialized memory
- fix multiple gnunet-nat crashes
- ignores
- only request reply if no key is known
- FTBFS
- ignore
- FTBFS
Always force queue local messages when the client was pre-authorized to
send. Fix #3999.
When cancelling polls cancel pending messages as well, as the message
will trigger a new poll upon sending. Fix #4010.
- add some assertions to make sure the poll mechanism is sane
Fix #4020, wrong assertion: if destroy is different from NO, state MUST
be DESTROYED
Another attampt at fixing #4010: mark unavailable neighbors in
FlowControl struct, test on POLL
- warn when trying to send traffic to disconnected neighbor
- typo, comparison order
- indent, clarify messages
- doxygen
- refactor connection destroy handling, fix #4020
- doxygen
- use proper setter method
- refactor, test orderd
- check return value on send_broken_unknown
- fix condition on connection timeout
- free queued messages on peer destroy, fix #4001
- make autoreconf happy by defining BUILD_EXPERIMENTAL_HELPERS (always to
false, since it's the same as GST anyway)
- ignores
Fix shutdown assertion failure in core_api: avoid double cancel of
pending traffic
Fix printf format causing SEISEGV
- reduce INFO level debug
Change CADET ports from uint32 to HashCode
- adapt test cases / framework to new port system
- open port after handle check
- fix CADET-using services
Port CADET API to MQ
- Add monitor API to MQ
- fix use after free (not removing single transmit handle from queue),
refactor queue logic
- missing implementation
Port CADET to CORE MQ API
- fix PID accounting without direct queue control
-use is_fwd to retrieve traffic direction
- initialize reserved to 0 to avoid valgrind complaints
- add queue control in peer and connection to cancel MQ_env when those
are destroyed, since callbacks could dereference freed memory
- minor changes, TODO, debug message
- cancel maintenance message if connection is confirmed while a
maintenance message is queued
- send connection keepalive if tunnel is unable to send traffic yet
- don't allocate message copies on the heap
- simplify connection-level processing of messages to be sent
Make a copy of the message on connection level only if it needs to be
modified (generated by the Tunnel)
- simplify KX code
- fix AX/KX CID initialization
- don't complain about AX messages pre-KX
- don't send ACK con CREATE if one is queued
- remove dead code
- improve logging of CONN_ACK
- reset timeout for early keepalive on CONN_ACK
- add connection info to timing log
- reduce log level
- doc
- measure encryption performance
Build messages fully in tunnel code, avoid copying in connection layer
- Fix Coverity #153011, abort on NULL path instead of dereferencing later
Un-flag in_flight to avoid integer underflow in GNUNET_MQ_get_length
Refactor encrypted traffic handling
- doc
- sending KX on loopback is very wrong, die
- use proper type for N/A
- don't complain about retransmitted connection_create/_ack messages
- always return a queue handle
- simplify ack handling, die if trying to send conneciton-level ACK to
clients
- fix logging
- refactor function timing code
- update doxygen / doc
- refactor connection upkeep code, remove unneeded functions
- simplify logging
-remove dead code
- use timing macro instead of log
- avoid connection timeout in valgrind tests
- axolotl initialization takes > 1s under valgrind
Bruno Cabral (34):
Fix missing test_connected
Test Commit. Fixing typo.
Add a third default. $HOME is not set on W32, and it refuses to work
without it.
Fix compilation
Don't use ancient -no-cpp-precomp flag on OS X. Straight copy from
WxWidgets : See http://trac.wxwidgets.org/ticket/14029 for a discussion of the
problems it causes when using non-Apple compiler. Even for Apple's GCC, it does
nothing these days and we no longer support OS X 10.3, so there's no point in
continuing to use it.
friend is a reserved work in C++. Rename to friend_id.
Missing extern for C++
Missing separator
Missing Social Service include
Mingw work around
Function signature do not match definition.
gnunet_psyc_util_lib was also missing from install
Fix infinite loop ( grothoff, please check !)
Don't destroy channel if it was already destroyed.
Avoid infinite loop. The loop go on backward, if we add 3 when we see a
double quote, we will get into a infinite loop
Do not crash if key is NULL.
Add fs search test with multiple files using an AND operator
Fix test case
Fix FS search with multiple results for the same keyword
Initial import of STUN support
Stub nat_stun.c
Commit rewrite before moving to GNUNET_NAT functions
Polish and simplyfy STUN code, move STUN code to GNUNET_NAT_
<none>
Get STUN to work with UDP plugin
Disable Stun
Fix #3918
Add missing file
Add test_nat to makefile
Fix crashes and memory leaks.
Forget to commit some files
gnunet-nat, tool to suggest changes to configure based on NAT
configuration
Fix review . PATCH 1/3
Fix review, - Many fixes. - Do not mark test as failed if no connection
PATCH 2/3
Carlo von lynX (45):
DESTDIR fix in Makefiles needed by distro packagers
configure: fix for "yes/lib not found" error
let's call it _notice_place_closing
typo
slowly aligning gnunet_social_service.h with PSYC spec
src/psyc only implements the lower routing layer of PSYC
more psycey method names in test_social.c
make gnunet-social.c build: resolve listen() collision
parsing parameters for gnunet-social.c a bit
cadet.. it's no longer called mesh
cmdline usability for gnunet-social
cmdline usability for gnunet-social
social: minor fixes
what is the proper way to ensure identity service is running?
social needs to start as user service, not system
social: put the sock in the right cupboard
typo in src/identity/gnunet-identity.c
added filters to gnunet-logread and an automation fix for gnunet-arm
taught gnunet-logread to also do what gnunet-logread-ipc used to do
improved usability for gnunet-logread -f
Merge remote-tracking branch 'origin/master'
conversation: non-invasive ringtones
social: fix a few doxygen complaints
social: fix various warnings
social: leaks and untreated error conditions
-unvoluntary edit
gnunet-logread with intelligent usage message and proper perldoc
fixed use after free and several printf warnings
fixed a memleak, a static string free, an access of freed memory etc
cadet: rollback + warning message for posterity
social: allow multiple calls to disconnect(), rest: handle warnings
social: crashing free() removed.. was it redundant?
leaks & warns
multicast: use INFO not WARNING
multicast: don't advertise the origin in the P2P overlay
probably the wrong place to address high-level features
testbed: STAR topology (SCNR)
Merge branch 'master' of /RAM/proj/gnunet
UX: message type for important things that aren't evil
attempt at a .gitignore file for those who git it
end-user friendly output messages, ok like this?
GNUNET_CADET_close_port doesnt work as expected. Bailing out instead.
protect CLI from port scanning: use string ports
ipv6 udp broadcasts are ending up on ipv4
-grammar & style
Christian Fuchs (207):
- added vpn helper stub for win32 builds - modified vpn-makefile
definitions to process gnunet-helper-vpn-windows instead of the regular one
when using mingw as toolchain
added libraries for setupapi & dependencies
Added logics to install virtual networks interfaces for gnunet-vpn in
win32 based on the info that can be found in the MSDN.
added teardown functionality for deconstructing the interface created by
the win32 vpn-helper.
solved fixme's, linked device handling functions to main program flow,
added some fixes here and there. added more comments to the code.
added unique additional hwid entry to allow us to find our individual
virtual interface again. Needed, because netsh requires us to resolve the
devices name as string, which can not be derived off the deviceinfoset
directly. we are now using the handed over devicename + our PID
added functionality to fetch the virtual interface's FriendlyName from
the registry. This information is needed for netsh later on.
grml, friendlyName is not friendly, because it does not help us for
finding anything. Switched over to backup plan: * lookup the devices's name
directly in registry via the PNPInterfaceID. Registry fun, yey! * added a few
fixes here and there
added a copy of the current sources(openvpn-2.1) and a signed copy of the
compiled drivers of tap32, which is used by vpn-helper-windows to emulate
tun/tap on win32.
Added basic functionality for setting IPv4/V6 addresses in win32 using
netsh. Added wrapper for launching programs in a windows shell. Added define
for _tpopen
reworked vpn-helper to now use regular(!) char, instead of wchar,
according to recommendations found in the C/C++-specs(wchar != unicode support,
but may break compiler compatibility).
* added tap version checking logics. Many tap32 versions are broken, Only
version 9.8 and later support IPv6, 9.8 is broken with IPv4. Thus, we are using
the same minium version as openvpn 2.3: tap32-9.9
* re-added mainloop for run * started work on our select-equivalent for
windows.
added initial stuff for overlapped IO and event handling
There are now four states for the OI facilities: * ready (to do work), *
queued (waiting for async-io to return), * waiting (a read-facilities wait for
its output partner to get process) * failed (if a socket error occured)
fixed a wrong conditional
In Windows, pipes, files and the console have to be accessed
differently(in contrast to *nix). overlapping does not exist for console
handles, and different functions are required to read/write/poll stdin/out.
as discussed, vpn-helper-windows will only function if stdin/stdout are
handed down as pipes.
consolidated read and write functions, as we can read from a named pipe
the same way as we can read from our TAP.
final logics for run implemented added teardown functionality.
a bit of bugfixing
more bug fixing
added calls for updating the driver in a for device. seems like this is
mandatory when adding a new device node. unfortunately this is pretty slow...
more debugging work.
* a little bit of bugfixing * added functionality to remove ip addresses
from our interfaces * some formatting work * added rety-loop for the
interface-name resolver (retry up to 30 seconds) * set new ipv4/v6 addresses to
be forgotten on bootup (store=active) * adjusted some debug output (added a few
\n here and there)
added ip-address remove functions
added gnunet-header stripping and adding functionality
added IOSTATE_RESUME and related code to state machines
helper now always uses BOOL(winbool) instead of boolean to stay consistent
added debug logics to print to and read from STDIN, instead of from named
pipes
re-fixed incorrect from last commit... :-D forgot to save.
added nice debug output for gnunet-herlper-vpn-windows.
added gnunet-helper-exit-windows, which currently is just a copy of
gnunet-helper-vpn-windows
some final code-beautification (coding style 4tw!)
helper-vpn and helper-exit now are 64bit capabled.
added support for partial reads from stdin
added support for partial write to stdin/stdout.
re-enumated argv to conform with helper-exit syntax
added my ssh public-key
* replaced no-nat delimiter which was % (reserved char on windows) with -
for exit-helper-windows and exit-helper. * added ipv4 NAPT and removal
functionality to exit-helper-windows * moved a few variables * next todo: ipv6
forwarding rules for netsh
* adde ipv6 forwarding functionality for helper-exit-windows
* fixed incorrect doxygen commant fields in exit and vpn-helper-windows
added 32 and 64bit tapw32 drivers for exit and vpn helpers in windows
* added vpn and exit to the default build options for mingw/windows
builds. This code is considered experimental and has not yet undergone
intensive testing. Use it at your own peril, but we would still be really glad
for feedback and bug reports! (thanks!) * fixed a typo in the exit-makefile
(buildflags) * adjusted output of gnunet-helper-exit-windows to properly
identify itself * updated driver-paths for looking for tap32/tap64
some fixes to the pt/vpn testcase.
started making test_gnunet_vpn a bit less linux-only. does not work on
windows yet, though due to permissions checking, see other commits!
removed raw-pipe checking code for windows. this will always fail new
windows. functionality will be replaced with windows-dryrun checking
added dryrun functionality to helper-vpn-windows
dryrun functionality added, removed irrelevant return code reset
extended the GNUNET_OS_check_helper_binary parameters to do previlege
checking in windows. To do so, tested binaries must still be supplied with
valid commandline arguments, but on windows gnunet will utilize the -d flag to
run the programs initialization phase or privileged operations only. In these
modes, a program will not enter its mainloop or communicate with the outside.
added privilege testing parameters for helper-vpn, exit and nat-client.
todo: nat-server
added privilege testing parameters for nat-server and added parameters to
currently windows-only tests (less work in the future, even if this is ATM
unused)
initial modifications to the set service to support intersection added
set intersection sources based upon union
a few more changes to the lib and the service for intersection
added vectorproduct protocol definitions added to be localized files to
POTFILES added (inactive) changes to the src/Makefile.am added manpage for
vectorproduct added apptype for vectorproduct added vectorproduct service,
client, API and related sources to SVN
adjusted makefile-template for gnunet-mainline toolchain
added vectorproduct's config- and makefile to configure-automake-template
removed texml-command-backslashes from comments which prodced warnings in
doxygen
fixed a couple of doxygen keywords added a missing license-header
merged reworked crypto and a couple of detail changes from the project
repository crypto now works properly, see comments in source
partial rework of the old API, includes simplifications and gets rid of
the extra connect. we now establish one client-connection per session.
fixed a couple of namechanges in comments
updated API definitions as part of the API rework
moved scalar-product protocol definitions to scalarproduct.h moved
service-structures to gnunet-service-scalarproduct.c
added new header for protocol definitions
updated includes of api and service to include the new headerfiles
remmoved deprecated headerfile from scalarproduct updated includes where
appropiate
removed much of the excell logics in the scalar product API finished the
the alice/bob API initiation functions in the SP API merged structes in SP API
reorganized SP bookkeeping of computations
rewrote API minus cancel function
synced API definitions again session now get properly terminated in case
we can no send to the service. implemented/added cancel functionality
re-added collective disconnect command to the scalarproduct API adapted
the scalarproduct client to use the new API modified scalarproduct.h to include
proper logging for gcrypt errors cancel is now quiet added proper gcry logging
to the SP-API
updated scalarproduct service to be compatible with the new mesh API
mini change to the SP-client output
SP-client now directly outputs its result to STDOUT, and ONLY the result
fixed missing backslash that broke buildchain/automake definition
fixed dangling tasks in SP-service removed destroy tunnel function, as it
is no longer needed with new-mesh tasks now get queued in a DLL and are removed
from it upon execution. do_shutdown de-queues all outstanding tasks
service no longer asserts if it can not convert the result of a
computation for the client, but just signals an error errors are signaled as
negative result with zero-length product. Bob determines his computation
succeeded if he gets zero-length and zero as sign
reworked task housekeeping\ added a couple of TODOs introduced due to the
recent logics changes
more work on the state machine. transmit handles are now reset in
do_transmit_msg session remain active/present until the client disconnects or
if a client sends a new message from the same connection
more rework of the logics
more reworking of the service's state machine
... and more work, rework should now be OK, testing
renamed the scalar/vector product manpage
sanitized the scalarproduct client fixed a typo in the SP API changed
SP-service to always use GCRYMPI_FMT_STD when sending the result back to
client. Even though gcry currently can not print negative numbers, we can just
remove the if-else block above it once that's supported
minichange
added propper shutdown task to SP-client fixed a logics mixup in the
SP-API between alice's client and bob's client-function fixed a memory
allocation bug in SP-API adjusted message formats to use uint32_t instead of 16
for all element counts and mask lengths
adjusted integer types in the SP-client
SP-service: converted integer types from 16 to 32bit unsigned whereever
needed
removed a couple of extra checks from the SP-api fixed a silly bug in the
SP-client removed two incorrect frees in the SP-service
the key entry in GNUNET_SCALARPRODUCT_ComputationHandle is now a struct,
not a pointer. re-added scheduler shutdown to SP-client
bugfix, forgot to reset a variable
excluded test cases for SP for now, while it does not fully work yet SP
service now properly checks for half-finished sessions in the tunnel teardown
handler for alice.
SP-client now reads the result from the correct offset (msg vs.
message...) upon shutdown, the server-API no longer crashes the SP-service by
calling the client-disconnect handler with a &client == NULL
fixed a possible free-of-null in the SP-API modified if-clause to check
for the opposite, correct condition
scalar product service now terminates correctly again, an if got into the
wrong loop.
removed old, deprecated SP tests added new test which is only a bash
script and computes the vectorproduct. TODO: proper logging output, add a
second test that kills one service instance
added first part of multipart-message support to SP extended and renamed
properties in existing structs and message formats for multipart support fixed
a memory leak
more work on multipart support
added scalarproduct multipart protocol numbers
towards multipart messages
finished service request multipart message sending
more work towards multipart sending, this time the service response.
added two stubs for the multipart chunk receiving message handlers
work on multipart request receiving
more work on multipart receiving
How much wood... more work on multipart receiving
removed a couple of excess variable declarations
more multipart work
onwards towards victory... more multipart receiving...
couple of summary fixes to the previous changes
made the service more resilient against out of order and simply incorrect
messages messages that arrive out of sync with the state machine are supressed
and logged fixed a couple of little bugs more work on the SP multipart message
support
more changes to support multipart messaging
extended cleanup function to take care of new variables in session
structure better cleanup for multipart handling
simplified SP cleanup a bit added a logics fix
finished multipart message reworking couple of enhancements to formatting
updated half the function comments inlined/removed generate_random_vector, as
it was only called from one source
second part of function comment doxygenization more formatting work
more work on doxygenization for SP
added/updated comments in SP client for doxygenization
more doxygen compliance, should no longer generate warnings, except for
one wired warning which is not in the sourcecode, will talk to christian
grothoff about that
updated test-config template for scalarproduct added declaration of
GNUNET_TESTING_PREFIX to the test script
updated config file for changed name of scalar/vectorproduct
more work on the sp_test
complete testcase template, currently functionally blocked by two issues
in the testbed profiler renamed a couple of cariables
fixed a double-free removed a redundant memcpy free_session_variables
always resets the freed pointers to NULL
finished SP-testcase SP now properly resets a couple of freed values
fixed a double-free fixed a logics bug in handle_service_response
fixed type in SP-test template
fixed PID checking and this the kill feature
added testcase for negative scalarproducts added testcase for
negative-zero scalarproducts
removed final non functional dead leftover code from gaurav... may that
code be forgotten testcases now do better logging together with make check made
testcases a bit more tolerat for really slow systems (5 seconds should suffice
even on older systems) activated new checks in Makefile
readded scalarproduct to the be built again
re-synced set intersection code with union as a starting point added
create for set-intersection
more work on set, commented out the intersection-change in set.c as the
relevant code is not yet done
more work on intersection
dokumentation monster session part 1
fixed guttenbergisch copypaste issue
added missing message type
added bloomfilter message struct renamed set-union specific message types
added set-intersection specific message type added intersection cancel API
implementation added intersection create API implementation added intersection
destroy API implementation added intersection peer_disconnect API
implementation added intersection accept API imlementation stub added many
prototypes for intersection operation handlers
workwork
added initial iterators for alice and bob to create their mutated
bloomfilters and set-versions accordingly
added functionality to send a BF BF size is now sizeof(GNUNET_hashcode)
updated some more dox
added many comments documented set-union state "machine" removed many
deprecated struct-members from state, ops-spec improved doxygen-compliant
added new msg type more work on intersection
more work on intersection
more work on set intersection
work work
fixed bug from set-union more work on set intersection, should now be
logically complete except for the adaptive bloomfilter size
added intersection to set's automake script renamed shared variable in
set-service's common structure reordered code in intersection i assume:
finished reworking intersection, still misses dynamic bloomfilter size re-added
struct SetState to intersection, which keeps track of the newest (and only
this!) element count in the set's hashmap. removed iterator over the set for
counting (see line above) a bugfix here and there
towards multipart msging for bloomfilters in set intersection
- more multipart fun - Make it so
- jummy, multiparts! - Hatchet
- more work on multipart sending - separated bf-generation and reduction
again, as this is now necessary with dynamic bf-lengths
- finished work on multipart sending in intersection - added getter for
addressesPerElement from a bloomfilter - added multipart message format for
intersection - added multipart message type for intersection
- begin work on enhanced multipart receiving - added msg handlers for
multipart bfs
- further work on multipart receiving - removed the multipart-state from
the statemachine again, as we can recognize multipart sending based on wether
or not the bf_data pointer is null or not - simplified & refactored the
multipart message format a bit
- added early success for fully disjoint intersection sets - some minor
bugfixing
- implemented output mode GNUNET_SET_RESULT_REMOVED in intersection
- converted peeridentity from hashcode to eddsa
- added extra parameter which i forgot in the previous commit
- first extension of paillier with foolproof homomorphism support
(ciphertext container carries information on support ops) - adjusted
encryt/decrypt functionality to handle gcry_mpi_t, not a plaintext container
- header sync
- scalarproduct element container was 1 byte too large - added logics to
compute the maximum supported number of paillier hom.ops for a ciphertext -
updated comments
- added logics for homomorphic operation in paillier - adjusted headers
- more work on crypto-paillier
- added testcase for set intersection
- activated testcase for set intersection, as it not works correctly -
some bugfixes for intersection - corrected a logics bug in intersection
- wrote a proper test for crypto-paillier
- omitted test for crypto_paillier for now...
- re-added testcase for crypto-paillier - fixed remaining pailier-test
bug - now: achieve world domination or off to new shores
- formatting - forgot to reduce the possible homomorphic opts by one, as
a badly chosen n might sometimes corrupt the final operation using a simple
check. As the overall calculation of possible hom. opts is pessimist anyway and
might become rather inaccurate if many operations are performed, reducing the
prediction by one is acceptable.
- purged leftover plaintext container for pailier.
- added more information on when paillier-encrypt soft-fails (too long
number, no homomorphic ops possible)
- begun migrating scalarproduct to use crypto_paillier and set
intersection - removed paillier from the regular build for now - partially
replaced pailier-encryption functionality in scalarproduct with the new API
- more partially replacing of pailier-encryption functionality in
scalarproduct with the new API
- extended paillier-api to also include caller-suggested maximum
supported homs
- added caller-hom-ops soft-cap logics
- removed custom crypto functionality in scalarproduct - scalarproduct
now uses the paillier-API + associated structs - omitted SP-tests for now,
scalarproduct is still heavily being reworked
WIP commit of scalar product 2.0. It is unfinished and does not yet pass
tests. This commit happens because of the move from mesh->cadet.
- sessions now terminate correctly for both sides if invalid/unexpected
msgs are received
- more abnormal session-termination handling
- session->remote_pubkey is not kept independently from the service
structure
- added status field to SP result messages
- proper service-side error reporting for the SP API
- added proper error reporting to the service - some refactoring -
bugfixing
- updated error reporting functionality for the service - client side
initialization adapted
- updated SP testcases to comply with the new key/value input format
- removed some optional parts from the SP test config - fixed two bugs in
the API, where the API would only append number-of-elements bytes instead of
the size of those elements - SP-service now retains the configuration handle
received during initialization
- updated NAT callback function defintion to no longer carry around
strings, - NAT lib now uses error-code enums to signal success and errors -
reworked related code to return errorcodes - added and updated a missing
doxygen documention - todo next: testcases
- adapted nat-tests for API changes - set NAT success return code to be
GNUNET_OK
- removed redundant and bugged check from SP intersection handling
- removed now obsolete check for the element count for an scalarproduct
calculation - removed matching for element count for the same reason - fixed a
safety check conditions logics operator from AND to OR
- fixed corner case for the set-intersection computation formula for
computation pair calculations with VERY few elements on one side
- forgot to set concluded intersection OP to NULL
- SP now uses the correct msg type for session initialization - fixed bug
when handling of failed set operations
- fixed bug in set which allows a peer to generate a malformed session
setup - fixed double-cancel of a set-handle
- removed double-free after set has freed its set already
- fixed a double-free-race upon failing a set op
- fixed the same bug on bob's side as on alice's side in copying over
element data in the SP-api. seems like this correction got lost somewhere
- no hton conversion for the set-API-add function
- added many new error codes for nat auto configuration, nat_mini and
nat_test
- added many lots of new error checking to nat_auto and nat_mini.
icmp_client server and client not contained in this commit.
- disentangled test_icmp_client and test_icmp_server's logics -
integrated more error-reporting changes for nat_test.c
- moved timeout handling responsibility from for nat tests from caller to
the library - updated NAT API accordingly - removed obsolete timeout handling
code from gnunet-transport - updated result callback for nat-test
gnunet-transport to not always just report success, but we now actually report
the proper result code we got from the NAT API - updated nat tests
- more scalarproduct bugfixing - added some more useful debug output to
scalarproduct
- added missing comma in nat-test
- renamed NAT's failure code to status code - updated status code
descriptions - added NAT library function to convert status codes to strings
- renamed NAT's failure code to status code across gnunet (not GTK yet) -
added proper error reporting to NAT-test-start/stop - timed-out nat tests not
automatically stop the test itself - fixed a design flaw in nat-test - fixed a
potential crash in the tests of the nat functionality (NOT NAT_TEST_xy...)
- added more debug messages to status2string - added remaining
functionality to GNUNET_NAT_test_stop
Christian Grothoff (10345):
ng
mx
ignores
ignore
removing server from argument list, other minor fixes
more ignores
fixes
better
debug code
nicer log messages
add possibility of passing extra options via configuration
fixes
keep addr
various bugfixes
dbgonly
dbg msgs
fixes
nolog
sn
reg
reg
reg
check for target repetition
bound queue size, clean up code
reg
fix
bound client queue size
fix
code cleanup
log
fixes
reg
reg
timeout code
syn
syn
reg
done
reg
reg
reg
reg
reg
reg
reg
reg
improving script
reg
Regression in tests on tripwire
script
Regression in tests on tripwire
reg
die
moved
API improvements
adding bandwidth management
API clean up
get time cfg api added
notes
removing dead code / unused variables
fix
fixing pid issues
fixing waitpid usage issues
more arm, waidpid fixes
client part
server draft
use external select
cancel task
first hack at topology -- incomplete
more topology implementation
implemented advertising
question
remark
move to minor
add
bld fixes
cov-data
better support for daemons
adding basic declarations for datastore implementation
importing older testcases / benchmarks, not yet adapted to new API
stuff
mtypes
h
improving datastore API
further datastore API improvements:
added bloomfilter
fixing testcases with new disk API
towards compiling tests
making tests compile
datastore
ignore
testing-design
fix
syn
stuff
airplane hackery
dsfixes
log
tolerate funky IPv4/Ipv6 localhost configs and getaddr returning only
IPv6 addresses for af-unspec
fixing missing address copy
fixing segv under linux due to unused w32-specific handle being
dereferenced on unmap
option section name changed
cleaning up disk api
fixing issues with hello handling
fixing testcases to match new api
fix
fixing types
nitpicks
forward-porting heap bugfixes
nicer debug code
notes_and_debugging
use libgcrypt entropy progress callback to fork a find to hopefully
generate entropy
fix
fixing find forking
fixes
API refinement
restrict to root fs
some documentation and bugfix in testcase
another fix
nodeb
api work
more API work
api
improving HELLO validation by transports
docprog
fx
transport fixes, hello API extension
fixes
disable
fixes
done
sh
diags
debugging
fix
less debug output
fix
update
simplify
debugging
terminate with error if message fails to come back instead of hanging
notes
bugfixes
docu
fix
seems to work
leak
leakfix
notes
bugfixes
leakfix
fixes
fixing compilation problems
repeat
stuff
importing Sailor's code
indentation
updating code to include first actual test
updating code to include first actual test
debug
bugfix
SN
notfixed
typo
typo
typo
sthuff
better
cpsing perf test
ign
fixes
stuff
implementing all-now iterator
done
debugstuff
64
print
test
sn
argh
work on datastore API implementation
finished first draft of implementation of datastore API
debug messages
bugfixes
bugfixes
dgbg
bugfixes
passes
termiante
quiter
stuff
reservations
syn
fix
quota management and better name for NO_TASK' '
fx
more tests
update bf
fix compile errors
done
dcl
done
towards datacache implementation
making plugin compile
done
docu
fixing bugs, adding testcases
done
quotatest
const-ing of config-handles
fixes
fixes
updated testing API
DHT and datacache API refinements
revising TODO structure
ignores
reorging work list
testing-update
die
compile-it
renaming for consistency
mktemp
working on testing code
more testing code
drft
update
travelhacking
hxing
more work on FS URIs
move-hacknig
fixes
stuff
hxing
hxing
hxing
syn
stuff
fix-memchr
more naming fixes
fixing remaining network-connection naming confusion in API
more naming stuff
fix
renames
more fixes
more fixes
fix check for close
indx
ignore
ignore
eta
more
event generation
finish publish shutdown code
finish publish shutdown code
todo
kblocks
sblocks
sblocks
updating TODO
clarification
eliminate warning
adding convenience API
adding indexing support
adding listing of indexed files
notes
working on unindex
refactoring publishing code
fixing fixmes
fixing fixmes
more
finishing first run at unindex
fixing datastore API, improving unindex
done
towards search
fixing warnings
fx
fixing types
fixes
mantis 1492
seach
cleaner crypto api
towards having download
towards having download
towards having download
towards having download
work on fs binaries
fixes
work on downloading
fs tools
fixes
better clean up
fixes
don't fail
fix
make gcc happy
disable
prep work
stuff
dead
helping llvm static analyzer a bit
fixes
more testing
fs
fixing docu issues
fixing compile errors
docu
better datastore API
eliminating not-used vector code
more FS coding
more work on fs
handling of unindex
done
not polaris
fs hacking
fix
fix
fix
td
more work on fs
move
stuff
more work on fs
heap fixes
types
block validation
more clean up code
update
documentation cleanup
useless
do not fail
adding missing function
typo
use static' '
improving style and docs
more work on fs service
fix
clean up afterwards
clean up afterwards
conf
fs hacking
fixing types
fix
fixing comments and minor bugs
bugfix
some doxygen fixes
fix
better diagnostics
code cleanup and m4 reverting
do not go via network to lookup loopback
moving resolver to util, making DNS lookups asynchronous in util
lnk
better ARM test script
fix
no-verb
fixing off by 1000 in select code
fixing timeout handling
wrong place for assertion
fix
add warning against busy waiting
add warning against busy waiting and mitigate
need host
redundant
fix
nospace
better
assert
support any numeric address
make verbose for buildbots
detect hanging select
absmin
clean up, more error reporting
misc bugfixes
done
fixes
more assertions, etc.
fixing leak
fixing inefficency
rename
need to sleep here
fixing slist api and implementation
use enum
debug
fix v6 test
fix types
style
undo
stuff
ps is not good enough to use for testing
sh not bash
fixing arm start
use arm instead of bash-ish kill
backup errno
more debug support
bugfix
silliness
fix
fix
improving ARM API
ignore
documenting
fix
better
dne too fast
sh
fix
bsdism
be less verbose
fix
fixing leak
code clean up
fix
fix
improving documentation
use mktemp
fix
check for FD_SETSIZE
fix code for certain types of /etc/hosts files
better comments
fix
fix
ignore
document params properly:
delete temporary files / clean up properly
fix
making flags, set non-blocking and set-inheritable internal issues of
network.c
use nodelay for a MAJOR performance boost for GNUnet code
fix
ignore testcase binary
minor consistency improvements to FS API and another testcase that
compiles but does not yet run
fixing config file
fix
fix
more consistent error message language and format
defservs
more work on bugs and testing of fs service
updating dep list
removing bootstrap-generated m4 files
updating ignore lists
fixes
use pipe instead of racy signal
use vfork if available, otherwise sleep after fork to avoid racing on
signalling the process vs execvp changing the signal handlers; clone without
signal-handlers would also do, but is even less portable
debug messages, disable default services for testing
vfork
no defaults
no defaults
fix
add assertions
leak
cleaner
leak
unload again
fix
leak
fix
fixes
fix (bad argument order), plus some more debug statements for now
debug code and fixes
fixing bugs, diagnosis summary
fix
fixing bio bugs
fix log calls
fix bugs and add comments
formatting
remove get_random from API
fix comment
dead
fix
do not report as critical
final fixes, adding testcase to official suite
fix
bug fixes, api fixes, extended testcase
done
draft for search testcase
cleaner
debug, maybe will help diagnose crash on ndurner gtwy
debug, maybe will help diagnose crash on ndurner gtwy
check return codes everywhere
bugfixes, including double frees
minor bugfixes
fixing leak
minor fixes
extra debug for datastore, stats api fix
fix
need larger quota
fix
fix
need other preps first
hacking on fs
use mktmp
report diagnostics
check-early
report
check flag
giving client API option for auto-retry, making more often use of
transmit_and_get_response API and removing auto-retry code where enhanced
client API can be used for it
fix
do propper delaying of notification past return
make compile without warning
better tmp dir name'
can now be NULL, check
fix
make sure stats that are being set can be committed on destroy
better
need to free current
fix
dead assignment
now exists
ignore
testcase for unindex
more consistent disk API
API fix
API fix
create dir if it does not exist
should report success here
more debug msgs
nodebug
fixing indexing and unindexing issues
noverb
done
exactly wrong
bad type, no cookie
fix
additional fs tests or fs test harnesses
better type
fixes
ignores
fix
fix
fix
fix
update
api
remove d everywhere
fix
report errors
check here as well
delete correct dir
delete correct dir
ignore test binary
allowing cancellation of resolver requests
making use of resolver cancellation to avoid callback into unloaded plugin
testing not yet ready for testing
update
getting testing to work
fix
only for headers
no cpp stuff here
fix
dup to prevent access after free:
inline
check for connect failure
testcase is broken, but claims to pass with a ton of warnings
remove
exclude tests
bugfixes
not yet ready
cloning cfg to make client shutdown function useful
adding code to ensure clean up of solicit state
tolerate empty message
wtf
remove temp files, indentation
kill pending transmit tasks on destroy
various fixes
do not do it twice
doku
nodbg
ignore
group test works
warn for unimplemented functions
update status
typo
fixing bio testcase and a bug in bio.c, also indenting
trick automake/libtool into NOT linking with g++ on non-W32 systems --
hopefully without breaking W32 build
check for NULL
missing config values:
basic topology test
adding some debug messages
fix
does not yet pass
docu
cleaner cleanup of DNS request code
bugfixes and redesigning scheduler API
immediately kill, then wait
send confirmation if kill request is pending
clean up for proper shutdown
ignores
fix
fixing core API for early shutdown handling
removing old comments
fixes
to be fixed by jilu:
fixing leak
misc. bugfixes and API improvements
clean shutdown
mark 2nd maint properly
fmt
update
reducing leaks, fixing shutdown bug, increasing timeouts
fix
fix
fix
formatting
indenting
no void in definitions
style and memory leaks
fixing more bugs in test_bio, largely leaks
code clean up
fixing out of bounds in testcase
indent
code clean up, rolling back mods to test_configuration
const
fixing compile errors
update
allow any function in a task to easily get the reason code
do not auto-reconnect during shutdown:
do not auto-reconnect during shutdown:
do not log if response was missed due to shutdown of our own process
use exponential back-off of 1ms to 1s for (connect) retries by clients
no need to sleep, we will re-try anyway
no need to sleep anymore
get rid of (most) unnecessary sleeps
not needed
not needed
update
ignore
either done or moved to Mantis
need break
check properly for mmap failures
fixing use-after-free
add assertion
add now
use add now instead of running immediately, which may free th and have
some odd effects
use add_now
increment RC to ensure that process_client_buffer does not destroy by
chance
fix
fixing doxygen and style issues in NAT code
better debug messages
fix
cleaning up
fix
set shutdown flag also for ready tasks
fixing doxygen comments
adding a few more asserts to help debug
fixing logging issues
done
doxygen-fixes
fix
ignore shutdown support for client/connection and shutdown fix in
connection
ignore shutdown support for client/connection and shutdown fix in
connection
clean up address probes during destroy as well
add shutdown-ignore option to server
better shutdown, error message formatting fixed
minor fix and better debug/error messages
cleaner handling of clients and service processes, ignore shutdown
do not use sleep, use scheduler
be less verbose
has been fixed
be less verbose
increase attempt count
increase timeout
fixed already
fixed already
bugfix
porting fix for 1467 to ng
fix
more
fix
proper error handling
more error handling
improving comments
formatting
existing code does not work on freebsd because there reading directory
files directly is fine, also fixing buffer size issues
work on DARWIN as well
nicer
why is const back? this must not be used
dead
fixing warnings, better structure
fixing bug in bio.c triggered by test_pseudonym.c
code cleanup
adding minor FIXME
fixing test bio
fix type
fixes
fixes
leak
fixing cosmetics and inlining of configurations
fix
stuff for jilu
support http redirect
fix
clean up ARM handling of child death, less busy waiting, faster response
times
fix
fix
fix
fix
fix
working on gnunet-directory
fixing compile errors
unused
fix crash
updating heap API and implementation
fix
fix
adapting main code base to work with libextractor 0.6 API; testcases will
follow later
porting testcases to LE 0.6 API, fixing issues with the ported code
making fs tests compile with LE 0.6 API
fixing comment
working on hostlist
add our own hello to peerinfo DB
support debugging
complete startup / shutdown code
clean up code fixes
formatting
fixing hostlist bugs
nolog
update
adding notification API to peerinfo
fix hang, use notifications
work if HELLO msg pointer is NULL
debugging, found a few more crashes with the hostlist test
fixing documentation
fix
fixes
fixes
fix
update
validate target
simplifying transport plugin API
allow cancellation of peerinfo iterations
further simplify API
adjust code to new peerinfo API
fix
fix
making code work better with dual-stack, preparing for triple-stack
fixes
fix
fix warning
further simplify transport API
make picky compilers happy
fixes
update
making peer_configure API cancellable
fix warnings
dead
peerinfo fixes
fix
ignore
fixing fixme
update
fix
cleaner shutdown
cleaner code
cleaning up topology code:
docu
fx
init
misc fixes
simplify
fix
clenaer
fix
fix
assert
fixed
fix
fixing core API issues
breaking stuff
making tcp plugin code compile again, but does not work yet
TCP plugin that might actually implement the new API done
no more TCP PING/PONG
fixing non-transport compilation issues
blacklisting API
blacklisting API implementation
ignores
stuff
fx
comment
api
stuff
fixing code to match new API
fixing code to match new API
stuff
fix
fix
fix
fix compile errors
indent
fixes
implementing get_hello_cancel
stuff
fix
fixing compilation issues
stuff
towards a new conveniance API
finishing nc implementation
drop
count
docu
finishing blacklist implementation
updates to peerinfo to use new nc API
stuff
fix
fix
udpate
set last_distance
use new APIs for connect/disconnect
stuff
fix
script to make buildbots more robust
disable prefix
API change for Safey/ARM
adding support for outbound message monitoring
adding support for outbound message monitoring
fix
nopref
add and update nate
moving indexing out of main file
stuff
support removing of current entry by callback
finally compiles again
fix
fixing drq code
fixes
die nicely
fixes
fixes
fixing fs FIXMEs
update
unused var
fixing test service running to retry if service is not running
cleaning up code from debug session
add-to-suite
fixes
code cleanup, fixing leak
fix
fixing leaks
fixes
gns draft
fixes
client
client
fix
fix
fix
fix
fixed
fixes
cleanup
fixes
indent
simplify
move
bugfix
does not compile on BSD and does not work yet, remove froms std build:
move log statement
stuff
SUID comments, review
flush
more
debug stuff
fixing shutdown issue uncovered by ARM test
make sure we run ALL shutdown tasks before resuming execution of new
'normal' tasks
assertion now wrong
remove debug stuff, make compilers happier
nodebug
fix
ignores
fix
cleaning up server code, subtle fixes
fixing the fix
fixing types and comment styles and struct alignments
fix test
fixfix
fix
permit NULL addrgen
fixing non-terminating loop
also match tname
min
max
debug stuff
towards better working transport code
reducing allocations
code cleanup
fixing session lookup
docu
reduce debug logging
no grothoff, no debug
found assertion problem in core
status update
revalidate before expiration
dead
block it
avoid linked list traversal, more code clean up
stuff
making things nice by breaking tons
fixes
fixing bad assumptions
update
clean up
done
ign
fix
fix
cleanup
disco
fix
fixes
fix
fix
leaks
timeout
stuff
stop
fixes
fix NPE, fixing HELLO timeout increment on HELLO transmission failure
fix state machine update
no log, no dummy
no dummy
nodeb
nodebug, test topo
tolerate client connect failure better:
update
use DLL macros
only if connected
nicer logging, activating testcase
nicer msgs
prog
fx
fixes, cleaner code
update
update
fixing doxygen warnings
starting library for FS multipeer testing
done with first draft of FS test library
ss
test
fix
fix
logs
debugging
fix
fixes
fix
allow remove of current entry
better cleanup
fix leak
fix leak, rename
stuff
fix cleanup
fix leaks, code cleanup
one more leak
ignore
fix
fix
fix
fixing leaks, use after free
fixing leaks
dead
dead
debug fun
rename
fix
fix
fix endianess conversion
nodeb
nodeb
bugfixes, extra assertions
fix
fix
doxygen
allow any for search
novalg
fix
reorder
fix
fx
note
fixing quota calculations, send initial quota with PONG
dbg
fixes
fix logging
fix quota calculation
log with pid
more logging
comment
fixing API issue of who is responsible for quota in
calc update
avoid tracking requests twice if multiple blocks in the same file are
identical
support multi-message encapsulations
debugging
use utils rand functions
dce
dce
bugfix
dce
legacy code removal
fix
doc
code clean up
fmt
init
check stat return value
code clean up
handle NULL return, fix leak
fix
clean up
nobuildfrag
bwlib
update-api
trying to fix bandwidth calculations
update
fast quota update
nicer logging
fixing corking
notify transport about bw quota changes -- always
do not double-schedule
nodeb
nodeb
db
make llvm happy
dce
fix
nodeb
drq clean up
wrong if
make cov happy
relax
nolog
OR
longer timeout, disconnect on zero quota
update
keepalive PINGs
fix
fix
bootstrapx
smaller timeout
defs
support GLOBAL_POSTFIX
cleaning up arm wrt processing options and prefixes
allow -l for daemons
enable $ expansion for GLOBAL_POSTFIX
fixing gnunet-publish for files
update
s
fixing shutdown
fixing shutdown
fix
nicer
fix
fix progress reporting
ETA calculations, filename
done
done
doxwarn
debg
dbg
update
mkdir
enable stats get cancellation
cancel pending write if we are just dead anyway
init
use
keep until server done call
dllfix
handle being already disconnected
disable
removing ps, does not work
only notify when desired
proper abort
better clean up
fixes
want
add stats
more stats
more stats
stats
formatting
nicer logging
update
more stats
more stats
update
fix
better defaults, nicer error msgs:
fmt
fix
more stats
fix
increase some timeouts, decrease others
copy component name for bulk due to possibility of plugin unloading
only start peerinfo processing after core init has finished
assert
fixing shutdown timeout issue
expand ~ in PATHS
expand ~ in log file name
more stats
more stats
logging
flush
fix
fixes
caps
logging
nonpers
stats
stats
stats
stats
stats
stuff
stuff
fix
misc
trying to fix revalidation PING and connect core notification
fix
fix
fix
extra checks, shorter timeout
more stats
debugging
dbg
crazy work
fix
fixes
fix
fix
fix
fix
fix
move
fix
fix
update
fix
update
fix
address resolution API implementation
stuff
link with stats
fix for 32-bit size_t negation
fix
tolerate NULL handle
fixing disconnect handling, making TCP plugin ready for bi-di use
better session selection
bi-di, no multicast PONG if reliable channel exists
fix
moving gnunet-peerinfo.c to allow it to use transport_api
done
fix
ignore
done
stuff
fix
better
const
done
check for NULL
fix
dbg
stuff
only for ksk
fix
const
trying to fix bug
clean up
consider corking
more gentle enforcement
violat
fix
better error handling
working on fs, bugfixes
fixes
stuff
work on namespace stuff
work on namespace stuff
fs hxing
stuff
style
fixes
back
making it compile
bug
fix
fix
fixes
dead
fix full data extraction for binary data
fixfix
more on directory handling
fix
towards suggesting filenames
fix crash
space
misc fixes
update
fix
planning
consistency
fix datastore quota management -- especially if stats is not working
fix
fix
faster
make more robust to disconnect
bugfixes, more tests
gpl
comments
update
pybld
fix
fix
run py script with config
fix
verb
verb
done
typo
nblock support done
update
dox
fix
fix
undo
warn about bad option combinations
nostop
report
additional tests
done
perfmfix
config per script:
cleaning
ports
be less verbose
unused-dead
bwtracking
details
rounding
better quota violation handling
stuff
ignore
dir needs cfg too
doxy
fs hackery
dead
make sure key stays valid for iterator even after entry is removed
fix
nodebug
mod
mod
done
do not use path
breaking DHT code
env
env
fixing compiler warning
name
cleanup
fix
done
work on HELLO
notes
fix refresh
fix
clarify
better
fix shutdown
update
done
fix
setkeyfix
logfix
fix
logging
proper PONG validation
stuff
fix
fix
transport API changes in preparation for the storm
transport API changes
notify
fix
fix
work in progress
work in progress
work in progress
maybe done with transport inbound-NAT verification and connect signalling
fixed
maybe done with transport inbound-NAT verification and connect signalling
fixed
fix
fix
persistence
logging
fix ttl underflow check
cleaning
fix
fs fix
fix
fix
fix
fix
notest
dist ready
make dist related fixes
more
checksrcipts
clog
back in post-release
fix
handle NULL better
done
cleaning up bloomfilter calculation code
peerinfo API redesign
change process priority based on scheduling priority
syn
fix
syn
debug
seems to work now
prioritize
working on topology
proper HELLO merging
use back-off and higher-frequency re-tries to start with for creating
connections
towards fixing blacklisting APIs and implementation
nicer error handling
fix
fix
fixing shutdown
fix
making parser more robust
use usr bin, env is not portable enough either
bsd is broken
towards implementing blacklisting
adding blacklisting code, untested
dead
fmt
done
move
skel
towards new peerinfo API
docu
doxygen fixes
new API implementation done
migrating testcase
activating new peerinfo API
create hostlist client if we have no MHD
make checker happy
for cov
make weakness more explicit
ignore addresses of other peers that are equal to our own
make clang static analysis happy
resume and less debug crap
fix
adding temp_filename
fix
code
testcase for gnunet-peerinfo
doxyconfig
peerinfo
off-line hack fest
blocklib
ignores
sleep
define loopback more broadly
stuff
fully implement in-line files in directories
towards job queuing
fix
bounded parallelism
docu
merge meta data for search results
store root URI in metadata
stats for datacache
adding availability probes for search results
one
added option to restrict search to local-only
doxies
nosvn
fixing warnings on OS X
first work on serialization
mark open issues
fix
fixfix
making py-tests work with amatus changes in fs
fix
cleaning up FileInformation struct
fi deserialization
details on missing features
fix
unversion
towards fi synchronization
fix
towards serialization
bad commie
publish serialization stuff
finishing initial coding of fs-publish serialization/deserialization
fixes
towards unindex deserialization
deserialization of unindex operations
ignore generated files
unindex syncing
todo update
search stuff
fixing missing free
fixing double free, missing free, use of undef key
fixing major issue with how IP addresses go over the network (previously
ill-defined) -- thanks amatus
reducing address size in hello to 16 bit
fix
fix
doxygen
inet_ntop
search deserialization
fix
fix
search syncing
more fs stuff
download deserialization and not keeping a file handle open per active
download
download serialization code
time subtract
fixing start time serialization
allow linking of downloads to search results
towards suspend
deserialize search link to download
generating SUSPEND events
calling sync, cleaning up files
code cleanup
code cleanup
cleaning up serialization pdir structure mess
cleaning up serialization pdir structure mess
doxygen
fix bus error
free active on download error
fixing mess with search update serialization and parenting
fixes
fixes
tests
test if dirs exist, create dirs if needed
fix
bugs
notes
init top on deserialization
dbg
igns
breaks and bugfix
dsh handling
dbg
bugfixes
better
note
add
removing debug code
fixing publish persist test
dox
fixes
update
fix
fixed
allow file hashing cancellation -- and make use of it
fix unindex persistence
better
complain louder
fix
fix
stuff
nicer error msg
fix
fix
make error more robust
typo
stuff
arm fix
fixes
stuff
cleanup
fixes
stuff
fix mockup
arm-specific
moving code where it belongs:
fix
fixes
fix
stuff
use IPV6_ONLY when binding sockets
ignores
fixing configs
dbg
ignores
no more allow shutdown
asserts
clarify
#ifdef debug code
info
nicer-log
die unstoppable multi api, die
year
update
update
minor fixes
fix
fix
clean up files
1
ds
stuff
dce
adding support for request queueing to datastore API
move
done with ds
fix defaultserv
do not use default services for things that ARM will start automatically
ignores
fix
fix
add cancellation
fixes
protostruct
assert
fixes
no log on shutdown
fixes
nodht
fixes
nodebg
nodbg
stuff
log
misc fixes
doc
debg
update
dox
nodbg
stuff
towards migration
work on migration
update
lc stuff
fix
fix
issue
nodbg
fix quota calculations
use v2
ass
check version
emsg
fix testing
nodbg
stuff
dead
fixes, comments
fixes
fix
better
fix
check
fixing leak, reducing amount of allocation
done
shutdown fix
fix
reduce reconnects
fixing reconnect issues
breaking FS big time, no idea why
fix
fix
abort start task on shutdown of connection
stuff
use DLL macros
leak
cleaner timeout handling
config fixes, do not start core, datastore, fs
config fixes, do not start datastore, fs
dox
newline
config fixes, do not start datastore, fs
update
fix
properly print v6 addrs
do not use link local addresses
do not queue SETKEY messages twice
fixes
EPIPE is normal
ndbg
ndbg
ndbg
ndbg
ecores
dbg
dbg
noverb
noverb
fix compile error on FBSD
nitpicks
nitpicks
more FBSD fixes
do not leak peerinfo
cleaner handling of gnunet-peerinfo results
testing
could theoretically be NULL, check
update
time tracking
formatting
handle errors better
fix bad define
reduce timeout
space
better messages
reduce work for processing duplicate HELLOs for validation, allow
receiving batch of messages for transmission
nicer logging
nicer logging
check that peers match, fix docu
send SET_KEY and PING message together to transport, nicer handling of
SET_KEY duplication
logging
noverb
fixing handling of transmit errors
better clean up
fix
check
renaming files to fit conventions
nicer error handling
nicer logging
fixing testcase: make sure content is not expired already
fix
keep /proc/stat FD closed unless it is actually needed
no need to link against arm
no need to link against libgnunetarm
indexing test
change unindex processing order to avoid transient dangling DB entries
harness
adding test for content migration
api for controlled churn
fixing mig test
nicer logging
logging, minor fixes
note
info
notes
testcase
note
done
test still needs LE
nicer
nicer
type
logm
nicer logging
die
fixes
extradist
logging
report port numbers
ign
fix
fix
reliability test for core
update
fixing bug with queue management in server nc -- and some additional
debug messages
more constants
logging and proper limiting of message size
use larger queue, proper handling of messages with short transmission
time, more logging
short timeout, transmission based on receiving to avoid reordering and
overflowing queues
away
done for now
make smaller -- for now
update
update
fixing Mantis 1564
fixing memory leak
fixing bad control flow
fix
decap
fix
better timeout, renew on significant progress means shorter timeout on
hard(er) errors and longer timeout on slow machines
add native setting capability
done
done
done
UNIX domain socket support
fix
install all of our headers, but not winproc unless on W32
not
check for headers at least conditionally
check for headers at least conditionally, also for MinGW
duplicate event type
have rank
use extractor
use extractor
fmt
more data
stuff
fix
done
fix
fixing bias
stdconst
fix
mykeys
fix overflow
fix double free
dbz
dbz
size
use after free
simplify
fixing leak
leak
error checking
nicer
extra semi
use constants instead of casting -1
use constants instead of casting -1
better type
minor fixes
minor fixes
dead
pass only unsigned char to isspace and the like
fixes
indentation
free
leak
fix
nodebug
indenting
indenting
fix
slooow bots
is dir test
pass do_index
need clear api
done
fix
fixes
fixes
fixes
fix
fix
logging
logging
check
cleaner
localization of scopes
code clean up
code clean up
cleaning
hmac
cleanup
cleanup
cleaning
not needed for good style
check len
publish start implementation
pc not sc
publish start, progress, complete handling
publish stop, suspend handling
kind-of works
ignor
style
fix
clean up
redundant
gcc
fix
nate on code sonar crack
yay spagetti
refine API to make it usable for server itself
mkdir
reorder
code review edits
mst fixes
dos2unix
server API clean up
make it easy to start service by hand
fix
cleaning up server.c code and fixing mst bugs
try fixing mkdirp macro
add mkinstalldir script
mkdirp-causes-trouble
unlink UNIX domain socket filenames
do re-try even for shutdown to make this reliable
removing testcases that fail too often for now from standard run
mst api fix
use mst API
ignore
support for systemd style listen fd passing
close pipes on exec
cleaning
fix scanf
listen socket passing support for ARM
doxy
diag
fix
check
do not bind if port is zero
nodebg
hacking on fs features
new transport address validation code -- see Mantis 1556
add option to disable socket forwarding -- useful for debugging
do not store duplicates in datastore
cleaning argz mess
check
leak
cleaner
cleaner
clenaer
fix
fixing common off-by-one error with respect to maximum message size
fixing common off-by-one error with respect to maximum message size
marking performance issues
scheduler optimizing
scheduler optimizing
fix
fix handling of search stop with active download
communicate download activity status with each event
return if directory is well-formed
tolerate md==NULL
properly check for gnd files
tolerate NULL mime type
support check for existing URI
investigation
eliminating testbed API that was never implemented and is not needed
fixes to plugin API for DV
implementing local address check for tcp/udp
fix
vpn build integration
no generated files
missing
fix
ignores
new tst
stats fixes
fixing 'resolver' double free issue
fixing #1561
unlink on bind
only run test if we can resolve our own hostname
slightly more moderate timeout
dbg-only
fixing #1566
dbg-only
trying to port statvfs call to BSD
trying to port statvfs call to BSD
trying to port statvfs call to BSD
logging
reduce to more realistic size
use constant, larger timeout
sparcbot needs really large timeout
ignore
deactivating testcases that have never worked
gpl3
gpl3
leak
redundant cases
fix
noinst
rename fest
fix
adding tests for openssl and making sure that emstp is available when
building monkey
nocpp
fix
fmt
fixing calculations and code for buffer realloc
use macro
fixing issue with sqlite not supporting unsigned 64-bit numbers
fix
do not run incomplete tests
indent
update
nicer build system
fixing off-by-one
nicer
no debug
fixing various dist issues
release
fix bogus gcc warning
switch to IP that is in 192.0.2.0/24 range which is specifically reserved
to be never allocated by IANA
hacks from trip
keep current priority average and use it
hackery
eliminate
dead
naming
done
perform reservation
fix
fix
ge
unindex on failure, location URI support
fix
minor fixmes
fix
fix 1579
fixing 1568
fix
order
inline downloads
fixing UDP bug
bind to v6 only
fix fs testcase
update docs
fast timeout on disconnect
fix
bad check
fixing 1581
fix
fix
log
harmless
defaults for dht
check and handle
use defaults
no autostart
no dht auto
no dht auto
no dht auto
no dht auto
student
renaming
renaming
prot fix
do not daemonize when running tests
do not daemonize when running tests
add code to enable checking if transport's idea of our peer identity
matches that of core, etc -- might help diagnose certain bugs
fixing bad memcpy
fixing 1584
fixing 1584
faster run
init
fixes
bugfix
misc
ignores
reduce verbosity
die debug code
fixing build
detect pcap
fixing accidental commit
fixes to build
fix
linux only for now
improved branch coverage GNUNET_TIME_absolute_subtract
fixing 1592
fixing #1591 with more detailed, consistent instructions
fixing #1589
update
fix
new pseudonym dialog
stub
update
update
missing
fix
update
done
done
mysql plugin compiles now -- not complete
missing 't' for 1595
mysql hacking
invert
invert
fixes
fixes
bugfixes
fix
fix
getting mysql to work
use test database
update
fixes
eliminate size field
size
options
implement update api
do not fail if datastore unavailable
ign
done
adding missing API call
done
done
misc fixes
iconv
mysql hackery
tolerate missing DB setup
fix
check
do not leak nullfd
fix leak of shc_pipe
fix hypothetical leak
check return code
less casting
stronger
solved
nitpicks
check for NULL
leak
leak
nitpicks
strict check
assert earlier
fix
fix
handling lost parent
defs
bughunting with assertions
fixing slist testcase
fixing compile errors
nodebug
nodebug
expose our hello to plugins
fixing fun bug
quiet
nice error message explaining failure
code cleanup
code cleanup
code cleanup
code cleanup
code cleanup
update
more code cleanup
more code cleanup
more code cleanup
more code cleanup
more code cleanup
fix
adding UDP server side
udp for w32
stuff
cleanup
stuff
stuff
stuff
renaming
cleaning
audit
whitespace
fix
postgres hackery
fix leaks
use big endian
fix
fixes
do not store size
do not store size
fixs
towards datacache
towards datacache
stuff
fix
implementing delete
fixes
fix
fix
fix
fix
done
train hacks
do not fail
fix
not here
cleaner
assert
snprintf
leak
fixes
nitpicks
demo test
stuff
cleaner
C grammar
stuff
nitpicks
asprintf
starting NAT code cleanup
hdr
nitpicks
nitpicks
nitpicks
add null check
nitpicks
nitpick
nitpicks
nitpicks
fix
fixes
fixes
skeleton for list_updateable testcase
fix
fix
NPEs
sparc
fixes
more doxygen stuff
more doxygen stuff
more doxygen stuff
update
leak
leak
fix
larger timeouts for sparky
fix warnings
backup errno
icondir
survive NULL
relax format check
issue
fix
W32 porting work by LRN
fixing bugs and compile errors
add newline at end of file
idea
fix present check for on-demand blocks
ok
warn
print source
ign
ign
ign
better
do not use zero:
clarify why we get these warnings
wrong name
towards pluggable block library
new block lib
nicer error msgs
move mingle function
fix
api
typos
use block lib some more
code simplification
fixme
plane hacking
no SUID on W32
match MHD 0.9.2 API
match MHD 0.9.2 API
fix doxygen warning
train hacking
mantis 1606
fixing warnings
stuff
fixmes
do not use default user, use my.cnf
new DHT API code
new DHT API code
new DHT API code
new DHT API code
new DHT API, tests do not yet compile
new configure flag to enable 'malicious' test code
make first test compile
make second test compile
make third test compile
make third test compile
making all tests compile
make code compile with enable-malicious option set
making test_dht_api work
fix
use propper blocktype
improved bf api
passing bf and xquery from client to service
client-side path tracking support
block library initialization
fix warnings
fix
mess
stuff
leak
fix
code cleanup
fix
indicate malformed blocks
dht get integration
use size_t
DHT PUT integration into FS
done
stuff
track delay between transmission request and satisfaction:
drop means max
entry lifetime tracking
adding support for artificial delays
fixing compile errors
small fixes and style issues
small fixes and style issues
fix
struct
unsigned
audit
enable out-of-gnunet build
for w32 port
fix
fix
fix
fix
better
fix
done
not in default
die cpu load code, die
nonce
done
fixing segv
draft
fix
fixing issue pointed out by amatus
break build
bogus
space
perf code
fix as suggested by amatus:
stuff
stuff
stuff
fix
style
stuff
corkstat
avoid disconnect on cancel
stats
fixes
faster stop timeout
log drop
bugfixes, more todo
fixing off-by-one
fix
nodebug
initialize reserved field, check it
largely for testing
fix
fix setting close on exec
do not set flags here
check reserved
check reserved
warn on problems
check reserved
close pipe on exit
logging
ign
ign
x
finding causes
fix
another place
off by one
less memory
stats and fixes
testing
fixed
improved load API to hopefully fix fs load mgmt troubles
bound as well
sane default
fix
fix
nodbg
cleaner stats
extended load api
allow zero
cleaner
more testcases, allow location uris in fs_download
do not reconnect, other peer violated, not core
fixing the fix
minus
update
smaller default perf size
sane defaults
refs to mantis
crash less hard
generated
move
newline
moves
move
fun
jccscript
fixes
api to add keyword
api to remove keyword
indent
done
fixes
seaspidering
proper transactions
ignore
removing unused
fix
fix
fix
ign
missing
stdarg
done
w32 patch from LRN
allow loc
fix
fixes from LRN for DEBUG code
fix DHT start, be less verbose
need longer timeout, better error message
fix as suggested by mwachs
ignore
cleanup
fix
defaults
stuff
docu
largely done
fix
Mantis 1613
original patch from Mantis 1614
style improvments wrt Mantis 1614 patch
fix
ignore
auth
fix
fix
LRN: llu
another LRN patch
LRN patch from Mantis #1615
cast
asserts, use static
dce
fix
fix
fix
moving plugin headers
fix
fix
rename
mention gnunet-setup
add API to test for plugin's existence
report config used
obo
fix
increase timeout
update
minor cleanup
allow not compiling malicious code
trying to fix grammar
more
seaspider fixes
useless
enable multi-dimensional arrays
less creative use of types
allow const static
seaspidery
seaspidery
report shutdown error
mkdir
fix
space
suppress warning
sockets patch from #1616
uint
lib
fix
stylistic fixes
enabling forwarding for AF_UNIX, no longer doing both IPv4 and IPv6 --
since this can never really be needed
better error info
indenting
not experimental
adding argument to GNUNET_CORE_connect -- not yet implemented
fix
fixes
import
import
stuff
dead arg doc
dead arg doc
the big core API/protocol change, breaks all testcases using core, since
the code is still buggy
fixes
hxing
hxing
fix
more checks
more checks
fix
fixing uninit fields
format
fix
fix
fix
fixes and more debug statements
do not use TC for transmission
misc
cont must be called instantly
fix cont mess
can take longer
another error to not report
assert
extra asserts for 1618, fixing one minor, likely unrelated use-after-free
code cleanup, set status to DOWN after connection goes DOWN, otherwise we
generate bogus notifications for clients and fail to re-do key exchange
fix
fixing 1622
fix 1618
fix
possible fix
fix
clarifications on r13759
fixing various issues Nate had: ready queue simply must be purged on
reconnect
ignore
fix
fix
ignore
full support for ATSi parsing
initialize uninit
log with pid
cleaning up TCP code, incomplete
fix for Nate
nl
style
style
memdup
adding UPnP/PMP library calls/support to TCP transport
order
format
indent
seaspider
random hack - 1618
fix
fix
fix
stuff
new vision for misc options
fixing bogus port bind by UDP
allowing DynDNS for external IP specification, option clean up
indent
bi-di migration options
docu
fix rmdir
fix rmdir
fix rmdir
track ATS in core
fix
nodebug
pass ATS data to clients
fix
pass ATS data to core clients
grab latency data
unit
removing fprintfs -- with bad %fmt statements giving warnings
fixing iteration code to handle ATS structs
removing unnecessary check and outdated comment
cleanup
minor style fixes
minor style fixes
indent
do not free if still connected
comments for Nate
air plane hacking
allow files of size 0
fix cast
fix segv on shutdown
viva vpn
fixing block reconstruction start/shutdown code
fix
LRN patch from SVN 1630
mantis 1627
fix
reconnect via testing testcase
reconnect test
allow transmission to self via core API
handle init NULL
fix
fix
fix
fix
fix
fix
fix
check
avoid %m
fix
fix
fix
fix
fix
fix
fix
assert
fix
fixes
fixes
fixes
cleaner
cleaner
fix
avoid inet_pton
fail on fatal connect error
fail on fatal connect error
cleaner
fix
fix
fix
plusplus
fix segv
fix
ignore self
fix
fix
init
dead
dead
exp
moved
fix shutdown order
doxygen
renamig pending_head/tail to control_pending_head/tail to avoid ambiguity
clenaer
fix
nolog
cleaning up core API with bugfix
ignores
fixing use-after-free
fixing #1632
docu
add keywords from metadata for files
retry faster
update stats, fewer status change updates
better stats
stats
track
fix
stuff
fixes
release
update
update
update
fix as found by LRN
fix
stuff
cancel hashing on shutdown
fix
fixing error message -- 1635
trying to fix SVN 1640
fix
fixes
fixes
fix possible infinite loop
rindex is deprecated
redundant
redundant
dead
Mantis #1644, unmodified, patch by vminko
ignore-generated
ignore-generated
code already there, doh
adding lifetime limits on migration content
contact page created
more done
date
not a test
breaking-indent
hxing
byebye
byebye
fix
abs_value
major rewrite of fs_download, hopefully fixing 1641
nicer
adjust
fix
note
defer
do not try TCP if port is 0
UNIX domain socket authentication support added
just use setpriority
indent
fix simulation option
docu
done
proper use of full data and fixing event signalling
docu
docu
docu
nosleep
append / to directory names (and remove later...)
gone
doxyfixes
assert element in/not-in list
asserts
fixes
name
api
undo
stuff
log
done
hdrs
improve weak random generator
stuff
noint
indenting
doxygen
ptoelke: please use this style...
stuff
stuff
fix
fixing bugs, declaring safe
docu
cleaning
fix compiler error for newer MHD API
undo
remove hard fail code entirely:
done
update
more chat code from Mantis #1657
undo
stuff
stuff
stuff
stuff
stuff
correct unit placement for GAUGER calls
stuff
stuff
stuff
use uint32_t for mutator
oops
be silent
gaugering
gaugering
gaugering
gaugering
gaugering
gaugering
gaugering
stuff
fix
fix
no includes on top level please
fx
move
die strtok_r
die strtok_r
fix
trans
trans
fix leak
fix leak
fix
fix
fix
fix
fix
fix
fs hacking
link with hello '
chat update -- Mantis 1665
chat indent -- Mantis 1665
fixes
fixes
fixes
fixes
fixes
doxygen for chat
more dox fixes
more dox fixes
argh
do not print to stderr
fixwarning
typo
idea
hxing
stuff
load
add cork option to core api:
implement cork option, fix testcases
stuff
changing heap remove node api to not pass heap; more fs hacking
prio
adding delay heap
help static analysis
fixes
assert
fix visibility
do remove
fix
fix
fix
doxy fixes
fix
void
fix
cleanup
assert for stat analysis
leak
leak
do not invert order
do not invert order
make peerinfo notify api more robust to serious disconnects
hate-it
assert for cs
that belongs into SVN...
make compile
make compile
ems
fixes
fixes
better API
adapt to new API
fixes-gaugering
disdebg
better msgs
fix
fix
handle indexed data correctly:
spell it out
polishing
fix test
fixes
communicate res delay in capi
fix
undo
delay as needed
fix
fix
verb
ugh
stuff
fix
zero
fix
fix
leak
fix
plat
mstdarg
ideas
ideas
fixes
make clang happy
ignore return values
spaces
communicate content already present better
fix migration support
noex
fixes
compile
log
changing datastore API to support replication level --- not fully
implemented --- and adding GNUNET_FS_BlockOptions struct to replace
prio/anonymity/expiration+replication args
undo
new plugin API
first hack at implementing new replication select code
improving datastore API --- not working yet
make compile
trying to fix autoconf warning
ign
sqlite clean up
rewriting perf test to match new API --- now it fails...
fix
undo
removing dead API calls
currently borken
renaming, fixes
move
test
fixes
fixes
dce
off
insanity
insanity
lnk
fixing 1673
indent
issue
going less bitfield crazy
1k buffers
need endian macros
ctype is required
nicer
nicer
nicer
die
indent
fixing build issues
fixc
nowarn
some optimization work...
getting mysql code to compile again, fixes to sqlite code
mysql fixes
fixes
fix
handle return value, kind-of
fix
ign
de-experimentalizing
update
fixes
fix
fixes
maybe
not anymore any more experimental than the others
disable
proper escape
debugging
fixes
fixes
syntax
nofs
fix
stuff
fixing shutdown
fixes
misc improvements
fix
cancel completed requests
fix
fixes
wip
disable-tests
datastore and fs fixes from Easter
new expiration api
fix
use correct expiration values
tolerate it if blocks already exist
nodebug
pass repl
done
fix
fix
clean
assertions that Nate should agree with
ign
nodebg
fix
watchtest
fixing 1671
nodebug
fixing leak
fixing leak
cleaning up learning code some
fix
fix
fix
stuff
ign
done
planning
fixing leak
done
fix
nocurl
dead
logstmt
sleeep
warn
eq eq
xtric
arg
link
linking
fix
novg
check
fix
fix
fix
proper shutdown
ignore
idea
fix
fix 1678
disable
indent, dce
clean up
clean up
cleaning
spacing
cleaning
linking
link
link
ignore
fix
fix
note
fix
more comprehensive error reporting
fix
simplify
fixes
nomesh
better logging
better
avoid restart
fix
fixing bug with connections that go down not coming back up
fixing bug with connections that go down not coming back up
additional test
ign
I guess this does not hurt
trying to fix function argument parsing
trying to fix function argument parsing
fixes
fix performance metric
fix performance metric
use fprintf for intended test outputs
nomesh
cleanup
warn if this happens
proper remove
fix
clean up logging, set rim ID always, even if neighbour was not found
no mesh starting
nicer debugs
def
wip
fixing major, highly annoying bug...
fixes
ignores
ignores
fix
minor code cleanup
better assertions, extra warnings
fix
more extensive error reporting
set sxn_len on FreeBSD where required
fixthefix
cancel task on timeout
notes on current issues
ign
extra check
docu, check NULL
set to NULL
null out
check DATASTORE return values
check return values, cancel timeout task
avoid direct call of notification -- except on error
make sure not to recursively call ntr
fix
fix
delay if disconnected
proper shutdown
indenting
indenting
track performance data
possible fix
do not remove from queue if already transmitted
stats
do not reset in_receive unless called from receive
only request next transmission if not already ready
nicer
fix
skip here
stuff
adding key for rupp, removing ayush
fix
fix
fix
fix
fix
fix
dmsg
fix
fix
stuff
fix
simplify api
simplify api
stuff
stuff
first draft for new transport api implementation
fixes
fixes
fixes
fix
reuse, add reserved for alignment
moving some tests to new API
fix
fix
fix
fix
adapting testcases to use new API
fix
use new transport API, minor core API change
ign
fix
fix
fix
cleaner
fix
stuff
update
fix
fix
indentation fixed, logging
disable warning if we intentionally throttle a connection
fix 1683
notes
fix
dbg
fix
enable test
fixed
fix assertion failure on shutdown
delay option
better
add support for delays
pcap no longer needed
stuff
fix
moving NAT options into their own section
fix
update
pass cfg to nat
fixes
update
move
fix for 1685
fixing confs
fix
fix
fix
fix
fixes
static
fix
fix warnings
fix
adding configure code for --enable-benchmarks, --enable-expensive-tests,
some clean up
update
ship and install man pages
fix
ign
check earlier for connect-self messages
extra checks against connects to self
nospace
include wlan files in build
ign
include
ship with conf files
ship with conf files
bump
refuse connections from self
include testdb
fix
add API to remove section
not today
fixing datastore schema for future change for improved performance
fixing 1689
fix
doxygen
doxygen
fix
W32 is too slow
fix
ifdebug
do shutdown on CTRL-C
pregenerate hostkeys
fixing 1692
asserts
format
fixing 1692
fixing 1691
fix
format
disable warning
dce
cleaner message
oops
cleaner
fix
fix
fix
fix
fix
chown
define htole16 if libc does not
disable vpn autostart
cleanup
fix
mark experimental code
do not crash if we run out of sockets
add connection limit to TCP
resolver API change
moving monkey
fix
fix
assert
make code more tolerant to re-connect failures, i.e. due to being out of
sockets
cleaner
cleaner
fix
fix
stat
better clean up, more stats
dnc
do not allow NULL for fn
fix
fix
stats
stat
fix
fix 1708
more asserts
fixing 1706
die harder
fix 1707
fix 1695
fixing 1696
fix 1698
fix
be more harsh
debugs
cleaner datastore handling, fixing test failure
fixing 1710
fixing 1711
cleanup
warn later
cleanup
cancel change preference request
more error checks
fix
more assertions for tracking down 1709
maybe cleaner to initialize explicitly
fixing typo, possibly over-eager disconnect
check if this pos was connected before...
checks
fix
trying to fix bus error
allow tests to pass on old sqlite installations
tolerate partial writes better
make resolver test pass even if primary DNS is down
high-precision LOG timers
avoid duplicate binds
Mantis 1716:
helping w32
fix
mantis 1718
fixing leak (1700/4444)
code clean up wrt timeout handling, etc.
fix
spaces
xtra checks
fixing 1721
fixing 1720
fixing #1723
fixing 1722
fixing 1719
possibly fixing leak at #1700/4451
dce
fix
fix
fix
info
mark as disconnected
fix
fix uninitialized timeout, use-after-free on mysql errors:
the big NAT change
partial fix of the NAT troubles
moving NAT client/server to NAT library
moving NAT client/server to NAT library
improving NAT code
make refresh frequencies configurable
support infinity/forever for time value in configuration
already done
renaming gnunet-nat-server/client for consistency, other minor fixes
finishing gnunet-nat-server
finishing gnunet-nat-server
ign
draft for NAT_test API
more hacking on nat test code
comments
more nat code
indent
testcase for nat test code
towards fixing 1714
timeout earlier as well
cleanup
fixing 1714
ign
invalid addresses cause transport trouble, removing to avoid for now
stuff
option to not do DNS lookup of hostname
fix
report IP from bind
stuff
hack for empty HELLOs
hack for transport bug
fixes
unable to rep
trying to fix 1717
fix
dead
fix
fixing 1729
fix
udp fixes
fix
undo
set af, better error msgs
adding code to get external IP from miniupnp tool
move SUID test code to util
fix
another attempt at 1728
still bad, but better
additional process start fun
fx
towards UPnP support
apifix
mini works
integrating miniupnp with nat library
ignore
endian
must not NULL
dbg
move
fix
done
new fragmentation API design - with retransmit
fragging
frag
data structures
defrag
add
bincheck
fix
proper cleanup
logging
fixes
better
frag
adding test -- currently failing
dce
frag
revised fragmentation API for blocking writes
dce
rename
nicer stats, bugfix
fix
indent
new udp code
node
fixing use-after free and/or possible bad recursion on disconnect
mantis 1735
MAntis 1616, 0001-Temporary-fix-for-process_kill.patch
Mantis 1616, 0001-More-permanent-fix-for-os_process_kill.patch
check
fix
never discard current mc
line
reliable udp
renaming
fix
fix
fix
add 1 std dev -- seems to help:
reduce size of benchmarks
fix
config
config
not zero
build frag
fixed
fix
leak
some code cleanup
leakfix
wrong check
fix leak
fix
zh
fix
move assert
eliminate leak
const
leak
leaks
leak
dead
stupid code
stupid code
fix
fix
fix
fix
fix
fix
fix
fix
fix
fix
fix
ign
fix
fix
nicer
fewer parens
fewer parens
fix
trying to do something about 1728
cleaner
update
dce
dce
dce
fix
fix leak
leak
leak
checkneg
assert
assert
assert
fix ats config
assert
assert
const
checkret
voidret
checkret
checkret
fix
checkret
fix
fix
sqrt
fix
duplicateline
dce
leak
leak
uaf
leak
defaults belong in defaults.conf, not in source
defaults belong in defaults.conf, not in source
assert
checkret
fix
fix
nse
fix
leak
ugs
fix leak
Mantis 1735
fix 1693
do not autostart nse
fix
tolerate duplicate fragment ids
leak
fix shutdown
const
done
dd dependencies to some programs
Include builddir include subdir with -I
Fix a heap corruption in envtable creation routine
cleaner
cleaner
run glpk code and tests only if gplk installed, fewer ifdefs
enable-ats configure option
fix
static
dce
fixing 1732:
fixing 1732
fix
cleanup
free
rename
usenew
link
build pots
simplify
first disconnect from core to handle disconnect events nicely -- likely
cause of #1724
ign
fix
cleaning up ARM shutdown
fix
dead
ign
just indentation / formatting
nse hacking
fixes
fix shutdown
fixes
fix starting stddevs/avgs
use messages with moderately futuristic timestamps
incremental proof calcluation, store on disk
fixes
success
fix
move to config instead of hard-coding for testing
fx
indentation
static
fix
generate histograms
generate histograms
no hard warning here
allow user to configure no deletion with testing
fix
exponential back-off
deduplicate requests
debugnotinfo
fix
fixing 1738
lm
can and should use longer timeout here
fixing fs
fix
fix
unix
init to zero
push back
fix
fix assertion:
mantis 1739
fix
option for not having delays
remove option
fix
production settings
fix
fix
fixes
rename
fix
cleanup
ignore
dce
differ
no range for filenames
simplifying configs with includes
simplify
simplify
simplifying config
optimizing cfg
simplifying configs
fix
LRN: Allow env vars to adjust and override loglevel
LRN: Bratao's select() implementation for W32
LRN: Be aware of absolute DOS-style paths as well
LRN: Fix _kill() debugging and I/O interpretation
LRN: Don't wait for parent control handler to quit
LRN: Fix debug logging in test_configuration
LRN: another logging fix
LRN: Let gnunet-transport-certificate-creation be a program
LRN: Fix debug logging in some transport tests
LRN: Make transport unreliability test less optimistic
LRN: Keep passing HELLO messages to transport
LRN: Bind test_testing transports to localhost
LRN: Fix debug logging in dv
LRN: Make dht test a bit less optimistic about MinGW
LRN: Fix test_dht_tools.sh to ignore CRs when diffing
LRN: Fix hostlist learning test to use /dev/null or NUL
LRN: Clear peer context before configuring it
LRN: Make transport tests less optimistic about MinGW
LRN: Resolve an overlooked merge conflict
fixing compile error, improving style
a bit cleanr
a bit cleanr
LRN: added new file
style
defaults
towards simplified configs
generatedfile
ignore
simplify
fixing double waitpid
fix double-waitpid here too
cancel on disconnect
fixleak
no hostlist
no topo
fix
LRN: fixing http issues
fixing parent_control_handler hack, adding lifeness attribute to tasks
fix
fix
force passing of sanity check
chmod
fixfix
nse
nat
natnse
significantly improving mysql performance - as planned a while ago
3 broken versions for sqlite
use postgres PRNG also on insertion
use sqlite PRNG also on insertion
stuff
LRN: Fix automake deps to allow -j* builds again
improving sqlite performance
improve API, speed up mysql
fix
LRN: Fix a rare arm-interceptor double-free bug
report issue better
minor bugfixes, switch neighbour list to hash map
rename neighbourlist to neighbourmapentry
make more resillient
fix
send message type maps to neighbours
move
fixnpe
done
rename
fix
draft headers
draft
implementing new HELLO api
towards plugins implementation
implementing plugins subsystem
implementing blacklist
init blacklist
improved client API
improved neighbour API
more client code
more client code
more client code
use relative timeouts over network
simplify
more refactoring
stuff
LRN: 0001-W32-create-parent-directory-for-a-new-file.patch
LRN: 0002-Make-group-shutdown-a-bit-more-verbose.patch
LRN: 0003-Make-pyexpect-a-module.patch
LRN: 0004-Remove-pyexpect-code-import-it-from-the-module.patch
LRN: 0005-Port-old-pexpect-tests-to-gnunet_pyexpect.patch
LRN: 0006-Use-plibc-to-convert-dir-separators.patch
LRN: 0007-Add-N-NUM-argument-to-quit-the-search-after-NUM-hits.patch
LRN: 0008-Specify-hostname.patch
LRN: 0009-Fix-.exe-extension-in-topology-string.patch
slight fixes to LRN patches
moving pyexpect, dummy init (missing from patch)
pymove
document option
nicer
ign
not needed
remember notifies
send validation PING
send target address with PING
transmit PONGs
handle PONGs
re-enabling signature caching code
new message types
towards neighbour management
implementing some neighbours fun
fixing compile errors
more neighbours code
peerinfo get missing from validation, fixing
stuff
stuff
likely duplication
stuff
fix
docu
noex
strndup
inits
new ATS api
use bind instead of connect to test if service is running to avoid long
timeouts during startup
ign
ats
more ats API hacking
more ats API hacking
fix
stuff
dummy ATS implementation
dummy ATS implementation
dummy ATS implementation
use NEW ats
fix
done or in Mantis
util
send HELLO with PING
validation should notify async callbacks about updates
implementing client's send
moving API around to make ATS implementable and separable
initial ATS service refactoring
initial ATS service refactoring
add address timeout
LRN: disable bind test on W32
switch api
track cont
no need for public key
finishing neighbours
docu
LRN: Fix .exe extension in transport tests
fix
notes
notes
fix
traffic report is not used
cost report is not used
stuff
stuff
HELLOs
setup handlers
adding missing config files
dos2unix
LRN: Added-the-ability-to-substitute-scheduler-select.
remove getter, make global static, add closure argument
LRN: Fix deps in transport
make compile
make compile
LRN: Fix deps in ats
LRN: gnunet janitor -- another killall variant
docu
LRN: fix typo
stuff
handle connect request, add message types
deps
demultiplexing
implement fast reconnect, handle disconnect better
more sane numbering
stuff
stuff
ign
process keepalive, simplify PONGs
cleanup
check
notes
make external-IP api non-blocking
indentation preferences and typedef finding code, mostly from LRN with
minor changes
LRN: Prepare-for-indentation
indentation
indentation
indentation
prec
indentdiff
even nicer indentation, thanks to LRN's indent patch
script to remove whitespace at the end of a line
run indent twice, it alternates between two 'canonical' forms, also run
whitespace remover
improved script
docu
foundit
fix
don't try to receive before successfully connecting and sending the local
connect message
nicer alignment, unsigned sizes
use unsigned int
minor cleanup
delay calling notify from notify_transmit_ready
malloc NULLs already
encapsulate payload, keep target/priority/etc.
timeout support, handle empty queue due to cancel/timeout
include message overheads in size request, handle change in message size
at queue head
communicate operation timeouts
use define
todo list
working on DV -- trying to fix bug found by LRN
fixing double-adding of the 'die-task'
LRN: fast and dirty transport crash fix
LRN fixes to DV config to make tests pass:
LRN: fix dht tests --- not sure this is the best fix
don't add addrs like crazy
dead
typos
fix
cleanup API for get_hello and get_hello_cancel
fix
fix doc
making mwachs happy
making mwachs happier
breaking stuff
doxygen
LRN: A bit less dirty transport crash fix
no longer true
fixing API and clean up issues in testing
fixing 1761
no whitespace after backslash
fix
misc train hacking
allow cancellation of certain transport API operations
allow handlers to be NULL
mark dv as experimental
status update
clarifying some FIXMEs
simplify
remove non-binary peer selection strategies
options should be plain 'int'
dce
dce
integrate DHT with NSE
doxygen
NSE also replaces max_hops
NSE also replaces diameter
stuff
removing dead republish functionality
stuff
use time api
stuff
dce
dce
dce
dce
dce
stuff
make new current
ignore
fix
updating po-files
adding missing files
adding missing files
fix
make bitching less verbose
fixing 1752
fixing 1753
fixfix
fix
fix
fix
fix
extra assert, use named constants
die
fix
indenting
indentation
fix: ipv6 IPs should be in []. strlen bad parens
use INDEXED BY only for sqlite >= 3.7
do not start mesh
simplifying configs
better reporting
fine
clean up
clean up
stuff
dce
check return values
reduce problem size
actually use external resolution code; fix SUID check
dbg support:
check for external-ip before forking
fix
towards fixing 1773
do use reuseaddr on non-W32 systems
c3018
c3015
c3054, also do not memset after GNUNET_malloc, malloc already memsets to 0
leak, c3053
leak, c3045
out of bounds, constant not named, c3042
statics are zero, so this is actually not even needed
fix
fixing FIXME, major bug with dangling pointer, c3142
c3133
use named constants
unsigned
longer exp, cleaner code
c3310
c3267
c3238
static
make 'obsolete' data structures conditionally compiled
no useless printfs
ign
fixing 1770, towards fixing 1769
fixes
fix
removing reply bf entirely, problem was with UID removal
fixing 1776 --- do not check uid in client and don't put it into recent
fixing 1767:
better error handling
SO_SNDBUF/SO_RECVBUF do not work on FreeBSD
more efficient implementation, maybe helping with 1777
fixwarning
linux-only
longer timeout
fixing #1781
marking HTTP as experimental
update
generate valid disconnect message -- Mantis #1789
done
indentation
no latency can happen, should be rare
do latency computation - #1768
limit mesh tests -- mark as experimental
disable debug options
done
fix
fixconfig
adding missing index
keep problem size
print emsg
print emsg
print emsg
fix emsg
fix emsg
cleanup
stuff
add
extend
extend
finishing new API
stuff
stuff
wild hxing
LRN: use bindto instead of use-localaddr
stuff
add
more wild hxing
hxing
making some files compile
making some files compile
move bloomfilter recalculation to block library
stuff
add new DHT code to build, it finally compiles
respect path tracking and demultiplex everywhere options
give priority to results
process find peer requests and replies
update
do not route requests that were perfectly satisfied
getting client API to compile
generate STOP message
helper
track HELLOs from peerinfo
fixes
fix
stuff
fix
adding some stats
fixes
better hello block lib
fixes
fix
fixes
migration
DCE
fixes
die
move
nonews
nonews
DHT api switch adjustments
DCE
make compile
MHD has a better way than a configure test for this
cleanup
paren
stuff
fixes
dbg
dbg
fixing warning
static
fix #1796
static
fixing 1802
cleaning up testcase some
dbg
const
use ring 10, test all-to-all GET/PUT
simplify code
fixing 1804
better reporting
nicer
disable for now
report disconnect reasons in stats
report disconnect reasons in stats
fix disconnect message transmission, send keepalives
cleaning up testcase, debugging
bugfix
log more details for debugging
do not add 1 twice -- callee also added 1, resulting in an increment by 2
per hop
use unique keys for each peer pair
fix code to perform GET retries
also demultiplex everywhere on put
nolog
use demultiplex everywhere
removing excessive logging, adding missing decrement (choice) to avoid
assertion failure; fixing loop bounds in non-random peer selection
remove excessive logging
nicer
speed up BF tests
re-add logging, move BF add operations to the latest possible time to
avoid polluting tests needlessly
fix
adding empty lines
fixing bloomfilter code, fixing testcase
no printf
fix #1713 (see discussion there)
fix 1814
factor out stop into new task (in preparation)
adding stats
LRN's big logging rewrite (#1805):
cleaning up test code to keep references to allocated handles and to
clean them up where needed; this should fix #1803
fixing memory leak (#1809)
add a few assertions
add statistic
print stat
unnecessary
fix stat, count 2nd case as well
more stats
LRN: Use GNUNET_EXTRA_LOGGING to manage compile-time logging calls
(#1805).
adding assertions to help with 1815
use load, not parse
use load, not parse
increase timeout to hopefully fix issues on slow buildbots
fix
add GNUNET_PACKED
pack block, use big endian for time
generated
allow for all CBs to be NULL, then do not send START
tell transport to connect, not core
tell transport to connect, not core
tell transport to connect, not core
tell transport to connect, not core
towards new core service implementation -- breaking core up into smaller
modules
eliminate last calls to GNUNET_CORE_peer_request_connect
build draft for 'new' core code
eliminate REQUEST_CONNECT handlers
eliminate last calls to GNUNET_CORE_peer_request_connect -- missed one
typo
cancel con_task
bad over-optimization
fixes
fix
p should be n
p should be p
moved
add packed
fix correction according to paper, minimum NSE should be our neighour set
prevent very low estimates from being used
dbg
updating ignores
testcase for runtime loglevels form LRN
fix
LRN: Fix select() on pipes to trigger read callback on EPIPE
LRN: Fix runtime logging test, since pipes work as intended
vminko: new API to delete a TransmitContext (slightly modified patch from
#1822)
change test to not require . in PATH
hxing
hdr
ugh
copy from API
stuff
stuff
moved
finishing clients draft
drafting typemap API
typo
towards KX
make opaque
fixes
hxing
hxing
hxing
removing 'publicKey' argument from CORE init callback
hxing typemap
stuff
fix
hxing
hxing
first round of simplification to change preference API
LRN: Small janitor fixes
LRN: Fix logdef processing logic
core-new now compiles
fix: update available recv window bw if bw_in changes
remove limit_outbound, DCE
remark
ATS API: do inbound and outbound bw
adding change preference API to ATS
renaming in code to use new ATS pref/reservation API
dead code elimination
fakeit
dce
removing dead bw code
removing dead bw code
DCE
fix
another core API simplification due to ATS introduction
should be done by ATS in the future
done
doku
lon
cleaner
fixes
cleaner
kill malicious flag, legacy protocol numbers
undo
update message
fix
use log from
log from macros
fix
new convention
LOG macro convention
use LOG macro in statistics_api.c
use LOG macro in core_api.c
use LOG macro in nse_api.c
use LOG macro in dht_api.c
use LOG macro in datastore_api.c
use LOG macro in arm_api.c
use LOG macro in peerinfo_api.c
use LOG macro in peerinfo_api.c
use LOG macro in resolver_api.c
use LOG macro in ats_api.c
use LOG macro in mesh_api.c
use LOG macro in dv_api.c
fix
use LOG macro in transport_api.c
convention
use LOG macro in nat.c
use LOG macro in nat.c
space
fix
use LOG macro in datacache.c
process inbound type map messages:
ugh
converting to GNUNET_LOG_from*
reudce insanity
fix
check vspnrintf return value, stack-allocate log buffer
fix
fixes
fix
make client with an empty list of handlers count as having a handler for
*all* message types
also allow deferral of encrypted messages
changing WARNING to DEBUG where appropriate
fix
do not dequeue here
use DEBUG here
fix
ignore
remark
LRN: Fix slist misuse in network.c
commenting out failing test for now
better logging
checknull
fix
log
log
set quota
ugh
log
debug code, some hackery around connected flag
we don't call GNUNET_ATS_peer_connect yet, cannot expect ATS to only tell
us about connected addrs
making new core default to see how (badly?) it goes
allow NULL
leak
leak
ATS API cleanup: remove 2nd callback type with virtually identical
signature
new ATS API -- code won't compile
sapi implemented
rename
commenting out old tests
renaming
documenting
API change to make mwachs happy
parse updated message type
parse updated message type
finish
finish ATS API implementation
adapting code to new ATS APIs
fixing compile errors
fix
copy and paste copies bugs
fixing late night hacks
make compile again
fix stack overflow
call receive only after sending INIT
merging
moving out addresses as well
removing client code
more client disconnect code
more refactoring
LRN loves slist: Use stack allocation for slist iterator
LRN: Use GNUNET_strdup() instead of strdup()
LRN: fix LOG definitions
more ATS implementation: message parsing and generation:
stuff
parsing reservations
complete reservations API
comments
implementing reservations
implementing reservations
cleanup
cleanup
cleanup
use after free
rcb, not rc
could be NULL
trying to fix keepalive logic
get quota config
move get quota config to addrs
more bw assignment code
fix
fix
send performance info to clients
fixwarning
parse preference change messages
move GNUNET_TRANSPORT_ATS_ to GNUNET_ATS_
restrict to 1 scheduling client
fix segv
towards nicer scheduling code
fixes
fix
fix
ign
configure ATS ports conflict-free for tests
fix
fix
fix
try to connect more than just once
try to connect more than just once
give ATS disjoint ports/paths
give ATS disjoint ports/paths
fix
only transmit if flag is set
smaller, not smaller-or-equal
bad free
remove old core
rename new core
new is current
simplify
removing bad reserved field, small refactoring
fix
fix
fix
more assertions
more log messages
bulk
needptr
cleanup
extra checks
fix
fix
send session release messages back to ATS API
send session release messages back to ATS API
fix
fix
try fix shutdown order
no longer needed
notes
fix
dce
leak
fix
fix compile problem
fix
fix
adding stats
cleanup+stats
fix
fix
major bugfixes in performance api implementation, some logging
major bugfixes in performance api implementation, some logging
even more stats, more logging
more logging
init
typo
do not loop back
Fix stat
weaken assert
bracket
prevent loopback routing, add some statistics, faster re-transmit
fixing 1830
fixing 1830
fix
fix
fix
not for 0
logging
do not destroy session twice, do destroy address on failed connect
fix #1826
fix
fix
simplify code, run longer
fix
trying to fix assertion failure at :496
fix use of ok
fix
be even more lax in the test
fix bad free
check
fix
do terminate properly
termination is NO
fixing bad fix
following Barts suggestion to raise quotas even higher to avoid connect
issues:
fixing 1792: validate signature in disconnect request
ignores
fix canonicalize
fix #1784
more keyword canonicalization, fix threshold, update testcases:
fix segfault introduced by recent refactoring
try to make sure to send connect on new neighbour even if tcnts are both
empty
did not help
eliminating GNUNET_TRANSPORT_set_quota code
fixing 1836
more stats, better names
use same name everywhere
add typemap retransmission code
add and fix stats
more stats
remove unused options -- these bits were always set, no need to pass them
consist
consist
replacing 0-terminated atsi-array with array+length in core API (and the
core-connect IPC)
getting rid of unnecessary ignore_shutdown calls
nse interval is too high for some test hosts
missing returns
larger interval helped, adding another 4x to be sure in the future
additional assertions to help with #1840 and #1835
incorrect to assert here
make writing data_output optional and disable by default
be more verbose
possibly fixing 1840 and/or related issues
simplify logic
another assertion, possible fix to 1840
:fixing another contributor to 1840
op
more precise log message
if CONNECT failed, move to disconnected
Adding FIXMEs: if delay is FOREVER, session should be terminated
undo 17963, better solution ahead
also generate PONG if PING contains an empty address:
fixing quota violation messages
packed
fix 1839
longer ARM stop timeouts, powerpc buildbot still has trouble
fixing bogus assertion / 1846
fix compile error
fixing 1845
fixes
done
record more stats
fixing 1837
possible fix for 1844
check return values
abort on error -- missing return
use long long
asserts for cov
fix leak
check return vals
check return vals
fix logging
more logging
more logging, cleaning up logging
check return vals
logfix
improving r17722
improving r17722
log eval result
log local operations
log local operations
fix
even more logging
problem is caused by client disconnecting before we can queue the reply,
more logging
more logging for insane bug 1829
logging
fixing 1829
put ifdefs around DEBUG logs
fix loop bounds
ob1
fix
improve error reporting, fix zombie attack (#1848)
trying to fix #1813 by actually respecting delay values, also some minor
code cleanup
increse delay to help slow systems
simplify
static
fix
longer timeout -- trying to see if this helps with 1851 ---- and using
named constants
fix
fix
fix
fixing #1850
help compiler
fix compiler warnings
start draft of gnunet-transport
fix
expanding gnunet-transport towards a benchmarking tool
dummy
fix
fix
kb
setenv to fix testcase under other circumstances
documenting gnunet-transport a bit
integrate gnunet-transport-list-connections with gnunet-transport
fix
reading for disting
dead
missing extra dists
reconnect on protocol errors instead of crashing -- see #1863
deleting old mesh API
making the new mesh the default
fix
fix
dead
handle NULL better
fix
fix
fix
do not call send_disconnect with NULL
check return value
more DLL usage errors
return on error
update AUTHORS
ready to dist
fix
infinite bw hack
ugh
disable delays if infinite bandwidth
undo 17914, 915, 916
not exp
fix
fix
fix 1817
fixing #1820
note for bart
implementing transport test / #1807
document transport test / #1807
try current dir
implementing #1747
fix 1866 -- stats instead of warning
increase timeout to fix #1872:
fix
fix 1821: only send payload to transport clients that care
implementing plugin directory scanner for #1746
fix #1746
implementing time and size parsers for #1875
use new quota parsing capabilities -- SVN 1875
nodebug
fix
fix
implementing last part of #1107: #1806
do not translate, ever
implementing #1785
cov fix
handle NULL
fix uninit
fix off-by-one and named constant
initialize count
fix leak and missing return value check
fm never NULL, dce
ignore return values explicitly
parse earlier, check sscanf return value
close stdin before dup-ing, check return value of dup2
ignore return values explicitly
curly wars / auto-indentation
use putenv instead of setenv for portability
LRN: Fix loglevel testcase (disable redirection)
LRN: Add '-Wl,' prefix to -no-undefined
LRN: Compile against libgnurx (W32)
LRN: Use regexps in logdefs
no in_addr_t on W32
LRN: Mysterious patch. I don't know why calling executable with a dot
forces it to create a new console window (which immediately pops up and grabs
the focus, then disappears) for itself, but that is the empirical fact.
Removing the dot fixes this VERY annoying W32 bug.
LRN: Add -lole32 (why had it worked without this until now?)
LRN: Add -lole32 (why had it worked without this until now?)
LRN: cosmetics
LRN: This is likely a typo. I caught it, because GNUNET_assert() is now a
function that takes 'int', but the argument was a pointer.
LRN: Call-transport-disconnect-to-clear-session-message-q.patch: This was
submitted to #1880, but i'm including this here as well.
treat empty string as NULL
LRN: fix gnuent_fs_Start arguments
introducing GNUNET_abort
fix #1883
PING regularly, track latency, towards fixing 1793, 1879
fix #1885
more code to get latency in ATSI working; not complete, also now
generating a warning -- for a real problem that still needs to be fixed
adding address abstraction
refactoring how we handle peer addresses in peerinfo/ats/transport/hello
subsystems -- use a struct instead of 3--4 arguments
fix shutdown order
fix
fix
cleanup
fix nested use of strtok
fix
assertions and minor cleanup
minor cleanup
clean up, asserts, FIXME
cleanup
fix
fix
fix
static
fixing 1787
nicer
fixing 1891
cleaner destroy by session ID
log code for debugging ATS session management
only do small messages on non-Linux UNIX domain socket tests
fix emsg
fix emsg
towards fixing 1903: break instead of assert
plugin name should never be NULL
asserts for 1903:
more asserts for 1903
ignores
ignores
ignores
fix 1903
dead code elimination, splitting fs.h into fs.h and fs_api.h
dbg
moving P2P messages out of fs.h
towards fixing #1786
set continued flag on client side -- towards fixing 1786:
fixing 1786
fix warnings
spelling
fixing 1908
only use good acks for delay calculation
consider number of retransmissions for delay calculation
consider number of retransmissions for delay calculation
consider number of retransmissions for delay calculation
not on OSX
add units to time, use configuration time api where appropriate, fixing
Mantis #1875
fix
dce
fix
fix
fix
fix
fix
fix
moved to drupal
fix leak
fix defaults
fix defaults
moving block plugins to respective subsystem
stuff
moved to mantis
moved to mantis
dead
adding svn2cl script for automatic change log generation from SVN commits
integrating ChangeLog generation with build system, removing ChangeLog
from SVN control
fixing binary build (file no longer exists)
updated file list in POTFILES.in after renaming operations earlier
automated gettext updates to PO files from 'make dist'
making sure ChangeLog is properly generated on 'make dist'
update ignore list to include ChangeLog
adding shell script to show changes prior to commit to reduce submission
foo-bars
simpler alternative to the West algorithm to convince ourselves that we
can still do averages...
fixing variance calculation - we calculate the average from the same
distribution, so the degrees of freedom is n-1
fixing west algorithm to also use j and to consider case where
estimate_count is 0
LRN: fixing 1956 by using a better random generator on W32
do not try to run with valgrind everywhere
LRN: Fix the use of SRANDOM
nse never queues more than one message per peer
fs also never queues more than one message per peer, so reduce requested
queue size
prevent same message from being requested twice by always skipping
head-of-queue when inserting (possibly related to 1957)
check return value
remove generated files from SVN
-ignore
increase
changing API in anticipation of future implementation of 1369
disable use of signatures if nse proof is disabled -- for example, this
is done in testing/benchmarking
adjust log level to DEBUG as this is normal
better error message
-cleaner
fixing mantis 1960
-minor code clean up and renaming dummy
inlining files from wlan, preparing to obsolete wlan subdirectory
removing no longer needed wlan directory
some more wlan code cleanup
-some more wlan code cleanup
-some more wlan code cleanup
-some more wlan code cleanup
-some more wlan code cleanup
LRN: I'm tired of seeing something like "ATS tells us to switch to
address '(null)' session 0x497198 for peer `IIH1' in state `S_NOT_CONNECTED'",
so i wrote this.
-pyexpect fix
-fixing string printing - size_t is not unsigned int
fixing mantis 1961
-files were removed earlier
-fixing 1963 by changing warning to stats call
-files were removed earlier
-reduce log level
-spelling
LRN: Use GNUNET_strdup() instead of strdup ()
LRN: Do not stop daemons twice / #1919
LRN: Fix dht multipeer testing crash - #1919
LRN: Fix dht multipeer testing crash #2 - #1919
LRN: increase dht multipeer test timeout
Implementing Thomas Bushnell's suggestion to work around the signal race
without needing special vfork semantics
-document desired direction of transport API change
-document clock skew variable a bit more
-doxygen fixes
introducing GNUNET_UNUSED macro instead of hard-coding gcc attribute all
over the place; also some further documentation fixes
-more doxygen fixes
-more doxygen fixes
-fixing use of uninitialized variable, some code cleanup, fixing memory
leak on error path
-typo
-improving NSE test error message
cache signature upon first generation to avoid useless re-signing of the
same data
-handle shutdown
-trying harder to remove useless signing in NSE
-do not send previous round messages if we are just going online and did
not setup those messages yet
LRN: new code for GNUNET_OS_network_interfaces_list on W32 improving
support for IPv6 and subnets/masks, see Mantis 1958
-nl
- removing files in flux
-set ignores for various generated files
-cleaning up message format and code related to recent transport address
stringification code
-no valgrind
adding authors header block following suggestions from Werner Koch
-fux
fix xslt invocation for change log generation
HACKING guide moved to CMS on https://gnunet.org/style
adding missing files to dist
-fixing indentation
-fix typo in configuration
-make compiler happy (presumed uninitialized variable)
-fixing minor typos, version bump
releasing GNUnet 0.9.0
LRN: Adding vectored exception handling for W32 (#1965)
-undo
-LRN: fix socket_close logging
-LRN: unset control pipe after use, add missing #define
distfix
fix
LRN: Fixing Mantis #1974: On W32 winsock2.h defines FD_SETSIZE (if it was
not defined before inclusion of the header) to 64, which means that it's not
possible to select on more than 64 sockets at once. This might work during the
tests, but in the wild people might want to have more than 60 connections, at
least in the transport service. The patch attached should increase the limit to
1024.
use uint64_t instead of long long for GNUNET_ntohll/GNUNET_htonll
-use uint64_t instead of long long for GNUNET_ntohll/GNUNET_htonll
-preping arm for test by LRN
Fixing #1976 by allowing allocations between INT_MAX and SIZE_MAX and at
the same time limiting BF size for datastore to 2 GB. Also fixing infinite loop
when creating BFs of sizes between 2-4 GB
Implement passing sockets in IPC on W32 (#1975)
LRN: Ensure that GTK can then do the internal call write(2, message,
strlen (message)); successfully by default.
fix compile
LRN: correct behaviour when logfile does not exist
-only with MINGW
eliminating last dependency on util code from wlan helper by inlining
-duplicated
-server_mst simplification as possilbe from inlining
-make mst code static, remove GNUNET_SERVER_ prefix
removing ARM interceptor connection forwarding post-accept code, now
using lsocks everywhere
-moving interceptor.c code into main arm file
-fix missing library link flags
fixing issue with sending back size estimate messages to peers that
already have good estimate information
add statistic to track estimated network diameter
-add logging for randomized delays
-eliminate special case, no justification for why it is there
major rewrite of ARM service and a bit of the ARM IPC to take advantage
of the simplifications possible now that we no longer intercept traffic; the
new code in particular is better at communicating what exactly ARM was doing in
response to requests. A major change is that gnunet-arm -i/-k now only impacts
if a service is running by-default, on-demand starting is no longer impacted,
option -t from gnunet-arm was removed
-LRN: misc. minor fixes
-LRN fix deallocation
-LRN: Fix for ARMed processes that have zero lsocks (such as hostlist)
fix crash
-revising stream library api
-careful WLAN code cleanup
-more minor cleanup
renaming WLAN helper process to gnunet-helper-transport-wlan in order to
satisfy naming conventions
trying to fix #2000
-fixing #2006
-api improvement
-api improvement
extra error checking in Bloom filter to check that the size of the file
on disk corresponds to the expected size for the given filter
store Bloomfilter with rest of datastore data
only try to read bf from disk if we didn't just create the file
-fixing pos
adding support for detection quota changes / missing bloomfilter files
and reconstruction/recovery code
use better heuristic to deal with datastore put failures; log migration
stop message transmissions
-LRN: w32 version of fstat
some very minor stylistic changes to gnunet-helper-nat-server based on
suggestions from Jacob Applebaum
additional stylistic changes to gnunet-helper-nat-client anticipating
next round of suggestions from Jacob Applebaum
-make assignment cleaner, move to where it is needed
-cast returns from close to void to indicate that we intentionally ignore
them, close socket on new error paths
-cast returns from close to void to indicate that we intentionally ignore
them, close socket on new error paths
-cleanup
cleaning up bloomfilter code and disk file size tests
really minimizing code running with root rights
really minimizing gnunet-helper-nat-client code running with root rights
-also minimizing SUID code here
code cleanup, also trying to fix #2011
implementing get_keys API for sqlite datastore plugin (#2013)
-implementting get_keys for mysql (2013)
-implementing get_keys for postgres
-fix error message
-even better error handling
-2011
adding apparmor profile for gnunet-helper-nat-server from Jacob
-tolerate bloomfilter creation failures bettter
-fix order
-another stats crash
-handle error codes from unindex
-add numbers to help devs
-track utilization by datastore plugin type
-use plugin-specific bloomfilter file name
-updating ignores
fixing calculation of Bloom filter size that was too large by 1024x
because it was not adjusted when the unit for the quota was changed from kb to
bytes
-typo
Jacob Appelbaum reviewed gnunet-helper-nat-server and affirms that the
code 'seems fine'
-actually use global_ret for return
adding min_expiration argument to GNUNET_DATASTORE_ContinuationWithStatus
callback to communicate which content has a chance of being stored in the
medium term
-actually initialize min_expiration field
actually limit FS memory consumption by limiting how many requests we
track from other peers (respective GSF-option had not been set; new code also
inverts the meaning of the bit, so it does not have to be set for peers but
rather is now set for clients to excempt them from the limitation)
block data migration based on lowest discarded expiration of content in
the datastore to avoid wasting bandwidth on migrating content that is just
instantly discarded again anyway
limit lifetime of migrated content to at most 1 year
fixing #2014
improving results seen communication to consider which exact keywords a
particular result has been seen for so far
-extra asserts just to be helpful in case #2000 resurfaces
-check for NULL
-do not crash on exit if we are not connected to statistics while trying
to submit mandatory values; just force instant connect
-not all uris are ksks
-fix
implement watch_cancel function
-hopefully fixing #1997, also renaming function to avoid double-use of
'transmit_task' identifier; minor clean up of assertion code
-properly use DLL apis where appropriate
do not run with full on NSE during testing
-static
-minor cleanup, adding HAVE_SETRESUID check
Adding optional compiler and linker hardening options as per suggestion
from Jacob
-undo
fixing 2012: network structure alignment now forced to be correct even on
W32 using #pragma pack from gcc 4.x
add version reporting for MHD to report.sh
-fix linkage
-fix
-fix uninitialized scope ID
-fix uninitialized sin_zero
-fix
-allow logging here
-LRN/CG: extract keywords from file names (#2032)
-LRN: new charset conversion functions / #2031
-LRN: add docdir to installation get path
-LRN: fix crash if name is not available
-LRN: neaten up, make debug log calls use DEBUG level
-install copying to share/doc/
-fix 2034, request can be cancelled in which case rh might be NULL here
-log to files by default
fixing migration stop delay calculation, largely by first calculating
datastore load correctly and then by better distinguishing between datastore
full, datastore timeout and success and finally by adding per-peer tracking of
the current block interval to adjust to repeated undesireable behavior. See
#2029.
-better logging
-maybe not the best default for now
-LRN: fix normal child death restart.
-remove code for keyword caonicalization/normalization, makes no sense in
international application, normalization methods are questionable to begin with
-also break at some other tokens
-fixing #2036
adding GNUNET_FS_time_to_year function to FS API
-fix
-dce
-version bumps for release
-cleaning up for disting
-cleaning up after mwachs
-more missing files for extra dist
-LRN: better size mismatch reports
-LRN: Also break on commas
-LRN: fix syscall name
LRN: normalize keywords (decapitalize, split) using libunistring - #2052
-add test for libunistring
-LRN: use FPRINTF -- #2051
-LRN: use FOPEN/CLOSE instead of fopen/close - #2051
-LRN: utf8-ization, #2051
-LRN: plibc header update - #2051
-fixing format strings
-LRN: OFF_T/off_t patch for 64-bit files on W32
-LRN: fix for OOTSD builds
-LRN: fixing issues that slipped through earlier
-LRN: ino diff
-LRN: 64-bit offsets for file locking on W32
-use test for libunistring from gnulib
-add to dist
-trying to pick better order of includes
-add revision check for libunistring
-I guess this can happen, but should be incredibly rare (or could be
provoked by attacker); let's just count instead of break - #2055
changing scheduler priorities to revert to DEFAULT instead of inheriting
parent-task priority unless something else is explicitly specified
-moving DNS structs to gnunet-dns-parser.h
-moving DNS code into its own directory
-ignores
-some code cleanup
-rename
-copyright
first quick hack to extract an initial DNS service API
-comments
-fix
dns hijacker code review
-small steps towards saner DNS API
-moving dns plugin to dns directory
-again moving towards DNS API sanity
-again moving towards DNS API sanity
adding GNUNET_OS_start_process_vap function
adding new GNUNET_HELPER_ API for communication with (SUID) helper
binaries via stdin/stdout using standard GNUNET messages
-adding new GNUNET_HELPER_ API for communication with (SUID) helper
binaries via stdin/stdout using standard GNUNET messages
-using new GNUNET_HELPER_ API inside of VPN, also eliminating bugs with
partial writes as part of this
-dce
-license
adding crc16 to gnunet_crypto_lib.h
-vpn should use crc16 from util
-vpn should use crc16 from util
-merging gnunet-daemon-vpn.h into gnunet-daemon-vpn.c
-dce
-minor cleanup
-minor cleanup
-copyright
DNS helper for DNS redesign
-indent fix
-comment fix
-fix shutdown order
-implementing new DNS client API
-implementing new DNS client API
-implementing new DNS client API
-implementing new DNS client API
-implementing new DNS client API
-dns not dht
-starting with new DNS service code
-starting with new DNS service code
-starting with new DNS service code
-towards dns service
-notes on the state of the world
-moving DNS API into the new direction outlined in my last comment
-fixing compile errors in tests
-more DNS service hacking
-more DNS service hacking
-fix
-misc DNS related bugfixes
-bugfixes and debugging
-ugh, found the bug
-remove debugging code
-dns parser API
-dns parser basics
-dns parser basics
-finishing DNS parser, alas untested
-bugfixes
adding missing file
-dns parser works now well-enough
-options
-formatting patch from GArik_
move monkey man page to monkey
-LRN: fix getopt
-bugfixes
-dns API improvements, towards serialization
-finishing dns packet serialization -- albeit untested
-adding tool to test dnsparser.c
-stuff
-try converting doubles to network byte order
-fix
-docu
-ignore
-bugfix
-debug
-addressing FIXMEs and adding statistics
fix crc16 prototypes
-remove useless casts
-towards new exit daemon
-ignores
fix
-const
-avoid warnings
-working on exit data structures
-late night code cleanup
-completing first round of the big refactoring
-document FIXMEs, fix some minor issue
-vminko: fix for #1930
-addressing some of the minor todos
-finshing first round of re-implementation; missing structs should be
done together with entry-component
-thoughts
-document libunistring dependency
-draft for service API to new VPN
-importing from gnunet-daemon-vpn to gnunet-service-vpn what can likely
be re-used
-moving tcpip structs to joint header, some more work on vpn service
-some more hacking on data structures, documenting open issues
-ignore
-LRN: define IPVERSION if not yet defined
-LRN: set bulk flag
-vpn api skeleton
-defining IPC messages for VPN
-defining IPC messages for VPN
-skeleton code for gnunet-vpn command
-rem
-more file name fixes
-add missing comments, expand error checking
-finishing gnunet-vpn tool and man page
-finishing gnunet-vpn tool and man page
-finishing first implementation of VPN client library
-finishing first implementation of VPN client library
-towards handling client requests
-also handle client service redirection requests
-implement local address allocation
-minor fixes
-LRN: my experimental block download time measurer. The results can be
found there - [1].
-fix
-fix
LRN: make disk iterator start return GNUNET_SYSERR if run on empty
directory
-LRN: make file filename alterable and initializable
-LRN add missing dependency
-defining proper structs for vpn-exit mesh communication
-dealing with new TCP message formats
-dealing with new UDP message formats
-towards having mesh messages in VPN
-generate messages in new mesh format
-parsing for UDP IPv4 replies
fixing bug to ensure that we properly descend into deep directories for
the various publish start/stop/suspend/resume event callbacks
-building IPv6 UDP reply messages for TUN
-building IPv4 TCP reply messages for TUN
-building IPv6 TCP reply messages for TUN
-updating todo notes
-implementing shutdown code
-fix
-LRN: fix fip_signal_start recursion
-fix
change default configurations on systems with UNIX domain sockets to NOT
specify any port for TCP-based IPC (and interpret that as no TCP-based IPC
desired), as we can and want to use UNIX domain sockets in this case by default
-ignore
improving code and build system to be in line with gnunet access control
model for services as described at
https://gnunet.org/gnunet-access-control-model
LRN: add function GNUNET_STRINGS_get_short_name to get basename
-LRN: fix callback arguments typo
LRN: new pipe creation function GNUNET_DISK_pipe_from_fd to wrap
existing file descriptor pair
LRN: skip short keywords when generating keywords automatically from
metadata
LRN: new threaded directory metadata scanner
-LRN: use correct character counting, instead of byte counting
-removing for now as it does not compile on POSIX systems
-file was missing, forgot to add earlier
-LRN: fix pipe writing and progress write
changing exit helper code to automatically do the network configuration
for an exit node (by running sysctl/iptables commands as necessary)
-changing exit helper code to automatically do the network configuration
for an exit node (by running sysctl/iptables commands as necessary)
-finishing first draft of new VPN implementation to the point that it
might run in theory
-sane defaults
-fixing crash, notes
LRN: enable more fine-grained control over blocking/non-blocking pipe
operation
-LRN: use blocking writes for scanner
-LRN: use signal pipe instead of a semaphore
-implement blocking write / non-blocking read for non-W32 systems
-LRN: harmonize dirmetascanner, plus fixing compile errors on GNU
-LRN: use filescanner in gnunet-publish
-adding some rudimentary logging
-adding some rudimentary logging
-basic setup/tear down for statistics in mesh/vpn
-adding statistics to VPN
-adding statistics to EXIT
-fix
-initialize len field
-initialize msg field in tnq
-removing legacy dns/vpn/exit code and renaming -new versions to current
-removing legacy dns/vpn/exit code and renaming -new versions to current
adding Ludo's gnunet-download-manager.scm back to SVN HEAD
-skeleton for new PT daemon
-expanding draft code
-finishing first implementation of protocol translation daemon
implemented new protocol translation daemon (#2063)
-fix build error on platforms that require -lpthread
-21 what -21
-more stats, more logging, adding access control checks
-sanity
-rename field
-track tmq queue length
-bound #messages queued, bound #destinations and bound #tunnels
-fixing #2069
-documentation related to #2069
-fix #2070
-fix #2079
-fix log call
-check NULL
-extra checks
-document expectations better
-rename header
-starting TUN lib
-renaming TUN structs to follow namnig conventions
-move IPv4 header initialization to tun library
-move IPv4 header initialization to tun library
-ignores
-move IPv6 header initialization to tun library
-move IPv6-TCP checksum calculation to tun library
-moving remaining checksum calculations to tun library, fixing #2066
-adding testcase
-fixing #2082 -- allow iptables/ip/sysctl to also be in /usr/sbin
-hints for stream lib development
implementing GNUNET_TESTING_get_peer_identity (addressing #2083)
-adding DNS exit-from-mesh functionality to gnunet-service-dns
-adding support for routing DNS requests via MESH to PT daemon
-fixing #2084
-need bump due to #2018
-fix renumbering crash
-initialize state_key to fix #2089
-partial fix to #2090
-fixing generation of responses from exit
-off by alen
-fix off-by-tcp-header size
-fix off-by-tcp-header size
-fix
-debug
-fix
-use different message types for different directions for TCP DATA
-fix getting output from sysctl into service-helper interactions
-mark function that should never be called as such
-adding ICMP support to TUN library
-adding ICMP support to exit daemon
-add support for receiving VPN_ICMP_TO_VPN messages
-add support for sending ICMP packets to service/internet
-misc bugfixes for ICMP
-ignore
-document ICMP MESH messages better, discard ICMP payload on
exit->mesh->vpn path for certain ICMP types
-avoid sending useless ICMP payloads and add ICMP PT support for sender
side (vpn) of direction vpn->mesh->exit
-implementing ICMP PT and type checking on VPN->TUN link
-make up ICMP payload for error responses on VPN->TUN link
-implementing ICMP PT and generation if ICMP payloads on EXIT->TUN link
(for both service and Internet exit)
-LRN: Give-debug-log-message-DEBUG-level
-LRN: W32-pipes-are-overlapped-by-default-now
-LRN: Pipe-errors-will-trigger-ready-to-read-event
-LRN: Adapt-fs-tests-to-new-gnunet-publish-verbosity
It-might-be-NULL
use noreturn macro for GNUNET_abort to help gcc and others
-fixing #2100
-a first VPN testcase
-do not compile vpn/exit/pt/dns on non-Linux systems
-fixes
-fixes
-fixes
-bugfixes
-bugfixes
-fixes
-fixes
-fixes
-fixes
-do not leave th non-NULL on reconnect
-fixing disconnect shutdown sequence
-test now passes
-handle mesh errors more gracefully
ign
-dce
-dce
-fix
-do warn
-dce
-dce
-fix use of extractor
-fix
-fix
-fix
-fix
-fix
-assert
-check cmd line argument
-check return value
-check return value
-fix
-fix leak
-fix leak
-cleaner
-fix leak
-fix
-fixfix
-doxygen fixes
-undo 19365
-leak
-fix
-do not abort waitpid just because of SIGCHLD
-always check that WE can execute
-terminate test after a while, check for SUID helper
-check for SUID helper
-quiet exit helper
-use proper backoff to make VPN connect/start faster
-fix call to stats destroy
-indent
-picking better default networks for dns/vpn/exit
-expanding tests to cover 4-to-6, 6-to-4 and 6-over-4 as well
-ignore
-call client last, so that if the client calls disconnect, it does not
mess us up
-nicer
-allow binaries without SUID if we are root
-fix IP header length calculations for IPv6
-fix
-fixing #2097
-follow naming conventions
-fix brainfart
-clarify
-fixing 2114
-removing outdated docs, some has also been added to the website instead
-check tcp off value
-check tcp off value
-check tcp off value
-fix tcp header definition
-fix IPv6 packet construction
-fix ret vals
-fix address order and mask calculation
-fix mask calculation
-fix
-sane default
-set tcp header
-actually copy payload
-facilitate debugging
-less verbose
-allow NULL
-warn if permissions are insufficient to run test
-fixing #2116
-more consistent naming
-log resolver errors always
-fix
-fix dead scheduler task cancellation on CTRL-c
-fix error reporting, remove unnecessary syscalls on some paths
-code deduplication, doxygen fixes
-doxygen
-renaming symbols to match naming conventions
-change callback argument from char to int, as that is cleaner for a
boolean; document API some
-cleaner version of sharetree trim code; should be fast enough to run
synchrnously
-implement tree freeing
-cleaning up dirmetascan code, still failing for me
-oops
-fix
-misc bugfixes
-towards resolving thread-issue
-finishing new helper
-towards external meta scanner process
-towards external meta scanner process
-enable helper shutdown during mst callback
-fix shutdown issues on error
-make mst more reentrant
-fs publish should basically work now
-fix
-fix
-misc fixes
-del
-check if /dev/net/tun exists and skip test otherwise
-fix
-all meta data extraction should happen in helper, removing from API
-add test for unsupported AFs (#2123)
-adding a few more packed statments
-renaming some fields, fixing byte order issues
-docu
-keep split
-simplify
-better error reporting
-trying to fix double-start issue on sparc
-fix
-possible fix for 2124
-be-fix
-be-fix
-docu
-also run if system does not support IPv4 or IPv6 at all - #2123
-improve byteorder macro checks for FreeBSD
-missed one
-extra size checks
-fix endian test
-ugh
-fix
-try with finished
-do not try VPN tests on FreeBSD and other non-GNU-ish systems
-fix for W32 pipes from bratao
-fix comment
-nicer
-barato's byte order patch, might even help with #2130
-add missing parens
-adding build system basics for stream library
-misc stream hxing
-misc stream hxing
-also user gnunet-original filename here more consistently
-always signal download start before 'completed', even for 0-byte files
-eliminate printing warning on MAGIC missmatch (#2138)
-fix generation of duplicate publish-stop events
-speed up logging test, fix logging testcase to always log at DEBUG level
adding -m option to gnunet-transport to enable monitor mode (see #1972)
-applying patch from vminko to fix #1972: adding support for continuous
transport-level connection monitoring
-new hostlist
-nl
-trying to fix crash from #1972 report
-fix
-patch from #1972 to display disconnects instead of exiting
-fix #2142
-updating DNS code to use GID match instead of source port match,
introduce source port randomization
-fix
-stylistic fixes
-fixing source port randomization for DNS service
-ignore
-fixing #2139
-doxygen fixes
-generate PUBLISH_STOP event later to avoid use-after-free
-do not tie verbosity to debug levels for fs tests
-unschedule found block request on resume to fix #2133
-fix
-log more details
-ICMPv6 has a different protocol number than ICMPv4, fix code accordingly
-fix
-fix
-GArik: fix typo
-GArik: size_t is unsigned -- DCE
-GArik: remove redundant comparison, also use %p for pointers
-GArik: fix assertion check
-GArik: fix infinite loop
-typo
-fix testcase
-add missing packed
-fix
-use GNUNET_free instead of free
-fix leak
-fix leak
-extra checks
-fix leak
-check return value
-fixing leak of transmit context (cov 10128)
-simplifying code
-use loop when looking up session...
-fix
LRN: Use binary mode on W32 (lol -CG)
-comments
-comments
-draft for new API
LRN: don't free memory on the stack
add GNUNET_FS_namespace_dup API call
-LRN: use GNUNET_strdup with GNUNET_free
-LRN: Fix-config-file-in-compliance-tests
-call continuation last
-avoid using add_continuation if possible
-die ifdefs, die
make all (?) asynchronously operating FS operations actually cancel-able
-bump
-dce
-use search URI, not result URI
-do not try to desearialize search directories
-sync after getting URI on publishing to resume properly; also bugfix...
-sync search results upon adding/removing downloads
LRN: check for tc being NULL
LRN: Iterate-over-a-copy-of-ksk-when-removing-items
LRN: Do-partial-serialization-in-fs-publish-helper:
-clarify
-plugin api design for namestore
-adding namestore directory:
-skeleton for sqlite database plugin
-DB API updates, creating tables and preparing statements
-DB API updates, creating tables and preparing statements
-again improving API, starting to implement some of the sqlite
interactions
-fix
LRN: Escape-trailing-slash-when-spawning-W32-process:
-finishing first pass at implementing sqlite statements for namestore
-fixing #2155
-increase timeout to help sparc bot
-build new namestore
-no verbose
-fix
-fix
-initial testcase
-extended test
-initialize udp header...
-one more
-ign
-namestore might just work now
-dce: can never be NULL
-stuff
-testcase harness
-ign
LRN: two directory patches change the way "is_directory" is evaluated.
Now it must be GNUNET_YES for the execution flow to switch to a branch where
something is considered to be a directory. The reason for that is that some
functions might return GNUNET_SYSERR when asked whether something is a
directory or not. Checking this value as "!= GNUNET_NO" will produce positive
result, even though it's not a directory.
-doxygen fixes
-fix type
-extended testcase plus misc fixes
-more bugfixes, more testing
-updated namestore API
-revised plugin API
-moving towards new plugin API
-finishing 1st pass on new plugin
-fix
TG: attached are the following patches for GNUnet:
-fixing bugs and testcase, plugin now works for updated API again
-fixing compile errors due to ns api change
-updating readme and configure
-ignore
-ignore
-fix
bratao/LRN: Use-bigger-buffer-for-EnumNICs3_results
-fix
-fix
-fix
-fix api
-updating PO files
-fix
-fix
Igor/CG: various minor updates to man pages
-extra logging is not the same as verbose
-eliminating #if DEBUG checks
-verbose is independent of logging
-verbose is independent of logging
-ignore
-usually should print nothing
-some more #if DEBUG elimination
-dont print useless stuff
-try larger timeouts for sparcbot
enabling use of pipes for signal communication also on UNIX to enable
future integration with Java services
-ignore
-novpn
-make disk IO blocking, otherwise opening signal pipe may fail because
child did not have time to start
-do use non-blocking opening of the pipe FD, and try again if it fails
-fix
-version bumps in anticipation of release
-monkey files should be with monkey now, not GNUnet
-fix
-fix
-fixing bogus assertion and improving error handling
add gns
-add stream
-commenting out testcase -- wrong c filename?
-commenting out testcase -- wrong c filename?
-fix
-fix
-autostarts
-fix
-ignore
-fix option test
-fix
releasing GNUnet 0.9.2
-fix compiler warning
-make no assumptions about the value of EINVAL
-some code cleanup
LRN: W32 pipe name generation needs random
LRN: Apparently cleanup is not for W32
-LRN: Add missing declaration
-LRN: fix gns makefile
-po dist changes
-make gcc happy
adding GNUNET_CRYPTO_setup_hostkey to setup a hostkey ahead of time,
using this function in the VPN testcases to avoid timeouts in cases where
creating a hostkey just takes too long --- such as on our UltraSprac
-LRN: fix
-LRN: fix
-LRN: fix
-LRN: fix
-LRN: set control_pipe to NULL if not controlled
-fix memory leak
LRN: Enforce GCC bitfield layout for some structs on W32
-LRN: fix fprintf
-LRN: Refer to own library by a relative path
LRN: updates to pseudonym API from #1952, change pseudonym management
-fixing sparc fun
-new api
-stuff
-use new serialization/deserialization API
-minor fixes to arm service list API (#2141)
-key mgmt
-fix
-LRN: gns block plugin depends on namestore
-LRN: Without this patch namestore tests can't even load the plugin. With
it namestore tests (at least some of them) fail two assertions, and then crash.
Not sure which is better.
-LRN: Namestore doesn't work on W32 without a port
-make gns non-experimental
-starting with a simple domain registration WWW service
-documenting todo, minor fixes
-stuff
LRN: adding generic functions for conversion of binary data to ascii and
back: GNUNET_CRYPTO_string_to_data and GNUNET_CRYPTO_data_to_string
-adding conversion of public key to string and back
-renaming
-fix
-doxygen fixes
adding gnunet-rsa, a new tool to create RSA keys and to print the public
key
-fix
-finishing gnunet-gns-fcfsd
-fix
adding gnunet-gns, a new tool for zone manipulations
-adding options to gnunet-gns
-fix
make gnunet-service-statistics not exit on external shutdown signal as
long as there are connected clients to be managed
-adding testcase for FS probes, minor bugfixes related to probes
-add priorities to download queue
-ignore
-zone to name --- API mentioned in #2194
-zone to name --- API mentioned in #2194
fixing #1927 by further limiting the time a download probe can be active
at a time in the download queue; this is equivalent to it having a low priority
-use sane defaults
-adding GNUNET_CONFIGURATION_load_from
adding fs/ to all default directory names relating to file sharing
-misc bugfixes
-misc bugfixes
-fix
-implementing zone_to_name
-parsing type and common values in gnunet-gns
-parsing expiration and some option checking in gnunet-gns
-more work on gnunet-gns
-LRN: add dependencies to namestore
-renaming gnunet-gns to gnunet-namestore
-more renaming
-reviving gnunet-gns for shortening
fix
-LRN: file checking routines
-LRN: experimental HELLO URIs
-fix
-ignore
-disabe namestore
-do not load namestore
-do not load namestore
-fixing #2196
fix
-fix fix fix
-LRN: fix peerinfo initialization
breaking stuff
-fix
-PKEY records should contain hash of public key, not public key
-additional namestore conversion APIs added
implementing more of GNUNET_NAMESTORE_value_to_string
-fix year'
-done
-disable extractor should also imply not using the filename as a keyword
-better interpretation of instant downloads from local database /
existing block
-misc major bugfixes in download code wrt reconstruction and chk index
calculations
-trying to fix #2198 -- or at least a related bug
-more chk index calculation errors, ugh
-a note to LRN
-fixing leaks/use after free
-brk
-fixeS
-fix
-fixing #2201
adding code to measure and report shutdown time for services to
gnunet-service-arm at INFO level logging
-additional tests for process starts
-removing 'proc' and replacing it with a different proc for each of the
different types of processes being started so that we can actually figure out
which ones are running
-fixing #2190 by actually stopping all child processes cleanly
-fixing #2195
implementing removal of KBlocks during unindex operation (#1926)
-clarify
-do not call uri_to_string with NULL pointer
-unindex needs to decrypt KBlock
-fix
-avoid problems with partial reads by using 1 byte for signals
-closing of control pipe is no error
add support for stub-mode for transport plugins
-LRN: Fix kill
-starting cleanup of peerinfo, crashed before by loading plugins which
failed if ports were in use, etc. --- not done
-LRN: Make gnunet_testing.py W32-compatible
-fix
-LRN: Don't call a NULL StA function
-fix
-LRN: Generalize ip-address str->addr conversion
-LRN: Add UDP StA function
-cleaning up get uri code
-LRN: use plibc SSCANF instead of sscanf
-no pblic in SUID please
vminko: implementing BINDTO option for hostlist service (#2140)
-towards addressing #2206
-fixing #2206/2212
-fix
-use zkey extension
-adding capability to install static list of HELLOs with the original
software to avoid need for hostlist server accessibility
-adding gnunet-hello tool to bump expiration times
-set expiration time for included HELLOs to infinity
-support truncated HELLO names to work with 99-character per filename
limitations of tar
-ignore
-fixing some of the issues from #1939 -- untested
-fix
-fix #2220
add -P option for printing peer identities with gnunet-rsa
-moving to port 80 as well
-fix newline
-fix testcase, reintroduce timeout
-never a clear warning, use INFO
-leak
-ease life of static analysis tools
-check return value
-check return values, fix leak
-uuuunet
-init
-doxygen fixes, add 'const'
-easier
-update config to make it ARM-able
-fix
-clean up HTML
-fix
fix
-fix
-more cleanup for #1939
-fixing misc. bugs and making wlan helper auditable, resolving #1939
-doxygen
-doxygen
-clone elimination, adding some documentation
-removing duplicate code
-removing code duplication
-removing duplicate code
adding API for short (256-bit) hash codes
-LRN: Have stat64 on W32 unconditionally
-LRN: calculate file size for single files when needed and use
GNUNET_DISK_file_size instead of STAT
-doxygen
add configure option --enable-javaports to open ports of services with
Java bindings by default (#2228)
-fixing minor compilation errors and warnings
-use short hashs in fcfsd
-allow registering one name per zone key in fcfsd
-LRN: Add dependencies to gnunet hello and service namestore
-fix
creating mysql helper library for the various mysql backends
-for max
-fix
-abstracting out mysql helper library
-end in helper library
-using new mysql helper library also with datastore
-ignore
-fix
-fix
-allow deletion without specifying expiration
-doxygen
-docs
adding libgnunetpostgres for shared postgres functionality between
postgres datastore/datacache backends
-indenting
-updating POTFILES.in
-check return code
-leak
-doxygen
-using cleaner message types for WLAN-HELPER IPC
-regex
-doxygen
-eliminating duplicate code
-eliminating duplicate code
-eliminating duplicate code
-eliminating duplicate code
-note \& ignores
-fix
-UID match is sufficient if GID match is required
-fix README to match default ports and proper config values
-extending defaults
-nullcheck
-fix error handling during publishing
-fix error handling during publishing
fixing issue with gnunet-publish not closing files early enough when
publishing directories with more than FD_MAX files (#2239)
-fix #2232
-fixing advertising constants, code cleanup
-fixing advertising constants, improving docu
-fix printing
-fix segv
-fix comment
-dce
-use define, fix timeout
-increase timeout
-code cleanup
renaming gnunet-core-list-connections to gnunet-core, adding man page
implementing rekeying, some code cleanup
-ignore
-fix
-fixing #2243
-doxygen
-removing legacy #ifdefs
-removing legacy #ifdefs
-removing legacy #ifdefs
-fix
-do not warn if stat fails, do not output DEBUG messages in test
-removing legacy #ifdefs
-fix
-removing legacy ifdefs
-make error message on timeout more consistent to allow tests to pass
-fix
-dot
-fix
-fix assertion, documentation, getting dummy to be closer to work on W32
-w32 fprintf fix
adding configure option to run GNUnet with monkey
-allow NULL for emsg
-fix
-ignore
-fix docu
-dce
-fixing outbound notifications in core API
-joint hdr hacking
-fixing core traffic monitoring assertion, code cleanup
-fixing 2253
-removing dead option
-improving configure test for MHD based on suggestions from cy1 on IRC
-removing dead API call
-removing 2nd argument from GNUNET_CLIENT_disconnect as it was virtually
always GNUNET_NO --- and all other uses indicate design problems
-removing obsolete argument from GNUNET_CONNECTION_destroy
-rename fest in connection.c
-simplifying destruction of connections, cleaning up server API -- all
for #2274
-fixing #2274 -- eliminating GNUNET_SCHEDULER_add_after
-fixing #2274
-dold key
-doxygen
-removing legacy debugs
-fixes
-do not test for prereq done, almost always set now
-fix
-fix
-fixes
-fix
-fix
-slightly better
-fix
-undo
-fix shutdown of server
-remove ATS 0-termination in NotifyTrafficMessage (#2275)
-there was no 0-terminator, fixing comment
-ignores
-misc offline hacking / code cleanup in util
-LRN: fix fprintfs
-LRN: remove references to non-existent files
changing bloomfilter to allow GNUNET_CONTAINER_bloomfilter_init with
sizes that are not powers of 2 -- GNUNET_CONTAINER_bloomfilter_load will
continue to round up to power of two
-doxygen
-doxygen
-adding improved test for libmicrohttpd
-improving doc
introducing soft shutdown concept for services; during soft shutdown,
services that are still managing non-monitor clients continue to run until
those clients disconnect; however, the services do stop to accept new
connections (will stop listening); soft shutdown is now used by ats, transport,
peerinfo, namestore and most importantly statistics; this should fix #2197
-fix
-test for LRN:
-fix
-eliminating legacy DEBUG ifdefs
-improve debug statement
-workaround tests hanging
-fix
-removing legacy DEBUG ifdefs
-fixing leaks and error messages
-ignore
-fix
-fix
-fix
-not -1 when using % instead of &
-trying to fix assertion from freebsd buildbot
-trying to fix test shutdown sequence if database not setup
-fix
-fix
-fix
-fix
-fix
-fix configs
-must be NONE
-fixing shutdown sequence
-fix
-simplify logic, connections no longer die during shutdown
-fix
-cleaner warning
-fix
-better error reporting
-fix
-minor fixes
-disable pow
-trying to resolve alignment issue on sparc
-doc
-trying to fix build errors on old gccs
-LRN: misc patches / fixes
-LRN: special W32 function checks; It forgoes AC_CHECK_FUNCS for some
functions in favour of manual AC_LINK_IFELSE, which is necessary to check for
API with stdcall interface (i.e. winsock).
-skip duplicate addrs
-handle timeout better
-handle shutdown better
-fixes
-fix
-fixes
-fix
-convert assert to break for now
-fix
-extra assertion
-fixing persistence cleanup
-fixing #2229
-static
-disable histogram, check return values
-fix
-fixing #2277
-LRN: fixing #2286
-updating AUTHORS, anyone else missing?
-dead define elimination
-fix build error
-document ignored return value
-fix
-fix
-fix
-new testing service API design
-further revisions on testing API and new testing IPC message definitions
-fixing #2289
-fixing #2290
-fixing compiler warnings
-fixing #2293
-fixes
-fixing #2294, #2295 and #2296
-improving error reporting on #2295
-improving error reporting on #2295
-improving error reporting on #2295
-improving error reporting on #2295
-improving error reporting on #2295
-make doxygen happy
-doxygen fixes
-fixing #2298
-ignore
-enable lm building
-applying fix based on fixed #2210
-fix
f-xi
-towards fixing #2299
doxygen
-fixing #2299
-fixfix
-draft of new testing code
-fix compilation of new testing draft
-fix
-fix
-fix
-code cleanup
-proper branch order
-extra debugging code
-fixes:
-cleaner
-extra IPC checks
-spacing
-fix expression order
-fixing #1969
-doxygen
-renaming new testing code to testbed
-renaming testing/testbed buildsystem updates
-ignore
-nodebug #2300
-build issues'
-draft of new low-level testing API
-draft of new low-level testing code structure
-doxygen fix
-draft structure for operation queues
-fix
-removing legacy ifdefs, fixing log statements
-fixing #2302
-fix compile error
-removing odd error message, fixing english, removing dead code
-changing message from ERROR to DEBUG as it is normal
-changing message from ERROR to DEBUG as it is normal
-fix linkage on FreeBSD
-fix linkage on FreeBSD
-eliminating legacy defines
-doxygen
-disabling failing test
-fix
-cleanup before and after
-noverbose
-removing bogus/legacy debug statements, not passing -L DEBUG to arm
anymore
-more to remove
-fixing #2170
-linux only
-cleaner
-fixing #2310
-fix
-fixing #2309
-fixing #2309
-sparcbot likely simply needs a longer timeout
-fixing error messages, making quotas a bit more lenient as we can
clearly -- based on timing -- be off by 1/3, see #2308
-report error, increase timeout for sparc
-avoid #ifdef, limit to tcp to avoid importing bugs from transport into
vpn tests
-improving error messages of failing test
-check return value
-fix uaf
-fix test if nls is not working -- OSX buildbot says hi
-only used on UNIX
-linux only
LRN: creating watchdog helper binary for W32
-fix
-ugh
-docu
-fix: min delay 50ms, do not skip mindelay if fragmentation block only
consists of 1 msg
-fix: min delay 1ms
-fix
-allow helper send to be cancelled
-log
-cleanup
-make helper start more robust
-clean up WLAN plugin
-fixing connection problem in WLAN. Someone should explain to me why the
other plugins work without this
-activate
-trying to fix #5817
-stylistic fixes
-comment
-fixing #2329
-fixing #2332, plus some minor refactoring
-fix #2331
-fix #2319
-trying to fix #2335
-convert SLL to DLL
-less crack smoking required: finding pos==client in DLL is pretty
useless now...
-fixing #2335 some more...
-improved fix for #2336
-fix leak
-better
-allow outbound monitoring of self-loop messages without bitching
-fixing UDP fix
-fix double free
-fix leak
-fix uninit
-trying to help/fix #2248
-note on issue
-fixing leak
-fixing DBZ
-fixing leak
-typo
-install dummy, even if it is just to make testcases pass
-fixing misc issues in stream, including compilation errors on bots, but
more
-trying to fix #2342
-fix
-eliminating cont-calls check, hard to count exactly with frag
-implementing WLAN string to address
-clique also causes unattributable trouble, commenting out for now
doxygen
-minor stylistic fixes
-removing confusing legacy ATS API
-regenerating PO files
-undo
-minor cleanup
-indenting
-major rewrite of gnunet-service-transport_neighbours state machine
-fix segv
-fix crash on error
-fixing #2347
-doxygen
-fixing #2345
-help safey
-fix use after free on shutdown
-fixing misc seaspider parser errors
-help seaspider some more
-help seaspider some more
-help seaspider some more
-really remove
-adding checks to avoid #2352 segfault
-code cleanup
-typo
-cleanup
-fix
-cleaning up TCP logging
-dead field
-fixing 2360
-fix loglevel
-use MHD_LONG_LONG
-code cleanup
-code cleanup
-more logging
-fix bogus assumption on my part
-fixing testcases, cleanup
-improving fragmentation performance
-fewer stats
-do switch statement cases in order of enum
-check for sks results --- #2352 first bug
-fix
-fixing 2352
-code cleanup
-cleaner linkage
-code structure
-some code cleanup
-fix #2378
-fixig #2377
-fixing testing testcase build issues
-fix warning with lockmanager port in testcases
-better error message
-disable lm during integration tests
-more uniform statistics naming
-more uniform statistics naming
-more uniform naming of statisitcs
-ignore
-undo
-no lm during test
-fixing #2340
-reduce too high keepalive frequency to more sane levels
adding GNUNET_SCHEDULER_add_read_net_with_priority
use GNUNET_SCHEDULER_add_read_net_with_priority instead of constructing
fd sets in server with only one active listen socket
renaming GNUNET_TIME_relative_get_forever and
GNUNET_TIME_absolute_get_forever methods, adding underscore, to make it clear
that the respective #defines should be used instead; replacing use of direct
function calls with respective macros where applicable; adding additional
GNUNET_TIME_relative_get_xxx-functions to avoid calls to
GNUNET_TIME_relative_multiply, which turn out to have gotten
performance-relevant
-typo
-minor code cleanups
-fixing #2373
-fixing #2365
-fix kill logic (#2381), interpretation of flag was inverted
-doxygen
-fixing #2382, also some code cleanup
-removing bogus const
-removing debug code that might be problematic
-fix memory leak
-dce
-install hostkeys
-better errno handling
-minor cleanups in testing
-javaport
-adjusting makefile to install NSS libraries of GNS
-LRN: make watchdog always kill its children
-fix assertion failure
-allow installation even as non-root non-sudo user to succeed with new
nss code
-fixfix
-document SUDO build implications
-actually start both DHTs
-test for nss.h
-fixing speedup test
-ignore
-ignore
-use fewer tasks for 2x speedup during test
-fix leak
-fixing assertion failure on sparcbot in line 2679, adding comments for
possible issues that remain
-removing assertion that seems bogus
-increasing timeout to hopefully help sparcbot
-fixing compiler warnings on FreeBSD
-fix nss.h test
-code cleanup and better error messages for failures
-code cleanup, more precise error messages
-fixing port choices that conflict with lockmanager
-LRN: fix dht tests
-dce
-also run fs by default
-code to debug #2390
-more logging
-trying to fix #2391: limit connect rate from topology
-fixing #2391
-doxygen
-fix port
-fix #2393
-trying to make issues such as #2389 less likely, but I doubt this fixes
it
-fixing minor bugs, code cleanup
-fixing minor bugs, code cleanup
-bratao: there is one little innofensive patch.
http://pastebin.com/iRharWUm , On W32 we waiting 10 microseconds on the select,
and its completely unecessary. Some tests are 30% faster to me.
-cleanup
-ignores
-fix
-fix
-fixing build issues for disting
-disable
-adding GNS fcfs import script
-unnecessary code elimination
-fix syntax problem for Java parser
-bump versions
-minor code cleanups
-fix
LRN: Update plibc and utf8ization
-hopefully fixing #2390
-hopefully fixing #2390
-fix
-LRN: fix a warning
#2404, patch 01: configure support for GNU Hurd
#2404, patch 02: include sys/vfs on GNU
#2404, patch 03: include net/if on GNU
-fix
fix arm termination on W32 by enabling signalling pipe to be used
-minor edits
bumping required curl version to 7.21.3
-7.21.3 required now
LRN: new select wrapper for W32 which avoids busy-waiting
-fixing #2405
-fixing linker error for now by adding -lcrypto, will need to be fixed
later
-fixing #2408
-fixing #2406
-fixfix
-fixing #2402, #2403, #2399, #2398
-fixing #2412
-fixfixfix
-LRN: fix udp plugin on W32
-fixing #2413
-doxygen
-adding assertions to help debug #2403
-handle disconnect error on shutdown better
-fix
-simplifying core API (#2400)
-fixing #2400
-fixing compiler warning
-removing useless ifdefs
-moving old testing code to legacy location
-fixes
-updating config files
-fix
-use human-readable error message
-dont test new in old
-fix configs
-fix
-fix
-fix
-dequeue download on error, another attempt to fix #2403
-zero out
-also check cth_in_progress here
-code simplification
-logging
-fix logging, revert to break instead of assert
-misc fixes
-fix config
-queue size is now at most 1
-renamefest
-slightly better in case free/strdup'd value is aliased
-auto-load hostkeys, minor bugfixes
-auto-load hostkeys, minor bugfixes
-ignore
-minor fixes
-doxygen
-use mmap
-moving defines to top
-using new testing library for peerinfo test
-integrate mkdirp with testing
-cleanup
-using testing in datastore tests
-adding helper function to determine plugin names to testing
-using new testing api to determine plugin name
-cleaning up template
-starting with namestore test cleanup
-more testcase cleanup
-typo
-LRN: fix mkdtemp on W32
LRN: fix RAND_MAX on W32: MinGW stdlib.h defines it to 0x7FFF, but our
implementation is better.
LRN: Fix-test_group-interval.patch: Just a wild guess - you wanted to
find ids either inside the interval or outside of it. Although maybe you should
have used <= and >= in one of the conditions then? CG: only mrwiggles knows the
truth. Anyway, old testing will be gone soon enough....
LRN: new utf8 argv converter for W32, converting strings on command-line
to UTF-8 for all command-line tools.
-fix
-fix
-fix
-po
adding check for gnutls
-fix
-doxygen
-LRN: fix dv test dependencies
-first draft of postgres namestore plugin (#2157)
-configure testcase
completed postgres plugin for namestore service
-debug
-adding block duplicating detection to template for block plugins
-consistently use struct GNUNET_HashCode
-fixing #2420
-LRN: block plugin depends on block module
-LRN: kind of fix for gnunet-testing-run-service for W32
-fixing #2240, untested on W32
-minor optimization
-fixes and cleanup
-LRN: typo
-LRN: fix r21946 to work on W32
-do not store/keep HELLOs with expired addresses on disk (#1932)
-cleaning up VERBOSE and check nonsense in util tests
-cleanup
-document error possibilities better (#2416)
-fixing #2153, code clean up in container_meta_data.c
-use underscores for testcase names
-LRN: fix binary checker to use proper API
-use const
-nicer name
-needing yet another API to box stdin for scheduler reading, not ported
to W32 yet
-needing yet another API to box stdin for scheduler reading, not ported
to W32 yet
-testbed hackery
-check for NULL
-special treatment for ./binaray-name
-special treatment for ./binaray-name
-pogen
-typos
-LRN: implement get_handle_from_native for W32
-LRN: fix testbed dependency
-LRN: more dependency fixes
-cleaning up fs-download testcases
-ignore
-simplifying tests with new testing lib
-simplifying tests with new testing lib
-remove duplicate setsockopt calls; GNUNET_NETWORK_socket_bind does those
already anyway
-dce
-use port numbers that are less likely to conflict with source ports
-use testing lib
-use port numbers that are less likely to conflict with source ports
-LRN: dependencies
-LRN: don't use Z
-pass curl flags for vpn test
-LRN: don't duplicate testbed.conf
fix compilation with c++
-fixing #2426
-fix #2354
-starting harness for gnunet-auto-share:
-LRN: isatty equivalent for w32
-more skeleton work for gnunet-auto-share
-implementing 'determine_id'
-implementing read_state
-implementing write_state
-implementing freeing of state on exit
-adding logic for starting gnunet-publish
-more auto share code and config
-debugging gnunet-auto-share, now largely works
adding man page for gnunet-auto-share, updating man page for
gnunet-publish
-fixing #2431
-fixing #2397 -- start put a bit after starting monitoring to observe all
events
-reduce size of test to help sparcbot
-porting gnunet-testing command-line tool to new testing library
-actually use testing libs hostkey functions
-handle error
-fixes
-use testing to create hostkey during peerinfo tool test
-modify timeout values further
-minor cleanup
-fixing #2397
-fixing IO-error bug in namestore tests on buildbots, simplifying
testcases by using new testing API, eliminating dead code
-no longer old
-starting a bit with namestore_api cleanup
adding API for parsing absolute time
-towards enabling relative expiration times in namestore
-fixing testcases with new expiration code
-fixing some testcase builds
-fixing testcases with new expiration code
-fixing #2434, plus some code cleanup
-towards block plugin for mesh
-fix use after free, nicer loop
-fix leaks
-check misc return values
-check misc return values
-fix #2436
-working on stream
-ideas for stuff
-fixing #2439
-fixing #2439
-some namestore code cleanup
-code cleanup
-fixing namestore signature bug; all that was needed was to force packing
on the struct
-improving comments, towards implementing #2268
-towards better download prioritization
-towards fixing #2268: srv parsing/serialization
-finishing SRV implementation (#2268)
-some more work on probes
-branching cleanup
-work on #2437, plus misc minor bugfixes
-print switch target
-print use shortened unixpath also when connecting...
-compilation issue
-notes for regex api
-LRN: Mark session as completely disconnected later.
-promoting regex to non-experimental as mesh needs it now
-fix #2122 and #2035
-comment / naming cleanups
-comment / naming cleanups
-comment / naming cleanups
-passing total trust offered along (fixing #1369)
-renaming for consistency
-removing #ifdefs
-ignore
-fix
-use testing lib
-use testing lib
-use testing lib
-use testing lib
-use testing lib
-removing dead old code
-removing dead old code
-LRN: Change logskipping to use functions only: Remember these weird
"skip_log" functions in my gdb backtraces? I hope that this patch will get rid
of them. Shouldn't affect existing uses of the API.
-LRN: Lock the output on W32: Purely W32 patch. It was needed for some
threading-related debugging i did for select() wrapper. Shouldn't harm to have
it, in case i do logging from multiple threads again.
-LRN: Support [] (formatted to be pid) in logfile names: I needed it
for mesh debugging. Wrote this patch before your log rotation changes, and had
to hastily adapt it after them, so i'm not sure it actually works. Did some
quick tests on it, and it appears to be working, so i'll just send it to you as
is, and see how it turns out.
-LRN: Better process termination for W32: Avoids spewing warnings about
killing processes that are already dead. Also removes a FIXME.
-LRN: Poisoning: Poisons memory before freeing it. It uses msize() CRT
function, which (i believe) is not available on other platforms. This patch was
intended to make the output of the next patch (see below) more apparent.
-LRN: Check session target: It changes the Session object layout to
make peer identity its first field, always, for all sessions. Code in
find_session() can use this to check that the session belongs to the right
peer. test_ats_api_add_address always triggers this due to the way it fakes a
pointer to session object.
-document new requirements on struct Session
-LRN: Wincat: I've been experimenting with an alternative non-MSYS
buildsystem for a while, and GNUNET_os_start_process() test kept failing due to
lack of cat. Wincat is a minimal cat implementation in pure WinAPI. It's not
named "cat.exe" to avoid clashing with MSYS cat (which can lead to some very
weird issues).
-fix
-LRN: Don't free shutdown handle while it's in use
-LRN: Don't use freed ic
-LRN: Log timeouts that are bigger than DWORD; fix format string
-attempting to fix LRN's NAT connect issue
-LRN: no more pexpect
-fix
-cleaner
-LRN: really no more pexpect
-LRN: fix integration tests
-fix include installation
-fix
-pogen
-optimizing regex
-boundscheck
-some cleanup
-fixing #2459
-LRN: reverting earlier attempt to fix LRN's NAT connect issue
-LRN: gives CONNECT message a priority (stops waiting for ATS to provide
an address to connect to, and goes straight to checking the incoming address).
-LRN: remove unused variables
-extend test to cover iteration over empty store
trying to fix #2458, largely works
-clarify comments
-fix rest of #2458
-use sqlite
-fix
-fix
-fix error message for unix paths starting with 0-termination
-more debug messages
-cleaning up logging
-adding support for importing PKEY records from URIs
-allow blank specification of URI-only
-add tool to handle gnunet-uris
-ignore
-LRN: fix host list freeing in testbed
-man page for gnunet-uri
-cleaner
-fix
-fix
-\'optimize\'
-fixing the schanzenbug
-major cleanup and various minor bugfixes for namestore_api
-fix
-LRN: new mesh build fixes
-more namestore code cleanup
-namestore cleanup and fixes -- wip
-namestore cleanup and fixes -- wip
-more code cleanup in namestore
-more namestore service cleanup -- wip
-more namestore service cleanup -- wip
-more namestore code cleanup
-fix
-do need to install plibc and other headers after all -- #2470
-check all fields for matches before removing
-fix message types for namestore, eliminate conflict with LM
-fixes
-fixes
-more fixes to namestore
-fix
-display expiration times with records
-better time conversion code
-seems to only work on GNU//Linux like this
-fix eternity issue
-be lazy about connecting to VPN
-rel time allowed
-use proper time API
-update docu
-committing relative/abs expiration time changes from a few days ago
-LRN: unlinux some linux-only stuff
-LRN: check for .gnunet TLD explicitly fail if it is not
-LRN: document namestore contents for given types
-LRN: fix af test on W32
-fix gns record deletion
-LRN: cancel the task when lookup is aborted
-fixing #2478
-LRN: initialize rd to avoid having garbage in flags
LRN: More logging for namespace comparison: Changes
GNUNET_NAMESTORE_records_cmp from a simple if statement to a chain of if
statements, each of which will log the reason comparison failed before
returning FALSE, making it obvious why comparison failed.
-LRN: Portable memory poisoning: Uses GNU extension function as a
non-W32 equivalent of msize() (there are subtle differences, but they are of no
consequence). Also swaps the poison bits depending on endianness (i'm tired of
seeing 0xdf0adba everywhere instead of 0xbaadf00d).
-LRN: Correct time == 0 handling: With logging in namespace comparison
it is now clear to me that deletion fails due to expiration times not being
equal. And sure enough, gnunet-namespace sets expiration to 0 when submitting a
template for seek-and-destroy. This patch: 1) Changes gnunet-namespace to
fail if the time is not absolute or unspecified. If it's absolute, it's used
as-is (hoping that the user got the exact expiration time somewhere - probably
from -D). If it's unspecified, it is set to zero (as it was before). 2)
Comparison ignores expiration time if either of the arguments has expiration
time set to zero (i'm assuming that real records will never have expiration
time of zero. If it's absolute, zero is in the past. If it's relative, zero
makes no sense). Also, i'm not sure it makes sense to have two records that
have different expiration times, but are otherwise identical (i.e. both are
public or private, same data, size etc). Lookups will find (and will be sat
isfied) by either, so there's no sense in keeping both (unless you're doing
some wizardry with expirations, but i'm not sure it's worth the compexity).
-LRN: give gns a port
-cpp fix
-cleanup
-LRN: fix new mesh to compile
-LRN: W32 NS provider for GNS
-LRN: adding missing file
-LRN: another one
-fix crash if private zone key did not exist
-LRN: Our std handles are not created by us, and thus are inheritable by
default. We don't want them to be passed to our child processes, because our
parent might be waiting for them to be closed, which would mean that our
process is dead.
-LRN: Python-based test script for statistics
-reverting SVN 22558 (new insights...)
-ignores
-LRN: Another take on std descriptor inheritance
-LRN: switching to py for arm test as well
-fix
-never start child processes on UNIX without stdin/stdout/stderr, pass
/dev/null instead if we want to decouple the child
-minor fixes
-misc edits working on #2494, but there are still API issues to be
discussed, tests do not work yet
-simplifying function checks
-simplifying function checks
-fix
-breaking testbed again...
-dos2unix
-breaking proxy as well
-test for execinfo.h instead of LINUX
-__u32 is likely not understood if we do not have in6_ifreq already; try
using uint32_t instead
-Bertrand Marc: fixing typos found by lintan
-code cleanup
-pass stderr to helpers:
-adding gnunet-config tool
-new API for asyncronous generation of private RSA keys
-code cleanup
-only test for exist, not for read
-actually make use of SGID permissions
-test for getresgid
-moving block test to resolve cyclic dependency
-cleanup
-cleanup
-adding support for heap stats gaugering
-adding mallinfo test
-fixing warnings from gauger header
-fixbuild
-try using gauger for memory measurements
-do not use signal pipe as gnunet-rsa is likely to be in a gcrypt select
loop
-clarifying inputs
-clarify name rules
-minor code cleanup
-possible fix for assertion in peerinfo
-trying to fix high CPU load caused by DHT
-trying to fix perm problem despite SUID
-better fix for perm issue
-skeleton dns2gns
ns2gns
-cleanup
-more work on dns2gns
-fix
-doxygen
-do not install w32 helper service conf on GNU
-debug code for new transport issue
-fix
-check return value
-preserve errno
-nocaps
-check return value
-check return values
-ignorerv
-dox
-fix
-fixfix
-fixes
fix for gcc alginment issue on sparc reported to Debian as #670578
-ignore
-fixing installation scripts for suid binaries
-make install scripts executable
better style
refactoring dns service to take stub code into separate library for use
in gns2dns proxy
updating report.sh, removing obsolete hostlist scripts
extend API to enalbe exclusive port ranges to be specified for
testing-system objects
only use control pipe with helpers IF the helper actually supports it
fixing #2530
-increase max nodes
adding DHT option to disable calls to try connect
commenting code
typo
always allow root
-bugfixes
better log message, fixing stale task
-fixing potential NULL deref
mark fcntl return value as intentionally unchecked
check return value from sendto
updating doxygen options
-ignore
fix use uninit
clarify - kernel
-adding continuation to testbed peer start/stop API
make fs tests build against new testbed library -- they do not pass yet
migrating topology test to new testbed API
doxygen fix
-update configure info
-update configure info
-fix
-update configure info
new getopt helper function to parse relative time command line argument
commenting out dead test
-fixes
-API comments
adding another testbed API function: convenience function to get all
stats from all peers
nse profiler _compiles_ against new testbed API
fixing Debian #686238
eliminating use of 'namespace' as a fieldname / variable name to be C++
compatible
triple star fun
triple star fun done
code cleanup
-doxygen
-ignore
fix
-fix
-fix
-fix
-fixes
-fix
-fix
-fix
-fix
-fix
fixing #2544
making nse multipeer test use testbed
-note duplicate API
-fixing #2546
fixing #1551/#2503
-ignore
-fix warning
-adding GPLv3 license, adding uninstall script to fix #2534
fixing udp busyloop reported by LRN on #gnunet on 8-27
fixing Debian #684317
-fix warning
-trying to fix bb errors
-use blocking wait as the pipe might be faster than the process cleanup
-use blocking wait as the pipe might be faster than the process cleanup
-use blocking wait as the pipe might be faster than the process cleanup
-allow test to pass
-fix compiler warning
-also exclude perf
-hope springs eternal
-peer review
-peer review
-code review, breaking build a bit
-typo
-code review
-core review
tot tot
-fix segv
-code review
give sysctl /dev/null instead of no stdin/stdout
turn gnunet-rsa into key generation tool
-bad whitespace
-fixing Debian #688486
-reduce workbits
releaxing libgcrypt version check, fixing Debian #684997
-trying to fix unindex trouble on keyword extraction
-trying to fix unindex trouble on keyword extraction
-do not start vpn when testing dht
-fixes
-no line
-do not restart helper if exp_cb is set
-fixing testcase shutdown
-fix leak
-fixing leaks, warnings, adding missing 0-termination checks for IPC
messages
-undo
-fix double free, linker issue
-state to ignore
-fix leak
-fix leak
-check return value
-check return value
-fix
-add cancel API
-use new cancel API, fix use after free
-fix
-fix tld guessing code
-some fixes for monkey
reducing error messages about missing configuration options by
introducing new helper functions to print them uniformly
-converting more places to use STRINGS_relative_time_to_string
-converting more places to use STRINGS_relative_time_to_string
improving PO set
-reorder
-dos2unix
-dos2unix
-dos2unix
-licenses, dce, etc
-DCE: _LIBC never defined
-no vpn
-simplify code structure for monkey
-docu
-do not do this for C code, only for C headers
-use sane data types
-simplify
-simplify
-syntax fixes
-avoid dirsep for seaspider
-use FIXME
-missing endif
-fix use of const
-fix
-nicer error message
-do not close stdin/stdout for standard system progs
-I think these should be .gads, not .gnunet at this point
-fix includes
-fix
-no arg
-fix cmp
-dce, stylistic fixes
-fixing #2569
-dce
-fixing style issue
-fixing regex linkage
-undo Martin undoing my 0-termination checks
-remove const const
-doxygen fix, fix off-by-one
-fix
-trying to fix bratao's 100% CPU bug, even though I wonder if this can
really be it, as there was a 1s delay already; anyway, exponential back-off is
better
-highly experimental patch to fix #2578, could theoretically cause
unexpected problems in unrelated places
-linker fix
-dead
-ignore
-use pclose instead of fclose as we used popen
-fix #ifdef HAVE_RESUID
-setpriority is priviledged, so save on those useless syscalls
-better output for 0 delay
-fixing #2578
-fixing qpe counter -- #2555
-reducing memory use for PendingRequestData by 30% by not allocating
space for target/namespace hash codes unless the request needs/has them
-chaning multihashmap API to allow option for avoiding key allocation
-add logic to measure heap size for all processes that use statistics,
reduce statistics, topology, nse and resolver heap usage using the same trick
we used for arm
-minor cleanup
-cleaning up arm test
-abort pending transmission before disconnect
-abort pending transmission before disconnect
-trying to fix #2572
-fix
-only run integration tests if MHD is available, as they need MHD
-fix
-getting rid of silly, stupid, useless, often wrong DEFAULTCONFIG setting
-putting comment back
-use defaultconfig if config is missing
-do not complain if ~/.gnunet/gnunet.conf does not exist, just use
defaults
-updating docu
-fix
-install sigpipe handler
-note authors
-update docu
-only generate keys of size 914
-only generate keys of size 914
-only build hostlist if we have curl
-fix for #688590: allow user to specify how to install nsslibs
-patch for ATS to fix compiler warning/error from Markus Teich
-use DLL, nicer use of indexing map, improve ability to use future hash
map memory optimization
-enable peer API to return pointer to interned peer hash code
-update ignores
-optimizing all FS multi hashmaps for key memory sharing
-enabling new optimized multihashmap
-cleaning up RSA code
-cleaning up RSA code
-fixing plan_map over-eager interning as reported by mwachs
-more rsa refactoring
-ecc draft code
-dce
-ecc man page
-code deduplication in rsa sign/verify code
-fixing obvious ecc issues, adding gnunet-ecc based on gnunet-rsa
-adding ecc test -- still failing
-adding libidn--towards fixing #2475
-ecc sign/verify only
-done
-fix the fix
-dce
-dce
-fix
-code cleanup, handle -1 timeout from curl better
-switch on int for cparser
-misc bugfixes
-more logging
-implementing #2475
-clarify API
-reduce stats collection by default
-doxhappy
-dead cast elimination
-namestore tests require sqlite
-trying to fix fragmentation / udp performance, not working yet
-fixing frag
-hopefully fixing #2596
-stop tracking requests from local clients forever
-make datastore queue longer, and length configurable, reduce statistics
(#2594)
-reduce statistics (#2594)
-do even fewer stats by default
-lots
-introducing MDLL macros
-using new MDLL facility to clean up bijection data structure and reduce
memory consumption in fs service
-add option to make expensive heap stat gathering optional
-fix #2598
-speeding up baadf00d code by 7x (on i7)
-fix double free
-fixes
-strange branch, strange code
-leak
-leak
-fix logic
-simplify
-doxygen
-check for sudo, nslookup
-fi
-fix
-fixing testcase cleanup on error
-typo
-rasp pi redefines slow
-check that which succeeded
-bugfixes, code cleanup
-fix
-indent
-test case cleanup
-code cleanup
-minor cleanup, fixing minor bug in auto-share
-wk says not to specify pkcs1
-ecc signing working
-man page for configuration file
-adding some missing man pages
-more man pages, typos'
-enable 'hello' URIs for gnunet-uri
installing all service, daemon and helper binaries to
lib/gnunet/libexec/; updating code to run binaries from new location, which is
no longer in PATH
-also to libexec
-fixes
-fix build errors
-clean up test
-pogen
-some gns code cleanup, still a mess
-fixing gns test issues
-doxygen
-more cleanup and input validation fixes
-fix path problem with libexec
-fix
-adding missing extra_dist files
-adding missing pkgconfig files
-moving DHT test to testbed API
-need 1.5.0 for ecc
-trying to fix #2603
-typo
-better logging in NAT
-more libexec fixes
-doxygen
-more libexec fixes
-minor fixes
-use loopback
-use loopback
-major DHT test cleanup
-more DHT test modernization
-misc minor fixes
-misc bugfixes, travel hacking
-add adv port
-fix #2609
-leak
-fix check
-fix missing check of return value
-fix leak
-fix use of PREFIX with libexec
-version bumping
-fix fix '
-fix libexec issue: SUID check
-fix
-actually enable free regex specification, in binary for now
-fixing #2617 --- removing dead autoconnect option
-allow config-less services to start
-allow config-less services to run
-link against gnunetutil
-ign
-fix no-dot-in-path issue on FreeBSD bot
-make gns independent of vpn on W32
-do not use libexec for absolute paths
-not experimental anymore
-put fcfsd into libexec
-yet another libexec fix
-fix uninitialized var
-fix fix
-tolerate NULL
-fix block download calculation
-ignore
-bump
-fixing
-fixes
ready for 0.9.4 release
-fix warning
-keep as const
-typo
-mteich
-starting build system for consensus
-updated and moved to https://gnunet.org/mysql-setup
-updated and moved to https://gnunet.org/postgres-setup
-rm
add option to disable use of stats entirely
-fix message recv accounting
-new breaks
trying to fix #2645 -- support lib64/gnunet/libexec/ installations
-more realistic default quota values
-implementing #2435
use new DHT API to block known results when searching in the DHT
"man gnunet-peerinfo" doesn't have the -g or -p options which are listed
in "gnunet-peerinfo --help". One spelling mistake is also fixed in the attached
"diff" (see: additional information). This was against r24891. - #2649.
-new datacache option
support running purely in-memory
-not a warning
-do not expire too soon
-adding plugin to keep datacache entirely in the heap
add option to disable Bloom filter entirely (only for benchmarking)
-wtf
-helper is not called from cmd-line, so no utf8 conversion required
fixing bloomfilter reconstruction after quota change, iterator for sqlite
failed to return keys due to off-by-one error
-actually, on W32 we do need to convert...
fixing #2641, datastore size way off
changing datacache API to separate put-paths from data (so that plugins
can test for duplicates), removing support for MySQL
-memory is cheap, default to heap
-fix #2652
-clearer logging
-allow override of NSE result for benchmarking
-delay between peer start and regex PUTs
use exp backoff macro, start with retry frequency of 1s, instead of the
insane 1ms
-extra logging, minor code cleanup
-no more monkey business in defaults for now
-doxygen
-removing useless casts
fixing crash of mesh service -- reproduced by regex profiler w 100 peers
-- on shutdown due to failure to initialize dht_handle field
-back to sane logging
-mysql datacache is gone
-fix the fix
-fix
-make use of new IRRELEVANT status code
polishing consensus API
-polishing consensus API
-more libexec fixes for OpenSUSE
-typo
-receiver hack
-assert privs when needed
-assert privs when needed
-use execlp
-fix
-also allow ad-hoc
-also allow ad-hoc
-also allow ad-hoc
-cleaner error reporting
-trying to add 802.3-encoding support
-trying to add 802.3-encoding support
-fix
-fix
-fix
-fix
-fix
-fix
-fix
-tolerate all-zero addresses
-dox
-fixing #2656
-pogen
starting to use stream in fs
mostly finishing server-side for FS-over-stream
towards using stream for non-anonymous file-sharing
-ensure that either stream_api calls callbacks last or that we don't
destroy a stream handle while it is in use below us on the stack
adding missing parens around macro arguments in gauger macros
combining test_fs_download.c and test_fs_download_index.c into one file,
adding additional test test_fs_download_stream
combining test_fs_download.c and test_fs_download_index.c into one file,
adding additional test test_fs_download_stream
-add multi-peer testcase for stream-based transfers (expected to fail
right now), and option to disable anonymous transfers --- to force stream-based
transfers
-config fixes
-config fixes
actually using stream for dblock/iblock transfer if possible
-updating dependency diagram
removing invalid assertions that shutdown must preceed closing
-enabling stream test fully
-disablefornowagain
-more logging
-fix double-sending in stream if finish_cb behaves in a certain way
-enable mesh, as stream needs it, activate testcase
-switch to hash map for replies to avoid linear scan, add timeout for
inactive clients
-clarify API (#2661)
-dox
-implementing limit to number of stream client connections
-add retries, and retry limit (#2659)
-docu
-you are welcome
-always use 'ret'
-make freebsd7 happy
-send content type
-introduce #define for + in GADS
-ensure labels are less than 64 chars, add test for full DNS names
-small fixes:
-typo
-fix options for gnunet-gns
-fix log message
-typo
-newline
-fix
-fix conversion
-fix align
-more logging
-fix availability rank calculation in probe signal, immediately signal
failed probes
-signal to applications that a probe is active
-sane freq
-feedback on style
-improving shutdown behavior
-skeleton
-write unique peer ID to file for each peer
-skeleton for the fsprofiler daemon
-oops
-more work on FS profiler
-more work on FS profiler
-function never existed
-docu fix
-implementing heap-based datastore backend
-avoid mysql warning (was harmless, but still not nice)
-options to play with
-virtually finished fs profiler
fixing #2680 -- config file lines can now have any length
look in
/opt/jdk1.6.0_22/bin:/usr/lib/jvm/java-6-sun//bin:.:/home/grothoff/bin:/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games
for gnunet-arm before trying /proc/exe
adding GNUNET_DISK_file_backup function; fixing #2646
-fix
-fix
-fixes
do not queue more than one write at a time to stream, -- hopefully fixing
#2672
-fix test
-fixes
-fix
-logging, initialize msize
-fix
-extra check
-auto reset non-progressing streams after 30s -- hides #2672
-fix
-fix
-fix
-fix
-fix
-fix crash with -D
-starting to move nat autotest code over
adding function to compute configuration differences in memory
nat auto code from gnunet-gtk now compiles as part of libgnunetnat
-dos2unix
-comments
-malloc zeros out
-pass NULL instead of 0
-use unsigned
-stuff
-reduxing regex dfa_merge_nondistinguishable_states memory consumption by
32x
-reduxing regex dfa_merge_nondistinguishable_states memory consumption by
32x
-reduxing regex dfa_merge_nondistinguishable_states memory consumption by
32x
-eliminating mallocing of state sets
reduce reallocing to improve performance
-reducing CPU usage from nfa_closure_set_create by avoiding
double-sorting and quadratic scan
-reducing CPU usage from nfa_closure_set_create by avoiding
double-sorting and quadratic scan
-fix
-fixfix
-nodebug
-fix compile issue
-improve swapping behavior
-allow NULL if n=0
-minor fixes
-going crazy on optimizing regex, fixing misc minor bugs
-fixing some leaks
-bugfixes
LRN: fixing #2701
-fixing heap walker
-vfork is back
-eliminate consensus warnings
-do not use consensus
-do not print warning if file does not exist yet
-add newline
-respect vfork semantics
-generate nicer error message
-fix option name
-new topo format
-updating ignores
-order
-fixes
-fix
-preserve command-line keywords even if extractor is disabled
-fix #2708 - pogen.sh
-must not log in signal handlers
-fix index out of bounds
-check return values
-cleaner
-unify workbits to sane value for test
-clenaup
-no alloca
-API discussion
-reminder
-quick and dirty use-after-free fix in LM
-nicer
-debug
-check for helpers in libexec
-die old testing, die
-renaming testing-new.h to testing.h, bumping library versions
-fix
-stuff
releasing GNUnet 0.9.5
-fix
-fixme
adding NO_IO option to peerinfo to disable disc accesses for experiments
fix
minor style stuff
-we are not ms
-no need to be nice
-note for API discussion with Sree
-python script to compute GNUnet CHK URIs
3
-try force binding client socket to loopback/localhost for localhost
connections to facilitate access control checks
-undo 25844
-tiny fixmes
-parse time nicely
-fix #2743
-check that this is actually a DNS packet
-style
-allow disconnecting from DHT during GET iterator
remove DNS exit for mesh from gnunet-service-dns
add DNS exit for mesh to gnunet-daemon-exit
start gnunet-service-arm nicely if configuration file does not exist
-fix
-fix
-arminteg
-def
-allow custom listen port
LRN: Here's a patch. See if it doesn't break anything for you.
-force no c format
-fix leak
-check for EINPROGRESS for UNIX, hopefully fixing #2741
API extension, make blocking socket API visible
-argh
-sanity
porting to W32 requires a port, yepee
-stuff
-clean up published files - #2725
-cleanup
-fix non-terminating index test --- do not remove file early
-fix warning about modifying const cfg
-zero out cdata here as well
-export ecc generation function
-trying to fix low parallelism bias
-pre-generated ECC keys for testing
-old testing is dead
-ecc key
-ECDH test
-dist
-do not fail hard on log skip checks
-do not assert on log
-clarifying what we need for ECC
095a
switching to ECDHE cryptography f, implementation is incomplete and
INSECURE, do not use for anything but testing
-make dv compile again
-restrict curl accesses to http, disable redirect for http plugins
-trying to address #2791
-fix state machine rekey logic --- #2790
-ignore
-key gen perf
-protocol change
-updated translation
-build fix
-trying to fix #2794
-check for NULL
-proposed mesh2 API draft
-use 2048
-proposed multicast API draft
-fixes to make stream tests pass with ecc in core
-big endian
-updated Spanish transltion
-note
-actually run the test
-disable extra checks
-disable extra checks by default
adding option to disable use of ephemeral keys for HPC testing
do not verify an ephemeral key again if it is the same as the previous key
reduce messages in initial KX
-fix
-remove 093 compatiability code, change validation order slightly
fix #2811 - missing options in gnunet-peerinfo man page
-fix
-include time offsets in logging as well
-nicer time format
-trying to fix rekeying (#2812)
-trying to fix rekeying (#2812)
-if we do not use ephemeral keys, use validity of 'forever'
-go back to 12h interval
-relax UID check
-dead
-dead
add argument to GNUNET_CRYPTO_ecc_decode_key to allow testing to disable
key validation
-go to more sane ECC curve to be more DoS-resistant
-go to more sane ECC curve to be more DoS-resistant
-transport must be able to run without stats
-transport must be able to run without stats
-removing spurious log message
-76 is right, but you have to delete your ~/.gnunet/private.ecc
-fix #2815
-rsa, not ecc
-improve strings API
starting major change towards implementing #2564, this breaks some FS
tests and FS APIs
-need 1.5.0
-fix
-fixes
-fix
-fix
-expand pseudonym test to cover crypto
-fix
-towards pseudonym crypto
-needs curl
-make DHt routing more sane
-no reason for continue, AFAIK
-I think this should be installed
-not a public API
-hacking on dv
-removing useless parts of old code, work on DHT API
-removing useless parts of old code, work on DHT API
-fixme-mwachs
-finishing DV API
-more dv bookkeeping work
-simplify design
-more DV hacking
-fix'
-more DV hacking
-const
-towards actually using consensus
-more DV hacking
-doc
-generate and process ACKs
-docu
-nack support
-adding distance change notification from service to plugin
-removing ancient code, comments
-dv work
-draft for DV plugin
-box DV messages when needed
-fix
-enable DV status monitoring
-count direct neighbors in stats
-fix
-always call CB with correct closure
-fix uninitialized sendto
-towards a DV test
-doxygen
-typo
-fixing #2563
-require 1.6.0 -- beta93 is enough
-doc
-do not call get callback after get_cancel
-make really sure 'cont' is never called again
-hopefully no longer needed...
-updating pseudonym crypto implementation
-enable evil
-notes
-idea for improved allocator
-notes
-never assert on bogus input, return error code
-cleanup
-install tap driver on W32
-include source in make dist
-fixing string decoder issue for input sizes of a multiple of 5
-cleanup
-simplify
-simplify
-towards implementing #2685
-doxygen
-fix type
-doxygen fixes
-towards using ATS for latency info
-update dependency map
-const
-use ATS from DV
-update structure
-doxygen
-fixing #2837
-code cleanup
-updating config file towards forcing use of DV
-typo
-test cleanup
-dead
-adding blacklist test to check that as well
-fix
-fix
-add timeout
-timeout
-fix shutdown order
-fix fix
-ignores
-refine test
-no longer pass distance information from non-DV plugins
-more logging'
-more distance removing
-typo
-cleanup
-indent
-this CAN happen, do not crash then
-need lockmanager
-ignore
-new core API doesn't expose ATS anymore, likely mesh2 should not either
-API cleanup
use SCRYPT - fixing # 2685, needs LATEST libgcrypt (Git from today)
-fix
-update
-first thoughs on a PSYC API, not finished
-doxygen
-doxygen
-consensus fix
-towards proper pseudonym support
-updating author list
-fix
-move pseudonym code to fs, mark fs as experimental for now
-fix #2857
renaming GNUNET_PSEUDONYM_-API to GNUNET_FS_pseudonym-API
-remove unnecessary arguments from ARM callbacks
-doxygen
-ignores
gitlibgcrypt
-no boolean
-no boolean
-move data resource
-use GCRYMPI_CONST_ONE
-gcrypt works better now, removing some debug logic
-fixes
-adding florian's proposed SET API
-minor remarks on Florians draft
-comments
-fix
-fs stable again
set close-on-exec file for opened disk file descriptors
-dead option elimination
-likely also dead
-evolve PSYC ideas
-fix glpk test (#2870)
-make sure result is 0-terminated for UDS
-comments
-comments
-hopefully fixing #2869
do not enumerate all IP addresses from our interfaces if the plugin is
bound to a specific IP address only
-do not call NAT with zero addresses
-comments for Florian
-remarks
improving example documentation
-fix link order
-link-to-doc
-fix error reporting
-fix error reporting
-update ignores
-fix crash if UDP port is in use
-stats is allowed to be NULL
-no peerinfo IO on tests
-formatting
-fully abort and restart existing iteration on notify
-indent
-no peerinfo IO on tests
-some transport tests require persistent peerinfo
-fixing pseudonym signing bug again -- how did the old fix get undone???
-quiet means quiet
-be less verbose about normal stuff
-fixes to error reporting
-hide unnecessary output, add notification that test was actually run
-handle case that we are in restart on stop
keep DNS helper running if IPv6 sockets fail, then we do IPv4-only
-test for mangle support
-test for mangle support
-indentation
-fixing indentation, logging
-be less noisy
-minor code cleanup
-indentation
fixing ARM restarting, adding monitor mode to command line
-document IPs involved in test
-fix CNAME handling after GADS to DNS delegation, need to repack query
packet
-cleanup
-update
-only run PT tests if we have libcurl
-fix W32 warning
-fix
-fix
-need to unset control pipe on OSX, check for bad FD in environment
variable
-simplify
-simplify
-simplify
-simplify
-fix
-nicer
-fix
-cleanup and some remarks
-towards using SET api in DV
-towards desired set API
-switching GNS from RSA to ECC
-removing chat from build
-preparations for fixing #2711
-towards allowing probes for directory entries
-fix
-fix
-export FS handle during events, as we might generate events duing
FS_start and client then doesn't know the handle yet
adding tool to convert/merge datastores (#2875)
-regex service API
-hacking up new regex client library
-skeleton for regex service
-ignore
-towards implementing regex service
-finishing regex service, but untested
-fix
-fix
-there is no test_mq here
-use set conclude API
-notes
-fix
-remarks on mesh api
-more comments on mesh
-updating desired project structure graph
-duplication
-update
-towards using mesh instead of stream
-file renaming stream->mesh
-rename stream->mesh in comments and symbols
-ignore
-moving to regex
-cleaning up client api to test for service availability
-cleaning
-more work on multicast/psyc API designs
-get rid of channel hierarchy for PSYC level, will be kept in user
interface
-further refinements to the APIs for PSYC/multicast
-allow asynchronous answers to membership test
-remove unnecessary client pointer from struct and avoid the keep/drop
calls that are not needed
-keep and drop are not needed here as we watch for disconnects
-keep and drop are not needed here as we watch for disconnects
-keep and drop are not needed here as we watch for disconnects
-keep and drop are not needed here as we watch for disconnects
-keep and drop are not needed here as we watch for disconnects, also
implicitly fixes missing drop
-use doxygen-style comments
-fix copyright year
first design for social service API
better name for entering a home
improving social API a bit more
-clarify need to decide on admission
first design for psycstore API
-make use of ability to reply on reject, i.e. to redirect to private room
-fixing issue with testcases and missing sqlite3 reported to bug-gnunet
-add option for future use
-remove dead option from defaults
-clarify which connection is closed
-spacing
continuing refactoring of regex library structure, disambiguating symbol
names between old regex library which is now internal and the public API;
moving IP-to-regex conversion functions to tun library, as only vpn is using
those to begin with
moving functions for testing and evaluation and experiments to the test
library, minimizing the internal library, renaming files according to which
library they belong to
adding file that was moved here from regex.c and forgotten in earlier
commit
renaming symbols from libgnunetregextest to have the prefix REGEX_TEST
renaming libgnunetregexnew to libgnunetregex
-fixing link order
fix linker issues on W32 -- regex needs DHT
fix compile error left from renaming fest
-fix double free
-fix double close/free
-fix use after free in log
cleaner to not use ptr after freeing it, even for printing
cleaner to not use ptr after freeing it, even for printing
-fix leaks
-fix memory leak (on error path)
-check for NULL (in testcase)
-stress that we do not care about return value here
-do check return value from remove
-check return value and report errors
-avoid side-effect in assertion
-avoid side-effect in assertion
-avoid side-effect in assertion
-avoid side-effect in assertion
-avoid side-effect in assertion
-sorry amatus, this breaks the build on other systems and libltdl should
simply not be installed as part of GNUnet, that's a bad idea
-fail hard if we could not create the directory for the HELLOs
-cleanup leaky test code
-trying to fix #2902, always build dns code and fix dependencies
fix #2901: fail hard if libidn or header are not found
-update svn ignore files
-proposing identity service API
-cleaning up block code, moving all of the block logic into the block
library
-renaming symbols of the block library to use the REGEX_BLOCK_ prefix and
not the REGEX_INTERNAL_ prefix
-duping comments from declarations to definitions
-inline trivial static function that is only called once
-follow naming and coding style conventions
-moving REGEX_INTERNAL_check_proof to libgnunetregexblock, integrating
proof check into block plugin
-renaming symbols of the block library to use the REGEX_BLOCK_ prefix and
not the REGEX_INTERNAL_ prefix
-combine log statements
-removing unnecessary key from regex blocks, compacting some of the other
fields
-introducing convenience function to load private key of peer
-introducing signing of regex accept states, removing PID from regex API
-move struct RegexBlock into regex_block_lib
-fixing #2585: optimized layout for regex blocks
-fix
-clarifying tutorial
-fix comment
-fix compiler warning, removing unnecessary attribute
-towards implementing improved namestore API
-fixing some namestore tests
-fixing some namestore tests
-fixing some namestore tests
-remove old files
-adding ECC-based keys
-updating PO files and man page
-fix build issue, function was renamed
-use valid ECC keys for testing
-fix build issue, function was renamed
-fix key name
-implementing regex test
-removing useless code
-removing dead options
-update ignores
-doxygen fixes
-doxygen
-skeleton for monitor API
-get rid of iteration response, use lookup response instead, message
formats were identical to begin with
-finishing monitor API on client-side
-typo
add '-m' option to gnunet-namestore, extend namestore monitor API with
notification about being in sync
-simplify zone key loading by using synchronous ECC key API
-simplify zone key loading by using synchronous ECC key API
-simplify zone key loading by using synchronous ECC key API
-resolve of 0 is well-defined, do NOT assert here, breaks tests as well
-remove async ecc key generation, not needed
-dead code
-towards implementing monitor functions in service
-towards having monitoring
finish implementation of monitoring, send change notifications to all
clients
-testcase for monitoring, fails right now, as a good test should...
-fixes in monitoring code
-update docu
rename 'leave' to 'part' as 'part' is the better opposite of 'join'
-underscores for lynx
-1:1 unicast --- not sure it belongs in social API
-more ideas on naming
-check for NULL
-add missing option
-const where const can
-no void
-extra checks
-document
-slight simplifications
-typo
-add extra checks
-fix
-fix
-no regex autostart
-no namestore autostarting
-disable regex autostart for tests
-adding missing file"
-adding API to associated user context with a server's client to avoid
ugly hacks, such as linked list searches and casting pointers to integers
(thanks to Peter-Vogginger for forcing me to fix this)
-fix doxygen
-return NULL if never set
-add dv_get_network to DV API, improve signature to use 'struct Session
*' instead of 'void *'
-bring set/gns into the fold of non-experimental subsystems
-move dv API header to where it would typically belong
-fix loglevel
-update test cases and improve output
-documenting and cleaning up gnunet-publish code
-fix year
-adding missing functions to identity service to make it more complete
wrt replacing namestore functions from FS
-updating ignores
-typo
-skeleton for identity service
-fix #2940 - FTBFS
-identity IPC definitions
-identity IPC definitions: message types
-license version fix
-create makefiles for identity service
-update ignores
-improved debug logic
-bad style, die
-renaming pseudonym to ego in identity service
-cleaner
-first steps towards identity client API
-addressing variable-size ECC private key len in IPC
-break in unimplemented functions
-doxygen
-hacking get, rename, delete on client side
-finishing identity client library, at least in theory (untested)
-suspend/resume is unnecessary now
-skeleton for identity service
-code to generate some of the responses
-towards handling rename/delete requests
-towards handling identifier creation
-notes on what is left
-handle get default
-fix: set default includes service name, not identity name
-handle get failure properly
-implement set handling
-update subsystem cfg on rename/delete
-handle loading, deletion and renaming of files of egos
-theoretically finished identity service (untested)
initial version of gnunet-identity command-line tool
-proper support for initial listing vs. monitor mode
-towards testing identity service
-add identity service to standard build
-doxygen
-add missing includes for FreeBSD
-silliness
-fix uninitialized fields
-fix typos
-identity API bugfixes
-use separate logging cateogory for util-server-nc
-improve test
-skeleton for 2nd test
-misc bugfixes and test-fixes
-code cleanup, tolerate e->vals of NULL
-modernize a bit more
-removing obsolte stream message types
-renaming stream->mesh where applicable
-minor code cleanup
-splitting mesh into server and client parts
-minor bugfixes
-missing file, more cleanup
-nicer logging
-sync for bart
-fixing bad replacement earlier
-fix shutdown logic
-enable buffering
-use sh instead of bash
-fix function signature
-add dependencies to hopefully make W32 bb happy
-removing last bits of RSA support, as this code is now dead
-updating ignores
-minor code cleanup
-pogen
-doxygen fixes
-fix uninit and fix leak
-initialize element type
-do NULL check before dereferencing
-move all code that deals with IP/policy to regex/string conversion to tun
rename GNUNET_TUN_ipvXtoregex to GNUNET_TUN_ipvXtoregexsearch to clarify
that this is for generating the search strings
-note where 2919 should be also implemented
-handle set API quirk -- #2955
-w32 bugfix -- compiler warning too
-fix warnings on W32
-do NOT cancel here
-fix #2970
-make use of deterministic ECDSA in FS, requires libgcrypt from Git as of
yesterday
moving to new, fixed-size encoding of public and private ECC keys
everywhere, also improving ECC API to better support ECRS/GADS operations
-updating hostkeys to new format
-removing RSA hostkeys
-again HELLO format change, so these are now bogus
-disable identity
-fixing main FS build, updating man page of gnunet-pseudonym
-fix, handle case where there is no update
-implement block decryption again
-rename as format changed
-fix W32 FTBFS
-fixing #2976
-testcase for new ECC crypto functions
-fix year
-doxygen fixes
-fixing comments, indentation
-raise loglevel
-fix loglevel
-fix
-minor fixes
-update testcase
-working on identity service create/rename/delete
-add extra context argument for key deriviation, so that fs and gads do
not collide
-fix
changing time measurement from milliseconds to microseconds
-be silent
-round expiration times to full seconds
-use 5 ms for timeout, not 5 us
-use ms for timeout, not us
-big bad bugfix thanks to Bart
-indentation
-some minor code cleanup
-more code migration to microseconds:
-fix #2969
-fix leaks (kind-of related to #2981)
-old fixme
-towards namestore support for the new privacy-preserving GNS queries
-more work on new namestore API
-fixing #2991
-doxygen fixes
-namestore service builds again
-build namestore
-adapt PT test to new namestore API
-working on new, simplified GNS API
-getting rid of duplication of record type values / enum
GNUNET_GNS_RecordType
-simplifying GNS API
-more work on simplified GNS API implementation
-removing test for dead functionality
-more coding towards working new GNS service
-documenting open issues
-more work on new GNS implementation
-dead code elimination
-adapting fcfsd to new namestore API, moving from gns to namestore
subsystem
-make build conditional on MHD
-typo
-adapting gnunet-dns2gns to new API
-gns compiles again
-mark one location of #2985
-fix
-fix build order
-dead code elimination, and docu
-dead code elimination
-dead code elimination
-dead code elimination
-dead code elimination
-minor fix
-dce
-more work on clean main GNS logic
-old code elimination
-implementing #2996
-fixing #2999
-undo stupid gettextize overwrite
-addressing #2995
-doxygen
-playing with doxygen groups
-fixing #2993
-if records are deleted, delete them from cache as well
-eliminating duplicate struct definitions
-handle client disconnects in GNS
-fix #2994
-enable resetting client context to NULL
-clean client context on shutdown
-be more permissive with label syntax
-gads to gnu
-updating ignores
-fixing #2985
-more work on GNS resolver
-proxy does not build yet
-implementing handling of VPN records in new GNS
-implementing #2990: integration of gnunet-namestore with identity service
-eliminate more uses of ShortHashCode, breaks signatures
-fix time test
-remove more uses of short hash
-dead code elimination
-remove ShortHashCode from API, deprecate SList
-dead header
-dead header
-handle CNAMES in GNS
-misc doxygen fixes
removing deprecated, dead test-connected API
simplified OR2 API
-simplified OR2 API
implement conversion to absolute TTL
document places where SHADOW records likely need to be handled
-towards .+ expansion, planning and documenting how and where
removing GNUNET_CRYPTO_ecc_key_free, use GNUNET_free directly instead
-doxygen improvements
-pogen
removing deprecated argument in 'init' callback of GNUNET_CORE_connect
-relative path
-also on W32
-doxygen fixes
-doxygen fixes
-doxygen fixes
-make clang happy
-dead code elimination
-code cleanup
-move shorten logic to its own file
-this header is not used here
-expanding namestore API with pkey <-> zkey conversion functions
-implementing zkey <-> pkey conversion functions
-converting basic DNS records to GNS
-start to integrate identity service with gnunet-gns command-line tool
-updating dependency graph
-starting to rename vectorproduct to scalarproduct, as this is not doing
a cross product
-more vector -> scalar renaming
-more vector -> scalar renaming
-more renaming, and ignores
-more renaming
-ignore
-again trying to delete old vectorproduct dir
-use GPLv3+ consistently
-use GPLv3+ consistently
-exposing dnsparser record packing API
-convert additional record types
-convert additional record types
-expose API for parsing individual DNS record types
-towards translating .+ names in GNS records
-expose API for freeing DNS record members individually
-this is dead, right matthias?
-implement .+ expansion in GNS record values
-updating gns status notes
-removing obsolete check
-more convenient API to lookup egos
-use identity service also for shorten zone lookup
-use new ego lookup function to simplify logic
-only run python tests if we have python
-report error if starting openssl failed
-needs openssl
-old comment
-gpl3+
-remove dead struct
-align core message, do not run psycstore here
-pogen
-fix u/clarify/improve util/network code doxygen documentation
-doxygen fix
-harness for multicast
-move messages out of API that don't belong into API
-integrating multicast with build system
-update ignores
-cleanup, FIXMEs
-docu
trying to fix #3006, libltdl detection on FreeBSD
-new key
-note
-only use x for ECDH
-improving gnunet-identity command line options and man pages
-fix date
-integrate identity service with dns2gns (#3010)
-update default zone initialization (#3009)
-allow passing NULL for zone to iterate over all of our zones
-doxygen
-implement iteration over all zones (#3016)
-activate shortening logic (#3012)
-link to bugreport
-fix clenaup
-use standard zones in gnunet-gns
-discussions on cleanup
-more precise check when to go to DHT (SVN 3017)
-update comments as well
-remove deprecated GNUNET_NAMESTORE_RF_AUTHORITY
-fix FTBFS
-doxygen, remove GNUNET_FS_uri_sks_to_string_fancy (dead)
removing gnunet-pseudonym, we're using GNS/gnunet-identity instead in the
future
-fix bad free, cut&paste error, might fix #3020
-comment out dead tests
-minor fixes
-misc fixes
-cleanup leaks
-code cleanup
-dox
-ignore
-require MHD 0.9.30
-generate tag file
-update ignores
-doxygen improvements
-moving time functions from FS to TIME
-check return value
-check correct variable
-minor code cleanup
-dead
-towards a list of symbols for monkey to filter
-eliminate LISTEN_PID check, silly idea anyway, better to just unset the
LISTEN_FDS after use; also this is needed to make things work with
gdb/valgrind/monkey/etc.
-monkey integration
-print main output to stdout
-fix crash in gnunet-namestore -a if no other records already existed
-fix missing '
-do not crash if cb was NULL on ego free
-finalize iterate_all_zones as well
-set zone on iteration over all zones as well
-go back to break
-remember that DHT put was complete
fix cleanup when request processing is done
-this is not a hard error
change .zkey encoding in anticipation of compact point encodings
-perform tiny wait if we do get worken up to do no work
-formatting, doxygen
-doxygen fixes
renaming fcfs to pin -- at least for now
-anonymous contribution: patch to add -c support to gnunet-gns-import.sh
-no longer experimental
-docu improvements
-fix assertion
-fix assertion
-cosmetics, trying to narrow down deserialization error
-do not clean up reader more than once
splitting 'struct GNUNET_CRYPTO_EccPublicKey' into one struct for signing
and another for encryption
-fix rename issues
-more rename issues
-fixing resuming publish operation
-fixing #3034
-dead test, gnunet-pseudonym is gone
-dead test, gnunet-pseudonym is gone
-dead test, gnunet-pseudonym is gone
-more complete fix for #3034
-updating man page, with contributions from Thomas Zander
-fix #3033
-avoid queuing large number of START requests
-properly clean up on disconnect, without memory leaks
-minor NSE cleanup
-stopid svn
-fix unindexing of KBlocks, make use of standard UBlock decryption
routines to avoid decoding failures
-fix check of return value
auto-correct missing 0-terminators in C_STRINGS/UTF-8 strings
auto-correct missing 0-terminators in C_STRINGS/UTF-8 strings
-doxygen fixes, and fixing publish URI serization for proper resume
-minor style fixes
-clarify API
-ignore generated psycstore.conf
-fix compiler warnings
-towards a compiling gns proxy
-towards a compiling gns proxy
-towards a compiling gns proxy
-ftbfs
-dead #define
add GNUNET_NETWORK_socket_free_memory_only_ API to allow getting raw FD
and releasing struct without leaking 'addr' member
-more cleanup, make it build at least
-update ignores
-eliminate ShortHashCode residuals (#2936)
-fix warnings
-more code cleanup
-rename structs to follow coding style guidelines
starting to clean up socks5 handling
-bump requried MHD version
-hacking more on proxy
-cleaner shutdown
more work on GNS proxy, adding back to standard build
require very recent MHD version
-adding logic for location rewrite
rework cookie rewrite logic
-handle failure to load certs more nicely
improve GNS zone / subsystem mappings
-fix segv
-fix bad assert
remove \r\n from values before giving them to MHD
misc bugfixes
-better error handling, GADS => GNS in CA
-cleaner shutdown
-fix double cleanup
-style
-logging, bugfixes
-avoid forever increasing backoffs on reconnect in normal operation
-handle case that we have no LEHO
-minor code cleanup
-fix shortening routine to deal with longer hashes
-handle NULL in unixpath
fixing issues and comments on multicast API
towards multicast IPC messages
-more issues with multicast API
changing DNS block plugin to represent DNS exit services instead of
hidden services
advertise DNS exit capability in DHT
clear private keys in static locations on exit
-misc doxygen fixes
-ftbfs
do DNS exit lookup in DHT
-fix naming convention
-doxygen and minor code cleanup
-update ignore
-doxygen, following naming conventions
remove 'nac' option from VPN, always return IP immediately, always build
mesh tunnel asynchronously afterwards
-dead code elimination in VPN
fixing #2916, but untested
-minor code cleanup
#2915: supporting multiple DNS exits and ranking among them
-doxygen
-doxygen
-cleaner
-doxygen, style
add option to cache blocks from result messages as well
-cleanup
-leave out proxy for now, as long as libcurl is not patched
-fix deprecated gnutls symbols
-update po.m4 to later version, update ltdl check
-uint32_t, not size_t
-run tests during 'make check'
-fix doxygen
-rollback accidental commit
-trap SIGINT for clean shutdown
-fix, is uint32
-simplify configs
-keep reading from client to support multiple GNS requests in parallel
handle empty result set gracefully
-formatting, doxygen
-removing bogus duplicate NTR cancel call
-more purpose.size 16/32-bit confusions
-fix PKEY delegation lookup
-fix testcase
-just whitespace removal
-not sure I see how this test can work, disabling for now
-set executable on sh
-cleaning up keys file
-report result of both tests
-adding missing GNUNET_PACKED
-doxygen stuff
-fix cname lookup
-indentation
-fix remove testcase
-set executable on sh
-get rid of warning
-fix MX testcase to use well-formed MX record string
-better error logging, return correct offset when parsing mx
-fix MX record parsing and serialization, others still to be fixed
-use dnsparser and consistently serialize DNS names in DNS binary format
-remove generated file
-activate tests, properly parse now properly serialized CNAMEs
-ignore
-ignore
-ignore
-misc fixes based on cppcheck
-check if timeout command exists before using it
-check return values, doxygen, code cleanup
-more logging, minor code cleanup
-fix memrchr
-fix uint16/uint32 confusion
-get rid of compiler warning
-get rid of compiler warning
-fix R-PI FTBFS
-minor cleanup
-undo const, it is not const
-testing LRNs fix from IRC to correct expression logic and move to 200
microsecond delays
-tolerate somewhat larger delays -- for slow systems
Applying 1st patch from Andrew Cann (see also #2887).
moving from abstract unix domain socket paths to normal unix domain
socket paths (#2887), removing now unnecessary/dead flags argument to
GNUNET_NETWORK_socket_bind
-ignores
-undo #29640, somehow causes problems
-just replacing if by switch, in preparation for future changes
using DNS2GNS record type instead of recycling NS record type in GNS;
fixing testcase by properly converting DNS records back to GNS blocks
-GADS is now called GNS
-fix interceptor's GNS to DNS conversion
Andrew Cann wrote:
-minor code/doxygen clean up
-fix shutdown segv if GNS connection was down
-commenting, removing dead message types
-adding missing PACKED directives
-simplify test
-typo/ftbfs
-doxygen, adding missing PACKED attributes
-disconnect on protocol errors, not on no-result
-add missing PACKED attribute
-fix block size calculation
-separate caching ops managment from request lifetimes, also separate
namestore handle, caused bugs
-shutdown if zone/ego not found
-disable NSS installation if configure did not set INSTALL_NSS
-eliminate dead option
-log if shortening is on
-delete old directory before running tests
-use correct subsystem name to locate shorten zone
-nicer messages
-minor stylistic cleanup, doxygen
-more logging
-removing test that is no longer relevant
-removing no longer used egos dir
-fix indentation
-more logging
-PSEU data is not 0-terminated, thus, we need to add 0-terminator before
using it
-fixing namestore to not disconnect on empty iteration result --- and
namestore API to better survive disconnect
-encrypt using both AES and TWOFISH, with independent symmetric keys
-rename files/symbols from _aes to _symmetric
-ignore
-enable looking up individual records
-fix expiration time, sqlite can only do 63-bit integers, as they are
always signed
-need sleep
-adding specialized hash table for peer identities
adding skeleton for revocation service
-revocation prototype
-draft for revocation service
-simplify logic, use modern APIs
-doxygen, todos, minor code cleanup
-fix
-towards addressing #3047, note this causes the code to FTBFS
-getting ATS and transport to compile again (part of #3047)
-getting core, nse, testbed, ats-tool and peerinfo-tool to compile again
(part of #3047)
-getting dht to compile again (part of #3047)
-getting topology to compile again (part of #3047)
-getting regex to compile again (part of #3047)
-fixing build issues in fs/set related to #3047
-fix FTBFS
-fixing build issues in namestore related to #3047
-getting psycstore to compile again (part of #3047)
-fixing build issues in vpn related to #3047
-fixing build issues in exit related to #3047
-fixing build issues in pt/exit/vpn related to #3047
-fixing build issues in consensus related to #3047
-towards fixing FTBFS in experimentation
-ignores
-fix ftbfs
-fix test FTBFS
-fix ftbfs
-fixing FTBFS
-fixing FTBFS
-fixing FTBFS
-just indentation
-just indentation
-messing up experimentation some more, towards using egos instead of peer
identities
-fixing misc FTBFS issues
-fixing last FTBFS issues
-ignore
-fix experimentation ftbfs
-fix warnings
-fix warnings/ftbfs
-legacy, again
-ignore
-test_namestore_defaults.conf does not exist, use test_namestore_api.conf
-spell param(eter) correctly
-fix config
-delete does not work in the strong sense due to caching of the encrypted
blocks, adjusting tests to reflect this
-address #3050, refresh encrypted blocks from plaintext blocks during
iteration, compute block expiration time correctly
-add configure tests for opus/pulseaudio for conversation
-towards canonical header structure for conversation
-fix warning
-adjust includes, add missing licenses
-remove generated binary
-ensure external symbols have proper prefix for conversation service
-indentation, minor stylistic fixes
-fix license
-fix indentation, minor leaks
-proposing improved conversation API
-split off microphone/speaker APIs
-implement microphone library
-fix typos
-implement speaker library
-tool to record and replay, to test microphone and speaker logic
-fix build issues
-fix helper: properly abort pending write tasks during helper destroy
-docu helper api better
-do cancel read task as well
-fix logging
-use util's mst, remove broken mst.c
-debugging logic
-doxygen, indentation
-much better latency
-trying to play a bit with the options
-nicer messages
Andrew Cann wrote:
-do not read via mst if there is no callback for reading
-ipc message design
-ipc message design
-finishing first round of conversation phone API implementation
-note on todo
-towards caller API
-more work on call API
-more work on call API
-mostly finished call API
-removing header to avoid double-encapsulation overhead when going from
helper to API
-service skeleton
-fix includes
-fix IV
-fix test vector
-also build meshenc lib
-reset statement
-reindent
addressing name conflict with PS_STOPPED on MinGW
-fix warning
-add missing #defines
-simplify script
-add logic to print expiration times for records
-do set expiration time when storing records, works better that way...
-initialize flags
-sync before server reboot, work on conversation service
-indentation improvements
-sending RING signal
-transmit hangup
-more control logic
-handle audio in mesh
-finish first draft of conversation service
-cloning gnunet-conversation.c
-updated conversation command line tool
-misc minor fixes
-doxygen/indentation
-cleaning up command-line tool
-man page for gnunet-conversation
-removing references to dead manpage/binary
-fixes to build system, integrating conversation in non-experimental
builds
-rename fest: new to default
-removing old legacy conversation API
-enable address printing
-enable printing and parsing of conversation's PHONE records
-fix ftbfs
-add newline
-logging, autostart
-fix double free
-misc minor bugfixes
-fix message type
-fix message type
-fix double frees
-fix getting peer identity via mesh_get_info
-update to reflect mesh API change
-disable phone on reconnect
-fix
-fix mesh tunnel destruction
-detect tunnel creation error
-fix double-free, clarify API
-ignore
-doxygen
-implementing revocation library
-handle client revocation queries in service
-doxygen fixes
-doxygen fixes
-doxygen fixes
-improving docs
-doxygen, coding conventions
renaming GNUNET_CRYPTO_get_host_identity to
GNUNET_CRYPTO_get_peer_identity
-load revocations from disk, misc doxygen/style fixes
-store revocations to disk
-update dependency graph
-updating structure
-removing dead lib header
-more dead code elimination / house cleaning
-adding skeleton for gnunet-revocation command line tool
-first draft for revocation command-line tool
-die
-also dead
-add missing pkgconfig generations
-add missing pkgconfig generations
-ignore
-install fcfsd in libexec, build dv by default, move fcfsd config to
namestore.conf.in
-remove trailing whitespace
-clarifications
-more whitespace
-adding man page for revocation, removing obsolte man page for RSA
-fix config, fix ntohs, must be ntohl
-finishing revocation command line tool
implement revocation broadcasting (part of #3057)
rename coreAPI to core_api to follow naming conventions better
-transmit termination reason also for caller hanging up
-moving friends file parsing logic into its own library
implementing FRIENDS file writing logic
-doxygen
-shorter line
-fix friends file parser logic, a NOT operator is a wonderful thing
sometimes
-document UTC vs local time
more strings documentation
implement #3035: add mimetype:type category keywords, also doxygen fixes
-ftbfs
should include util_lib, not common
-do not include twice
-actually just include util
-implement #2992 (but reverse lookup itself seems to not yet work
properly)
-ignore
fixing #2992: do not use hash of pkey and then match against pkey, that
will not work...
-bump name to 97
towards implementing postgres backend for namestore (#2987)
fixed #2987 -- at least the testcases now pass
-preping for XDG
fix script to use file names from configuration and support -c option
migrating towards XDG configuration specification (#3000)
-fix
-clarify comment
more work towards XDG spec (#3000) -- config file location
last XDG change: use GNUNET_RUNTIME_DIR instead of /tmp for UNIXPATHs by
default
-use get_value_filename with UNIXPATH to get himBHexpansion
-do not autostart revocation
-xdg changes: docu
create directory if one does not exist before trying to output
certificates
-fix header inclusion
-say that we ignore retval on purpose
-fix warning
-switching (again) to named sockets, see #2887
-rename serverAddr to server_addr to follow naming conventions
changing UNIX domain socket access control to file permissions checks,
instead of UDS credentials (#2887)
removing remenants of abstract unix domain socket handling, this finishes
addressing #2887
-replace stderr with /dev/null, instead of just closing
-fix warning
-fix leak
-fix leak
separating ECC crypto into functions/structs for ECDHE, ECDSA and EDDSA
-updated tests
-ignore
-fix eddsa api migration
-fix ftbfs
-fix ftbfs
-fix eddsa ftbfs
-fix #3037: support IPv6
-eddsa not yet in gcrypt
-remove legacy
-just .conf, not .config
-doxygen, code style, indentation fixes
-doxygen / coding convention fixes
generate progress events when publishing directories (towards fixing
#2230)
-fix crash on recursion with NULL def
implementing #3064: enable split-personality ARM to run some services as
'gnunet' and others as grothoff, depending on level of personal information
available to the service
-fix possible NPE
-fixing minor memleaks
-fix leak, doxygen
-fix test ftbfs
-needs env to link properly
-fix includes
-doxygen
-pogen run
-need to do mkdir before testing UNIXPATH, not sure how this worked before
-mqtt needs a port
-doxygen fixes
-create directory, 2-digit CC
-updated libcurl API, do IPv4/IPv6 with separate sockets
-moving code to new libgnunetgnsrecord in preparation for addressing #3052
-rename fest for symbols moved from GNUNET_NAMESTORE_ to new
GNUNET_GNSRECORD_ library
-specify gnsrecord plugin API
-define gnsrecord plugin for DNS
-adding plugin for fundamental gns records
-add gnsrecord plugin for conversation
-updated structure
-SERVICEHOME is now GNUNET_HOME
-use plugins, fixing #3052
-fix shutdown
-start tests with fresh/empty database
-make tests executable, enable himBHexpansion
-correctly implement handling of '+' records
-copied block-related functions from namestore to namecache
-also generate makefile for namecache
-moving namestore_common functions to gnsrecord library
-another renaming fest for GNUNET_NAMESTORE_ to GNUNET_GNSRECORD_ symbols
that were moved
-build namecache by default
-finishing split of namestore into namestore and namecache (#3065) -- in
theory; in practice, somehow something broke badly, so the tests are now failing
-not a user service, this was the point of splitting it off
-dead
-dead
-add man page for namecache
-style fixes
-fix plugin linkage
-start namecache as well, perform NC registration for client during
lookup handling
-perform NC registration for client during store handling
-fix use after free, cancel of dead task
-start namecache as well
-start namecache as well
-indentation, code cleanup
-fix testcases
-doxygen, etc
-need to connect
-fixes
-fix linkage
-fix linkage
-minor cleanup
-start each round with infty as the minimum interval, otherwise it would
always be zero
-nicer
-remove expired records immediately after decryption
-check for expiration and do not cache expired records; allow
zone_publish_time_window to go back up, so reset it on new iterations
-fix testcase to not use expired records
-better sleep pattern
-check for DHT refresh, let it expire in the DHT once
-fix termination
-convert to absolute time instead of failing
-conversion to absolute time is now done in GNUNET_GNSRECORD_block_create
-do not use expired records, they'll be removed internally now
-use fixed records
-use configuration option for reset of zone_publish_time_window, also
frequency should be often enough to put every record 4x within its expiration
window
-ignore
-doxygen fixes
-fix bad allocation and message size calculation
-fix leaks, add assertions
-adding proposed namestore lookup call (API only for now)
-removing bogus comment
-fix indent
-fix leaks
-improve logging
-check return value
-dead code elimination
-trying to implement revocation
-fix testcase
-fix indentation
-fix GNS2DNS resolution, make SOA test pass
-towards implementing #3051
-enable monitor over all zones
-allow namestore to monitor ALL zones, and to optionally only monitor
changes
-simplify to use new namestore API to skip initial iteration, document
now passing quickupdate test better
-32 bit, not 16
-implement certificate validation against DNS names and for LEHO records
-adding logic for TLSA/DANE verification in combination with recently
submitted patches to GnuTLS
-require gnurl
-whitespace
-moving from libcurl to libgnurl
-bump gnutls requirement
-eliminate dead code
-doxygen
-moving code to parse IPv4/IPv6 subnet specifications to strings and
making it part of the public API
-converting regular expressions of vpn/pt to non-binary format and adding
proper policy parsing
-faster port range generation, O(n) instead of O(n^2)
-make proper use of new policy parser
-improvements to parsers
-log parse errors as WARNINGS
-better comment
-better doxygen
-indent
-match exact types from message struct
-man page for gnunet-conversation-test
-fix gnurl checks
-more fix
-getting rid of some warnings
-simplify utf8_tolower/upper APIs
-doxygen
-do not set PENDING on shorten
-handle PENDING
-florian wants sane dot graphs
-improving API design
Werner Koch wrote: Hi,
Werner Koch wrote:
-license bump
-initialize flag, do not use expired record for testing
-initialize flags to unproblematic value (hence usually zero)
-fix include order
-now featuring short peer identities, yepee
-use valid public key in new format
-fix cname test
-clarify docu
-tolerate lower case letters
-fixing zkey resolution
-update ignores
-stylistic fixes
-do not start those during testing either...
-addressing final crypto fixme
-bump configure test for an even more recent libgcrypt symbol
-implementing updated GNS plugin code for new GNS2DNS records (#3093)
-improve testcase to actually test #3093
-rename testcase to match record type
-fix crash on GNS2DNS resolver completion, and properly translate names
ending in .+, fixes #3093
-fix SOA test to work with new GNS2DNS record format
-allow tests to pass if system is offline
-fix doxygen
-fix #3095
-fix confusion between service name and default ego name
-fix shutdown
-dead
-do not allow SSL connections if we are only given an IP address by the
browser, as then we cannot check certificates
-complain if this happens
-fix potential leak and awkward test behavior causing it
-operations can be there, but none with active cont
-fix memory leak on shutdown (3104)
-fixing crash in #3101, but just replacing it with a break, so there is
likely still an issue left
-fix minor leaks in gnunet-arm (#3105)
-doxygen, memory leak in heap plugin on shutdown
-check for NULL, doxygen
-do not call with NULL
-extra checks
-fix memory leak
-remove noparam, as suggested by Werner Koch
-avoid bus error on sparc
-fix message alignment to avoid bus error, indentation, style fixes
-align message buffer
-make buildbots happier
-be less insanely verbose, especially wrt util
-only run tests if we actually had pulse/opus
-multiply infinity without warning
-fix #3101
-fix warnings, use proper types for addr
-towards enabling call waiting in conversation -- creates FTBFS
-fix FTBFS
-reworking connection to allow multiple waiting/active connections per
line
-updating gnunet-conversation to match current API
-implementing missing resume/suspend functions for callee
-completing API implementation
-build fix
-doxygen fixes
-doxygen, minor bug fixes
-only notify AFTER sending is really close to finished, not before
-fix include order
-remove GNUNET_MQ_impl_send_commit, make it part of send_continue, to
ensure calling the continuation _last_
-avoid possible problem due to clean up during deeper recursion by making
continuation asynchronous
-return after handler was found, as handler might have destroyed the mesh
handle already
-avoid use after free
-do not send done after destroying channel
-fix
-clean up during shutdown
-fix missing endianess conversions
-fix shutdown
-add assertions to ensure there is no stale reconnect task
-more debug messages
-use two-digit country code, as required
-fix test
-towards a real test for conversation
-expanding conversation test
-document test
-doxygen, and likely fix for #2727
-fix
-doxygen fixes
-fixing crash from reset_mesh due to mesh API semantic change: cleaner is
now always called, but we don't want it to release the maps in this case
-not const
-fix channel_get_info, return pointer to peer, not interned value
-cleaner callback sequencing
-fix test
-fix
-fix
-do not crash on exit, even if mesh died
-implementing revocation set union
-doxygen
-handle case where channel is NULL (failed to be created?), see #3125
allow spaces in binary file names (#3094)
-attempting to address #3133
-fix #3141
-cleaning up UDP broadcast code, preparing for cryogenic
-fix leak
-fix bad sizeof
-check return value
-doxygen
-adding missing close
-fix memory leak
-check return value
-fix left-over confusion between hash code and peer identity size from
recent peer identity format change
-doxygen fixes, fix starting gnunet-service-resolver as it is no longer
in
/opt/jdk1.6.0_22/bin/:/home/grothoff/bin/:/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games:.
-initialized uninitialized 'key' field in GetHandle to fix assertion in
gnunet-service-dht_clients.c:721 due to uninitialized key in IPC message
-doxygen
-doxygen
-doxygen and logging
-doxygen
-fix #3133
-fix #3147: mesh server should properly interpret 'no answer' from
datastore, instead of passing along the 'emtpy' block
applying patch from #3148 (by shedatc) to make gnunet-helper-nat-client
compile on FreeBSD
-undo, this was a bit too early
-card template
fixing #3146: peer identities no longer have the size of hash codes;
breaks FS protocol compatibility with previous versions slightly...
-doxygen
-html form
-run refresh_routes() asynchronously, as the recursion in #3152 looks
suspicious
-fix paths
-reduce stat calls by caching result of GNUNET_OS_installation_get_path
-template for supriti/xvine
-make cryogenic code Linux-only
-rename and install
-degermanize
-dead code elimination
-improving TeX
-dead
creating gnunet-bcd
adding man page for gnunet-bcd
only try to open /dev/power if it actually exists
-fix timeout
-fix #3153
-fix #3157
-safer, found by clang
-dead assignment, reported by clang
-dead assignment, reported by clang
-dead assignment, reported by clang
-dead assignment, reported by clang
-dead assignment, reported by clang
-dead assignment (and leak), reported by clang
-dead assignment, reported by clang
-dead assignment, reported by clang
-fix leaks
-fixing indentation
-indentations, comments on API
-do not leak 'elements' array
-indentation fixes
-do not leak 'mask' array
-fixing indentation/doxygen/style issues and some memory leaks in
gnunet-consensus-profiler
-check return values of hashmap operations
-fix leak
-check return values of hashmap operations
-fix leaks on error-return, fix indentation
-simplify code
use 'svalue' that is actually a signed integer, as otherwise cmp with 0
is always false
-typo
-add missing break
-add missing break
-fix allocation type
-use doxygen-style comments on membes
-fix leak, use 'op'
-address warning, indicate not checking return value is intentional
-trying to fix leaks
-adding missing break statements
-fix leak (rather theoretical one)
-fixing earlier fix
-always initialize emsg_
-r was captured in request, do not free here
-r and r_prime were captured in request, do not free here
-fix leak of emsg
-fix leak, indentation, make start_cb static
-initialize to NULL to use non-free if 'goto except' is hit
-fix leak
-fix leak
-use modern GNUNET_new, instead of GNUNET_malloc (sizeof
-fix #3158
-dox
-fix indentation
-move base64 encoder/decoder to util
-adding CERT record support to dnsparser
-adding CERT support to DNS plugin
-note missing features
-doxygenfixes
-perform mic/speaker operations before event callbacks
-test two incoming calls at the same time, plus suspend/resume
adding test for immediately rejected phone call
-doxygen
-remove object from all_routes peer map before freeing it, should fix
#3161
-fix #3152
-doxygen and style fixes
-correct use of consensi
-default to https
-set session to active on initial creation and notification of the
transport; this should fix #3162
-fix
-freebsd wants malloc_np for malloc_usable_size
-test for tm_gmtoff and use it instead of 'timezone' as suggested by GNU
libc manual
fix use-after-free: the call to mesh_channel_destroy will ultimately free
'op', so we must not update 'op' after the call; however, it is a good idea to
clean up the 'channel' field just before the call (#3164)
-fix trailing spaces
mark timeout task as done if it is run, fixes #3171
-doxygen and formatting and style fixes
-minor code cleanup and doxygen/style fixes
-fix #3168, cancel retry task AFTER GMT_cancel, as that triggers a
callback in gnunet-service-mesh-peer.c:952 which may then CREATE a new task
fix #3170, again by cancelling scheduler tasks last so that they cannot
be re-created by callbacks triggered during shutdown of the rest of the struct
-ignore
-ignore
-rename enum MeshOption to GNUNET_MESH_ChannelOption to ensure we use
GNUNET_MESH_ prefix as we should
-doxgyen, indentation, code clean up
API cleanup, rename GNUNET_TRANSPORT_check_neighbour_connected to
GNUNET_TRANSPORT_check_peer_connected
-complete comment
-cleanup
-cleanup
-cleanup
-style fixes
-make checkers happy by always initializing map, also extra check just in
case the branch conditions are not quite right
-clarify prescedence
-clarify prescedence
-assert MHD_VERSION is always above 09E00, remove dead struct
-remove duplicated return
Stephan Posselt writes:
-set charset to utf-8
-increase timeout for slow systems
-update bcd template and adjust code accordingly
-doxygen, indentation fixes
-doxygen, indentation fixes
-doxygen, indentation fixes
-doxygen, indentation fixes
-bump deps
-doxygen fixes
-doxygen fixes
-consistently update consensi map whenever peer distances change (should
fix #3183)
-minor fixes
-fix check, do not drop our own messages, but count foreign drops in stats
-fixing #3181: close all connections on DV service reconnect, also track
active connections and check that transmission requests are valid
-trying to fix #3189
-fix
-do not crash if we get NACKs or ACKs for disconnected peers
-ignore unmatched NACK/ACKs
-fix distance
-fix #3192, as suggested
-misc dv fixes
-do disconnect after continuations, to hopefullyf ix #3194
-doxygen
-fix leak
-style and doxygen fixes
-avoid duplicate remove
-close /dev/random on shutdown
-fix path
-fix use after free
-indentation, doxygen, logging
-remove find() forking, we pretty much should not need this anymore, and
it confused users in the past
-remove useless/broken configure options, and gnunet_directories.h
-indentation, doxygen, logging
-more logging, etc.
-fixing #3202
-undo, breaks multiuser more...
-trying to fix #3202
-dead
-extend transport plugin API with capability to just disconnect a single
session (not yet used)
-fix NPE
-be stricter during handshake, close sessions with broken interactions
aggressively
-no vg
-dead
-doxygen
-fix hostlist warning from #3206
-fix leaks
-fix leak, style
-fix leak
-removing code that just dereferenced uninitialized 'm' and then decided
to randomly increment memory
-fix leaks of MPIs
-fix
-doxygen, style, possibly fixing hostlist-client missing first core
connect notification
-missing header
-fixing includex for xdht
-fixing 'make dist' issues
-doxygen, style fixes
-update doxy file
-code cleanup, fixing #3207: timeout can be NULL in
reschedule_session_timeout if we are 'in destroy', in which case we should
simply do nothing
-code cleanup, fixing #3207: timeout can be NULL in
reschedule_session_timeout if we are 'in destroy', in which case we should
simply do nothing
-undo
allow distance change to communicate changes in network of next hop as
well, related to #3191, but not sure if it fixes it
-adding a few more assertions to debug #3191
use enum GNUNET_ATS_Network_Type instead of uint32_t where appropriate
use enum GNUNET_ATS_Network_Type instead of uint32_t where appropriate
-sometimes bugs are really simple -- fixing #3191 but creating fresh
crashes
-fix 3208 by making kills async
-fix #3211
-fix off-by-one so we can also select on FD 0 (stdin)
-clean up network select code, avoid insane #ifdefing for MINGW
-fix #3212
-remove break for something that theoretically could be OK
-move tcp session check into extra checks condition
-crash if this happens, so we at least get a core dump
-removing useless session counting logic, doxygen, code cleanup
-code cleanup, in particular moving towards using client context for
storing TCP sessions to be more efficient upon receiving
-fix warning
-doxygen, indentation, minor code cleanup
-doxygen, indentation, fixing comments
remove external IP address from previous DNS lookup only after DNS lookup
for new iteration has completed (#3213)
-doxygen, indentation
-simplify logic to help diagnose #2858
-ignore
-cleanup, doxygen
-fix FTBFS
-ignore
-move to experimental
-do not (auto)start certain unrelated services
-adding example 'def.tex' file
-fix message types in comment
-fixing leaks on error paths
-fix use of uninitialized variable
-indentation fixes
-never store NICKs anywhere but in '+', do not display nicks in
gnunet-namestore
-never store NICKs anywhere but in '+', do not display nicks in
gnunet-namestore
-no warning if dir does not exist
-reduce testcase output
-do not log at DEBUG in testcases (at least not in SVN), do not print
random stuff to stdout
-require gnurl 7.34
-fix #3220: typo in SQL
Fix assertion failure seen on buildbot ubuntu-armv71-evans, build 769:
-trying to fix double-cancel
-ignore
-pi can be slower than this (see build 1241)
-pipe garbage to /dev/null
-handle errors in address part
-clean up error message
-require just sh, not full bash
-bash is in deed required
-properly unload incomplete plugins
-complete API implementation
-fix endianess before printing DEBUG message
-bump
-bump
-bump
-misc fixes
pass ATS information for incoming sessions immediately to ATS service
-more debugging, fixing targets
-learn routes from forwarding/receiving as well
-fix distance calculations / ntohs vs ntohl
-style fix
-update PO
-towards changing use of target distance to be from source (incomplete)
-misc fixes
-style fix
-style fix
-fix transmission logic in DV API
-more logging
-consider possibility that DV message arrives after direct connection was
established
-fixing #3222
-doxygen, fixing #3223
-undo
-fix single-page layout to include gpg key
-check for 'timeout'
-gnunet, not gnunetd
-larger delay for Pi
-reduce workbits to be more Pi-friendly
-adapting style for 5x2 86x54mm bcs
-undo, rd[i].expiration time are in host-byte-order
-fix integer overflow on 'forever'
-add missing alias for namestore
-indentation
-enable url
-set receiver handler to NULL on timeout
-typo
-/dev/power was renamed to /dev/cryogenic
-check for working app, use correct option
-use default policy that works now...
-give nicer warning about big regex
-fix VPN resolution to A/AAAA records
-use the same hash function to map service name to hash, not once strlen
and then strlen+1
-indentation, cleanup
-fix service name construction, remove .gnunet.
-fix nss installation invocation if nss is not to be installed
-fix DHT lookup (thanks, Sree)
-fix free after free
-support weeks as well
if records from namecache do not match desired record type, go look in
the DHT
-this should fix #3228 (communicate conversation/reconnect errors to
client)
-clean up authors file
-remove useless const
-add missing return
-fix leak
-handle partial writes and IO errors, even on stdout
-fix leak
-revocation should be in defaults
-autostart set
-use @AUTOSTART@
-fix calculation of block size, doxygen
-fix #3233
-code hygene
-code hygene
-move DV to experimental
Stephan A. Posselt writes:
-handle new error code from enum
-reduce log level
-amatus says this happens with just new peers, boo
-trying to fix amatus's KX issue
-fix casts
-debug logic for NAT external IP resolution
-remove OLD, not new
-doxygen
-more logging and doxygen fixes
-fix-ftbfs
-add assertions on non-existing sessions
-fix build system issues
-fix build system issues
-fix build system issues
-disable set
-ignores
-fix
-fix printf
-fix tiny leak, doxygen
-version bumping
-fix issues with multi-user setup and ARM
-gns is user service
-avoid bashisms
-fixes
-switch to case-insensitive LOC URIs and use compatible encoding for peer
identities with the rest of the system
Here's another patch for GNUNet's scheduler. It's a smaller version of my
last patch.
-fix check for DANE and ftbfs
-let's not introduce stdbool shortly before the release
-fixing misc format strings
-remove trailing whitespace
-fix possible assertion failure if rekeying is triggered on session that
is not yet up
-fix setfault (op->vt can be NULL)
-doxygen, indentation fixes
-make sure to link against right libgcrypt, enable abort of
gnunet-revocation even if identity service is not running
releasing GNUnet 0.10.0
-fix assertion, FD 0 is fine
specify proper access control rules for conversation service
-improve docs (#3244)
-fix build with older sqlite3
-fix npe (#3248)
FreeBSD has ucred.h in sys/ucred.h, check for this (#3249)
-modify NAT API to return error messages about problems detected
bad comment, option actually not required anymore as we do not use
VERY_STRONG anyway
-eliminate repeated calls to gcry_md_open, use reset instead of close/open
-revert 'optimization' that broke test_crypto_hkdf
adding benchmark logic for HKDF
generate compact regular expressions if range is negated (#3226), switch
default policy to not-25 (SMTP) as we can now work with that
-doxygen, indentation
-updating potfiles
-typo
fix #3272
add configure test for libogg to fix #3278
-document libogg requirement
should be uint32_t
use -Wl on -no-undefined as it is a linker option:
fix #3283
-have default subsystem name for phone
-fix #3285
-fix linker error for pallier test
-doxygen
fix #3284: support lib/MULTIARCH/ paths in installation, use
address@hidden@ so that we can extract the value of MULTIARCH during 'make
check'
towards reviving priorities in core API, this time with enum to make
classes clearer
-towards fixing #3295 (core traffic prioritization)
-doxygen
-fix typos
importing man pages written by Bertrand Marc for Debian
another one from Bertrand Marc
Bertrand Marc wrote:
-indentation, doxygen
-indentation, doxygen
removing depreated 'priority' argument from
GNUNET_TRANSPORT_notify_transmit_ready
add API for #3296
-doxygen, indentation, code cleanup
-doxygen, indentation, code cleanup
add api to bandwidth tracker to notify clients about excess bandwidth
available
implementing #3296
-eliminate odd global variable, indentation, doxygen, stylistic fixes
implement #3295: only transmit background traffic if there is excess
bandwidth
trying to fix #3281
-try to fix check for DANE if --with-gnutls is passed, also fix output at
the end (thanks to E.E. for reporting)
-fix test
force not using DHT for lookups that are only in the master zone (#3325)
-typos, doxygen
-indentation, superfluous asserts
if we found any non-expired block in the namecache, that is sufficient
reason to not go to the DHT -- and essentially quickly return NXDOMAIN (#3325)
-fix copy-n-paste-comment and doxygen
-add a few comments
add timeout option to gnunet-gns
-doxygen, indentation
make sure to create friends directory before trying to write friends file
avoid problems with different users having different values for
XDG_RUNTIME_DIR by not using that variable for system-wide paths of UNIX domain
sockets
handle case where partial download prevents proper bottom-up
reconstruction as not all child-nodes are even attempted; handle case where
probe returns results immediately while being created (do not break on that)
-disable POW calculating during test
-avoid defining 'cfg' symbol each time ats-testing.h is included (fixes
FTBFS)
-check for NULL fd
-doxygen
-mmissing prototype
-presumably better fix for instantly resolving names in master zone
-fix dependency
-doxygen
-fix removal of block_mesh file (#3281-8125)
-bump gnutls version
-fixing wrong bump
-fix ftbfs with older libgcrypt
-fix disting -- include sp
-fix FTBFS
-ensure timeout forces call to continuation even if statistics service
cannot be reached
-no set autostarting
-fix map cmd shutdown
-fix linkage issue that could cause a transport service crash on exit for
certain NAT configurations
-add missing header
-one more ATS header to be included
-ignore
-bump version
-XDG environment variables cause more trouble than they are worth
-allow users direct core access
-fix #3355: disable warning
-make sqlite3 hard requirement (#3341)
-pogen
-doxygen improvements
-doxygen
-fixing hypothetical leak on error path
-process in logical order
-bump version in C tutorial
-doxygen
-reenable shipped hellos test
-comment out all of directives on test
-version bump
-pogen
-fix build issues
-ignore
-fix test as prefix likely now has lib/ suffix
version bumping for 0.10.1
-more version bumping
-fix dep
towards fixing #3363: replacing old iteration API with new monitoring API
for core (needs testing, gnunet-core incomplete)
-also notify monitors on timeout change
-improve/simplify gnunet-core tool and update man page
-fix crash on shutdown
-do call receive done on monitor calls
-comments for surpiti
-grab source IP from outer IP header, not from inner IP header
-attempting fix for FreeBSD configure error on netinet/ip.h reported by JS
-rebuilding PDF
-doxygen, indentation, new FIXME: disconnect needs to notify clients
-signal monitor disconnect via callback
-signal monitor disconnect via callback
-indentation
-doxygen, plus FIXME request for clarification
-rename file to match updated functionality
hack namestore pending capability
-comments
-comments
-comments
-rename to conform with naming conventions
fix #3348: send typemap confirmation messages, perform faster typemap
retransmissions for unconfirmed typemaps, restart retransmissions on reconnect
-dead now
-update ignore
-fix (mostly harmless) use of uninitialized variable
-indentation, cast to avoid compiler warning
-fix compiler warning: do not return const double
-only complain about upnpc if it is actually enabled by config
-fix error message
-new hostlist
-indent
-indent, doxygen
only notify clients about significant (>5s) timeout changes
-mesh -> cadet
-note to supriti
-big cleanup of test_server code, also fixing #3394
improve error reporting for malformed URIs in search results to help
diagnose #3395
-po update
-ignore (mesh->cadet renaming...)
-LOC, not SKS
-do not require UID match for ATS
fixing #3302: match record type and value as well during deletion, if
specified
-doxygen
add logic to handle SRV/DANE names (#3003 and 2526) in GNS resolver
-doxygen, some additional error checking
cleaning up scalar product client API
-a bit more towards following coding conventions
-indentation fixes
enable DHT monitors to distinguish routing requests from requests that
terminate at the local node (for GET/PUT, not for replies)
-fmt
-fix install
no comma, breaks gnunet-statistics-gtk (fixes #3417)
-have default line, complain about missing options to log
-doxygen, indentation
-ignore
-indentation and doxygen fixes
add separate BINDTO options for IPv4 and IPv6
-doxygen, code cleanup
keep exit process running even if it cannot work to avoid ARM restarts
-indentation fixes
-fix crash observed on FreeBSD
-use GNUNET_log, not fprintf
-better logging, doxygen, indentation
-better logging
-nicer logs
-initialize cid in hangup message
-update to po
-indentation, logging fixes
-fix #3428
-rename
-remove mention of GNUNET_PREFIX, should no longer be needed
properly reset name resolution pos for CNAME handling
-report if typename given is bogus
permit 'ANY' for typename
-indentation fixes
get rid of phone ego, we use the master-zone directly now
-reindent
-doxygen, indentation
addressing #3431
-allow caller ID to differ from zone used for resolution
-ignore
-preliminary support for BOX records
-box from/to string conversion
-year
handle box records in gns proxy
towards sanity in SRV handling
-implement SRV serialization
-avoid 'hu', as it is unsigned short, not uint16_t
-fixes for consistency
fix use of %c for ints
more tlsa fixes
draft hex2bin, bin2hex
-testcase and bugfix for hexcoder
handle SRV
-fix return value check
fix boxed value parser
-indent, doxygen
-indent, doxygen
-doxygen fixes
-fix build order
-fix leak
-check return value, doxygen
-fix #3301
-fix #3301
-doxygen
-fixing doxygen, indentation
-fixing doxygen, indentation
-fixing doxygen, indentation
-die dead stuff
-making test nicer (indentation)
-fix timeout handling for resolver
clarify prettyprinter API and protocols, make sure implementations are
consistent in their implemenation, doxygen fixes, indentation fixes, subtle
semantic fixes
-code beautification
-ignore
-fix compiler warning, fix test logic
-fix log level, logging, indentation
-fix status code
-send proper termination on error
-fix counting
-fix handling
-misc fixes
-misc fixes
-fix
-cleanup
-indentation, doxygen
-make retry nicer
-doxygen
-trying to fix #3426
-simplify plugin, remove unnecessary header, make internal plugin
functions static, doxygen, indentation, code simplification
code cleanup
-revert 33649, see #3426
-fix warnings
-fix #3443
-indentation
-doxygen, indentation
-doxygen, indentation
optimization: only iterate over mamultipeermap entries that match peer id
bugfix: only create another address entry if we did not have one already
minor code cleanups, doxygen
print active/inactive information
doxygen, make local symbol static
-remove dead include
-doxygen, indentation
-doxygen, indentation, nicer stats
do not use session_id in GAS_performance_notify_all_clients, needs to be
'active'; also indentation/doxygen
notify performance monitors about destroyed addresses (for #3406)
-fix memory leak for 0==p->length
-fix use after free
-fix use of uninitialized memory
-fix socket leak in error handling branches
-check 'socket' call return value, handle failure
-handle case where sendsocket is still -1
-add missing break
-assert zone is non-null, simplify branching logic a bit
-bail out properly on error, fix leak
-fix channel data range to make -1 legal value
-check return value
-check return value
-mark return value as intentionally ignored
-check return value
-check return value
rename hellos to use new base32 encoding
output public key of peer
add check for certtool
fix #3451
-indentation, doxygen
-simplify logic a bit, doxygen, indentation
-doxygen, indentation
-split up transport_api_monitoring into monitor_peers and
monitor_validation
-doxygen, cleanup
-doxygen, cleanup
-doxygen, cleanup
-reordering API in header to make structure more obvious
adding design for session monitoring API to transport plugins; skeleton
implementation for UNIX plugin of that API; clean up of UNIX API code (removing
of duplicated/dead code)
-fix doxygen
implement monitoring API in UNIX, simplify code a bit
-actually perform monitor calls where appropriate
-updateing tempalte
-fix regex
-simplify logging using macro
-misc stuff
-fix SKS publishing in simulate-only mode
-fix URI test to handle new encoder, print both SKS and CHK URIs when
publishing to namespace
-fix time assertion introduce in last patch
-fix URIs
-removing dead config
-removing global state from UDP plugin (statics), doxygen, indentation,
starting support for monitoring
adding monitoring support to UDP plugin, plus some doxygen/indentation
fixes
make myoptions member of 'struct Plugin' instead of a global
-towards having the monitoring API supported by TCP
-indenting
-rename tests to caputre better what they test
adding missing monitoring callbacks for TCP
-fix uninit
-missing commit
-rename config files to match test cases
-use LOG macro instead of GNUNET_log_from
-doxygen, indentation
-remove logic for somewhat overkill check to see if session object is
valid
-simplify logic, just use clearly working direct method to extract
session from server's client handle
-indentation, doxygen
use session map instead of DLL
-code cleanup, doxygen
-avoid passing ambiguously typed 'cls' pointer to address_to_string
routine in http_common, code cleanup
-minor code cleanup, indentation, doxygen
-add support for 'update_inbound_delay' to HTTP client, complete plugin
monitoring implementation
-const
-move struct def to where it is used
-indentation fixes
-session header is now dead
-fix indent
-simplify deletion logic, tell monitor that we are initiator
-drastically simplify http server session cleanup (and startup) logic
-add monitor support to http server, more code clean up, add http server
MHD suspend feature, support quota reset API from transport in HTTP server
-adding missing API functions to WLAN, skeleton for monitoring support
-simplify logic
-add missing calls to monitor
combine sources for wlan and bluetooth plugin logic into one file
-tolerate it if gnunet://gns/ URI prefix is in upper case
-fixing doxygen, indentation
fix #3446
-define port for W32
-pogen
-doxygen
address Debian #758937: allow use of curl if it is really the gnutls
version and libgnurl is unavailable
libcurl.m4 added
patch was correct, but no need to have the bug as a comment
-indentation
-remve legacy option
-remve legacy option
fix assertion, ensure 0-termination
trying to do something about #3540
fixing parallel build, getting rid of useless _DEPENDENCIES declarations
fix #3544
-doxygen, code cleanup
Adding man pages from Bertrand Marc. Thanks!
-noinst
build fix from Bertrand Marc
-fix typo found by Bertrand Marc
-fix typo found by Bertrand Marc
-pogen
-is pending is dead
bratao's patch for #3545: suspend/resume of downloads
-fix NPE in NAT test
-fix format string
-return non-NULL even if we instantly fail
-set socket to NULL on close to avoid double-close, do not cleanup on
timeout, leave that to app
-doxygen
-fix return values, add missing return
-init rand
-fix leak
-check return value
-fix leak, uninit mask
-fix i/j indexing confusion, resulting in array index out of bounds
-avoid leaking trail by avoiding allocation in the first place, use
unsigned as appropriate, eliminate useless return
-remove superflous semicolon causing branch to be always taken and
significant logic to be omitted; pass pointers to HashCode and PeerIdentities
instead of making copies
-remove superflous semicolon, causing logic to be executed despite
condition not being satisfied
-remove dead assignment
-eliminate leak
-eliminate dead code
-fix leak of sockaddr
-fix leak
-add assertion
-add missing break
-actually enable logic to drop part of path to fit message size, check
against correct constant
-fix test logic
-add missing return
-fix use-after-free in error handling
-fix leak
-fix #3550
-fix dbz
-too sleepy
Hello Christian,
-do not log 0-byte callback
-towards #3452: implement monitoring API in DV
add function to obtain number of subscribers to NC
define transport plugin monitor IPC format
implementing monitoring functionality in transport service
-fix ftbfs
-doxygen fixes
-fix #3556
-fix ftbfs
adding TCP_STEALTH support to TCP plugin
adding TCP_STEALTH option to default configuration
-use const for the array
-update and improve visualizations
implementing plugin session monitoring API (#3452)
implement transport plugin session monitoring support in gnunet-transport
(#3452)
make set testcase fail to highlight issue #3490
-fix #3558
-fix ftbfs
-move welcome creation to startup
adding support for TCP STEALTH integrity protection
adding TCP STEALTH support to HTTP client (without integrity protection)
-dox
-fixes
adding TCP STEALTH support (without integrity protection) to HTTP(S)
server
-fix warning
-rename constants
-no args for -P
-indentation, doxygen
-fix message type
-fix sync signalling for plugin monitoring
-fix misc monitoring issues
-bad printf
-fix docs, fix shutdown
-fix indent
fixing collect_generation_garbage() complexity, doxygen, etc.
-do not do GC for each destroyed operation on client disconnect
-inline
-indent
-indent
eliminating useless salt argument, salt can be generated inside of
gnunet-service-set
-do not run tests if disabled
rename element's type field to 'element_type'
remove additional variant of operation state
-return const, do not mangle const input
again, do not pass salt around, code cleanup, etc.
clean up internal set API, avoid copying context message needlessly
use and respect send_more field in IterAckMessage
adding logic to allow GNUNET_SET_iterate_cancel
-simplify logic
-comment
-renaming
-dox, indentation
-rename file
-stylistic fixes
use separate struct for just sending #elements in set, and check message
size before casting
-doxygen
-fixing misc issues and bugs, including better termination logic for
intersection and salt handling
-fix nested context copy
-fix iteration set
-misc fixes, in particular actually build bf
-doxygen, style
-avoid use after free during set intersection completion
-remove dead state
-fix ftbfs
-use keep as RC to address #3561
fix #3561
-fix #3510, ensure revocation test times out
-handle msg NULL
-indentation fixes, some fixes to message size checks, some code cleanup
simplify logic
-code cleanups
-fix nat test timeout/termination issue
-fix ftbfs
-fix misc compiler warnings
-ignore
-simplify logic
-simplify logic, eleminate dead code
allocate r and r_prime MPIs before use
-fix MHD test
-fix ftbfs on W32
-remove anicent NEWS
massive rework of scalarproduct service, splitting into Alice and Bob
-mark minor todos
-dead
-ignores
-fixing minor compiler warnings
-extra checks, plus indentation fixes
-internal logic consistency check
-fixing misc minor ats test/driver bugs
-fix leak
-check return value
-fix some of the obvious issues
-bad NBO conversion
-more SP fixes and todos
fixing #3568
-zero-terminate
-fix FIXMEs
-sending session ID back over same CADET channel is redundant
-simplify further, too much checking...
-clarify docs
-indentation, doxygen
-style
-style, actually call GNUNET_CADET_receive_done to receive more messages
-initialize public key
ensure that a bogus public key cannot send us into an infinite loop
-indent
-fix shutdown
-fix leaks
-disable vg
fix #3569
-fix rps config
-fixing paillier bug, improving testcase
add paillier benchmark
add paillier benchmark
-bad file
-handle negative results correctly
-running at scale...
-code cleanup, some data
moving away from DEFAULTSERVICES to per-section FORCESTART, thus
addressing #3565 indirectly
-expect PREFIX to point to lib-dir
-missing section
-simplifying configs
-simplifying configs
adding single location for no_forcestart configuration list
-missed hostlist
respect new CADET limit
-warn on error
-could be other peer's fault
-fix bugs
-commit with dbg logic
-indentation, doxygen, style
start fewer FS probes concurrently, run them all in the same task to
avoid undue pressure on the scheduler
-bad alloc, just leaks
-fix leak
-fix leak
do check message sizes in union p2p protocol
-split off MQ API into separate C file to reduce visibility of symbols;
reindentation, stylistic fixes
-doxygen, one more statistic
properly clean up pending advertisement message requests with CORE on
peer disconnect
-style fixes
-more logging in try_transmission
trying to fix #3576
-do forcestart for gns; doxygen fixes
-fixing peerinfo part of #3559
-dead vars
fix #3570: finally remove SList API
fix key management issue with LOC signing identified in #3559
-fix crash in testcase
-doxygen, indentation
removing dead code, improving comments, simplifying iteration logic,
making iteration over all addresses also faster
-bump year
-fix test: only run test logic, not full peer
-simplify configs, do use defaults where they work / are identical
-fix test logic
-fix double free
do not poll empty database every 100ms for migration, doxygen fixes, code
cleanup
-fix bad CMP
-simplify config
-fix test: needs correct port in correct section of config file
-build scalarproduct by default
properly report NSE estimate to statistics (#3563?)
-fix leak
-fix stray semicolon
-fix leak on error handling path
-check return values
-check return values
-check return values
-fix use of possibly wrong or uninitialized session
-check return values
-fix leak
-fix leak
-check return value
-make compilers happy
-adding debug logic for #3555
-fix warning
-assign ports to sensor*
-fix bus error
-slist remains
-unnecessary
-pass peer identity pointer to clients that will remain allocated
-remove TH from map upon completion of transmission
-disable forcestart
-removing tons of options that should be dead
-ugh
-adding single file for testing without autostarts above core level
-hostlist tests need topology
-no NSE
-no nse
-no nse
-gns tests do require autostarts
-must notify client on timeout
-disable nse pow
-towards addressing #3581
-simplify tools test
-simplify tools test
-remove dead api
-leaks, leaks leaks
-check rval
-check rval
-more benchmarking results
-old-style comments, avoid duplicate comments
-disable NSE POW during cadet tests
-use const in peerstore callback
-do not use NBO double for stats setting
finishing fixing #3581, with also simplified logic
-fix use after free
-fix sec_name leak
-check NULL, check RV
-fix leak
-trying to fix transport service assertion in time.c:393: overflow on
multiply
-check return value
-fix use after free (GCCH_destroy frees 'ch', and then ch->t dereferences
it --- Bart, please check
-fix leaks
-fix leaks, check RV
-fix leaks, use asprintf
-fix leak
-check RV
-check RV
-use static
-check connection state / rval
-fix double-free
-support join_msg=NULL properly
-and this is why the NULL comes FIRST
-check RV
-check RV
-do not warn about libcurl if we have libgnurl
-do not modify 'const' argument
-static
-fix warning
-clarify logic
fix #3588: avoid bash-isms
-check RV
-fix unused initialization of locals
-check RV
-add assertion
-eliminate duplicate all_zeros globals, eliminated dead 'latency' field
in SendOKMessage
-use separate message types for TRY_CONNECT and TRY_DISCONNECT
-make static
-simplify logic
-doxygen and minor style fixes
-disable unused services during test
-assign port for DNS
-do not start unrelated services during ATS tests
-doxygen
-undo
fixing datastore-statistics interaction to ensure stats are properly
written to disk on exit
-doxgen, use modern client API
-doxygen, indentation
-cleaning up stats code, use separate structures per subsystem for faster
processing
-fix dbg level
-fix call to MHD_start_daemon, properly pass MHD_USE_IPv6 if needed
-do run with proxy type NONE
-actually, require proxy option to be empty, cleaner
fix #3589
-fix trivial double-free
making GNUNET_SCHEDULER_cancel() perform in O(1) instead of O(n) to help
or even fully address #3247
-delay serving requests until plugin is ready
-shutdown alone does not move to ready list
-use flag to check for readyness
-update priority to shutdown if that is being used
-fix doxygen issues
-stylistic fixes, also, only block re-validation for 2s IF we actually
did send a PING, not if we didn't find the plugin
-docu, style fixes
-check once, and consistently
suspend server while handlers are not in place
-simplify apis, doxygen
-do not wait for timeout on hard error
-undo, breaks stuff
-ignore
-fix compile error
-update docu to reflect current use of GNUNET_PREFIX
adding support for blind signatures (modernized version of Taler logic,
with variable key length)
adding API for incremental hashing (from Taler)
moving hash_context testcase over as well
-.po update
-clean up testcase
fixing #3610
-no hash
fix assertion violation if active call is cancelled
fix #3611
use GNUNET_snprintf
use enough space for storing full address
do not print two messages, one is enough
-oops
use ATS functions to convert quality network type to string, instead of
using string API
eliminating callback argument and its implementation which are never
called in the first place by ATS
simplify ATS API and plugin API by returning the network type, instead of
an ATS_Information struct
-dce
exit loop once first handler is found
use modern MQ API, use multipeermap instead of DLL
-fix indentation
-indentation, doxygen
-indentation, doxygen
-test cleanup
-doxygen, link against libgnunethello
-test cleanup
always use a slot, even if the session is NULL, simplify logic and
document what still needs to be fixed
-just indentation, dead code
removing dead logic to copy and pass around unused ATS values
remove ATS information from callback from ATS address selection, as the
argument was never used in the first place
-remarking on things to fix
-fix compile issue due to api change
-indentation, doxygen
-towards improved ATS API, adding return value with address record when
adding address, adding new subsystem with peer-to-address map to transport;
causes various new assertions to fail, but no major regression -- not finished
towards a cleaner ATS scheduling API
splitting add/update logic in preparation for changing msg formats
split structs for address adding and updating
simplifying IPC for address in use and address update
simplifying IPC for address destruction
also simplify message format for AddressSuggestionMessage
-documenting
-improving docs
finish ATS reconnect logic to retransmit state on reconnect
-indentation, do not pass stats arg as const just to then cast the const
away
use statics instead of singleton struct, simplify internal APIs
use const in api, avoid cast
-use static where allowed
-removing useless/dead/redundant tests, adjusting other tests to fit
current API
-eh, static sh plese
-fix testcase assertions
-fixing testcases
handle mq being NULL
store suggest handles, do not drop
only add addresses as valid from peerinfo if we have the respective
plugin loaded
-indentation plugin_transport_udp.c
-undo bad commit
-indentation
always create session before giving performance metrics
-use static where applicable
-pass closure, indentation fixes
-indentation
-address dying doesn't depend on bandwidth suggestion of ATS, just on
in_destroy flag
fixing wrong assertions
fix shutdown order to destroy sessions first, fix logic failure with
respect to bi-directional transports like UNIX
-stupid check, remove
-correct a few overly optimistic assumptions about what can legitimately
happen...
-make sure we get the correct errno back after
GNUNET_NETWORK_socket_create()
-fix warnings created by ARM opening too many sockets with limited
sockets testcase
-udp broadcasts should not be flagged as inbound
-style fixes
-constants first
-fix indent
patches from Hellekin to fix #3577
-do not send any messages until we have our own MAC from the helper
-fc not done here, do not null
-fix logging, set quota to higher for testcase
-fix logging, set quota to higher for testcase, also WLAN
-update ignores
add GNUNET_CONFIGURATION_get_data from Taler
adding GNUNET_CRYPTO_eddsa_private_key_from_string
add GNUNET_STRINGS_data_to_string_alloc from TALER
-fix leak
-fix use after free
cleaning up UDP code to use only on address format throughout the code as
much as possible
fix #3630
-reduce log level, this is expected on shutdown
-improve logging to help diagnose #3593
-doxyen, comments, indentation
avoid duplication from storing PID in both address and in validation entry
-cleanup
avoid passing both PeerIdentity and Address (which contains PeerIdentity)
if address is always present as then this is redundant
-removing dead code, chvc_ was not used at all
add statistic on size of validation_map
remove blacklisted addresses immediately from validation map
check that we support the specified plugin early, not late
GST_validation_set_address_use does not need the session
get_address_latency also does not use session
get_address_latency is actually dead
reduce callback insanity for callbacks that are always bound to only one
value
dramatically simplify quota notification: avoid registration, simply
always notify plugin
-fix fix
adding statistics on why transport may drop ATS suggestions
do signal disconnect to transport service, even if specific address is
in_destroy
when ATS suggests an address that neighbours does not have a peer for, do
not ignore it, setup the neighbour record instead
-fix FIXME
document API, do not pass unused 'session' argument
-document struct
separate ATS connectivity suggestions from ATS scheduling API
separate ATS interface scanning logic from ATS scheduling logic
-mesh is dead
-dead code elimination
-fix NPE
-assert to check API use
never destroy sessions on free_neighbour(), they should be destroyed by
their own timeout
get rid of unused state and dead code: GST_neighbour_get_latency is never
used
-docu
add internal API to enable telling ATS about 'failed' suggestions
-forgot to add file
doxygen, new define
remember which port we picked at random on port-autoconfig
also allow local addresses for P2
cleaning up test a bit
Various changes:
big ATS refactoring, no serious semantic changes should stem from this
-cleaning up gnunet-service-ats_addresses.c
-cleaning up gnunet-service-ats_addresses.h
-cleaning up gnunet-service-ats_connectivity-*
-cleaning up gnunet-service-ats_performance-*
separating out feedback from preferences
-move header
cleaning up gnunet-service-ats_normalization.c
-misc cleanups
avoid void *, use proper type for client
dead code elimination
dead code elimination
starting with cleaning up gnunet-service-ats_preferences.c
starting with cleaning up gnunet-service-ats_preferences.c
first pass at cleaning up ATS plugin API
-fix segv
-use only one nc per client
get rid of stupid network[i]=i array
-fixing silly NPEs
-fix misc minor crashes
-fix ril delete address handling
test ATS tests to pass again
-sane timeout
simplify logic
simplify logic
use lower-case
simplify logic: just use string directly for the plugin name
minor cleanup
deduplicate code for quota parsing
deduplicate code for quota parsing
DCE
-bringing copyright tags up to FSF standard
-bringing copyright tags up to FSF standard
-updating gettext m4 macros
more dead and duplicate code elimination
simplify normalization logic, also have clients access norm array of
address directly
fix testcase FTBFS
-rename for naming consistency
-fix config, shutdown issue
-adding comments, minor cleanup start
cleaning up gnunet-service-ats_preferences.c
cleaning up gnunet-service-ats_preferences.c
cleaning up preference_peers as well once they are no longer needed
-clarify API
adding FIXME about 'struct Session'
expose connectivity desire of clients to ATS plugins
use new 'get_connectivity' API which is also RC-counting, instead of
another map
avoid several linear scans and allocations by combining AddressWrapper
and asi structs
removing dead argument
check for nan
cleaning up address_decrement
cleaning up get_active_address
clean up find_quality_property_index and get_active_address_it
eliminate super-lazy functions
-remove fprintf
-restructuring, adding comments
-simplify logic, convert breaks into asserts
changing s_get API to return void instead of address, have plugin use
existing mechanism to signal address activation
-indent
-useless assertion removal
-simplify logic
-simplify flow
-simplify flow
do not force disconnect just because client does not request address
anymore
-simplify flow, redistribute bandwidth after adding address
-getting there...
-further simplify
-further simplify
-further simplify
-when selecting addresses, kick out peers with insufficient support if we
are at the limit
-cleaning up bandwidth allocation routine
-updating comments
-fin, mention future work in TODO
-fix shutdown order
-fix shutdown
-that the tests do not catch bugs like this is a real issue...
-use backoff for reconnects
-better debug output, indentation
-instantly report addresses during NAT initialization if possible
-tolerateNAT instant reporting
-do only send HELLO to clients once on startup, do only send fully
initialized HELLO to clients if we can help it, even the first one; do not
perform blacklist check twice for try_connect
perform proper locking and unlocking
do proper check for stability, in particular UINT32_MAX-0 is not a delta
of 1...
-minor indentation/logging fixes
-do not continue in case of transmission error
-report if testcase failure is because switching is not enforced, then do
not fail hard
-reorder tests to have slow ones go last
-eliminate dead argument, do use quota immediately
-improve UDP logging
-fix
-just eliminate useless check, can just be false
-do not reschedule validation if in_use did not change
fix #3664: during PS_RECONNECT_ATS, primary address may be NULL, causing
an NPE if we continue here
-fix #3665: hello can theoretically still be NULL
-fixing some memory leaks from #3667, also reindentation and code cleanup
-fix typo
-check RV
fixing #3657 (replace ATS_Information with struct), but WIHTOUT fixing
ATS testcases yet
-fix crash
-fix NPE
pass correct closure
do not call receive before sending first message
eliminate dead call from API
-dead flags
-check for in_destroy when looking up session IDs
new testsuite, obsoleting (most of) the old one
-also obsolete
-ignore
-fix badness
-testcase was long dead
-testcase was long dead
-testcase was long dead
add command to test reservations
add proper test for bandwidth reservation
-remove unused exported symbol
finish implementing reservation test, and fix bandwidth reservation logic
-allow UINT64_MAX without crashing
-allow UINT64_MAX without crashing
-style
-improve logging
keep track of number of temporarily blocked addresses
reset blocking time for addresses upon successful connection
-better stats for KEEPALIVEs
-better stats for KEEPALIVEs
-fix typo
generate our own HELLO before starting validations
-fix shutdown issue
-nicer logging
log TCP server suspends
-fix leak on transmission error in combination with disconnect
-fix memory leak of address on shutdown
-reduce loglevel (#3682)
-fix callback after cancellation issue
simplify logic, avoid using uninit variable
-fixing misc minor bugs and style issues (unchecked return values, memory
leaks, dead code
-fix
-tolerate NULL
-doxygen
fixing #3683: proper counting of TCP connections
fix blacklist checking logic, integrating tracking of sessions with
blacklist module and fixing dangling session issue which caused misc. problems
when blacklists were in use
trying to fix #3687/3683/3684 by avoiding server's client reference
counting (GNUNET_SERVER_client_keep) which should not be needed in the first
place
-fix scope assignment
be more careful with errno, so we can be sure it is not overwritten by
'free()' and the like
fix use-after-free on exit
-signal connection failure to receive even if receive is triggered after
failure is observed
-indenting...
fix TCP plugin RC issue (#3687) by changing util API (!)
-stylistic fixes
simplify logic, do not report monitoring events for sessions in destroy,
indentation, doxygen -- may help/fix #3591
adding additional checks for invariants to help find #3690
disable solver during aging, ensure there is only one aging task
fix #3688
initialize 'client', so that disconnect can remove entry from DLL, may
fix #3689
Fix use after free:
-doxygen, i18n and indentation
-fix indentation, logging, i18n, doxygen
ignore EAGAIN from accept()
-check on cur_connections
-shorten stats labels
-dead
-dead
-updating report.sh
-no longer needed
generally use GNUNET_assert() instead of GNUNET_abort() to also log the
error
-count total TCP handshakes
notify monitor about existing connections
call GNUNET_SERVER_receive_done() also on internal error paths
-indentation, comments
-fix ftbfs with debug option
-improved FD issue debug support
-extend block API to allow specifying flags, use this to allow reducing
checks for replies from local DB
count number of pending replies and refuse to process queries if queue
gets too big
-simplify logic
-simplifying event loop for UDP, separting v4/v6 for better performance
(in theory at least)
major code cleanup in UDP plugin, seems to also fix bugs; specifically, I
think I fixed a leak
Fix curl SNI handling with LEHO. (by Martin Schanzenbach)
-reimport from 0.8
-doxygen, comments, indentation
define quota message to transmit quota requests, towards addressing #3652
make output of gnunet-transport -P more readable (#3662), and in
particular support -n and defer printing until address resolution is complete
finally fix #3698
fix #3649/#3591
possible fix for #3690
trying to fix #3702 by removing session from session map earlier -- and
adding additional check for messages being enqueued for destroyed sessions
fix #3701
-fix segv
-fix misc disting issues
add --enable-wachs option
add --enable-wachs option
-fix dist issues
-ignore
fix #3704
improve test to check set union operation and to test with empty sets
-indentation and comment fixes
trying to fix #3694: reduce duplicate ACKing
this can be very harmless, reduce logging
further delay ACKs if we get duplicates
simplify test, log failure cause
act on received SessionQuotaMessages
-try to avoid being stuck with key sent/received if PING/PONGs are lost
-always assume at least MIN bandwidth, even if we got no quota or a quota
lower than min
-proper log message
updated DE translation from Mario Blättermann <address@hidden>
finishing implementation of #3652
do not do cryptographic verification for results from datacache
-doxygen and stylistic fixes
add GNUNET_CRYPTO_rsa_public_key_hash
-doxygen fix
fix testcase failure if firefly is localhost
-minor fixes
-include conf files in dist
-disable tests we know to fail until MW hopefully fixes them
try to fix timing issue in ATS test library
-try to fix this again, other call to run_interpreter() this time...
preserve errno
check for existence of 'getopt' command line tool
-more logging
consistently call deq before qc
-undo
-undo
-get test to work, but with ecdsa instead of eddsa
-get test to work, but with ecdsa instead of eddsa
-fix makefile
-also cover private key case
fix #3709: bound encrypted message queue
-simplify logic
fix #3721
-fix counting
-eh, can happen
-update doc
-start DNS service to do external evals even if we have no hijacker
return GNS2DNS records if explicitly requested
fix segv if additional records are more than authority records due to
field confusion
-fix misc core issues
-no NSE
-clarify error
-doxygen and indentation
-re benchmarking and adjusting MAX_QUEUE_PER_PEER
-re benchmarking and adjusting MAX_QUEUE_PER_PEER
-do not autostart new services either
-trying to fix warnings
-more cosmetics to reduce compiler warnings
-fix bad array init
-fix doxygen, make function static, fix indentation
-updated french translations
-trying to fix AE's problem on Guix with more fancy retransmission logic
add multihashmap_clear to API
address #3745
-updated French translations
-doxygen, build system fixes, minor API extension
-doxygen, indentation
-indentation and doxygen fixes
-fixing PORT incongruency (LRN's W32 issue)
-static
move ARM/config logic to util/ where it doesn't need to cast away a
'const' and is _slightly_ less likely to be forgotten, also add big fat comment
to hopefully clarify why it is there
starting with whanu
eliminating dead HELLO logic, not needed for x-vine/whanau
-also don't need xdht routing subsystem for wdht
-reducing to required parts of skeleton
-fix FTBFS from wdht
-dead decl removal
-work around ftbfs
-sending messages / starting random walks
extending datacache API with function to return random element,
implemented (only) in heap plugin right now
-add counter for #items
-doxygen, updating template to reflect extended API
implementing sqlite version of get_random
-doxygen
-add counter for #items
-doxygen
implementing postgres version of get_random
-doxygen
adding API to get random key from datacache
-doxygen
-lower to debug level (#3783)
-fix ftbfs issues
-major wdht hacking / refactoring -- still not finished
adding GNUNET_CONTAINER_multipeermap_get_random
implementing 'get_closest' API for sqlite and postgres datacache plugins
implementing 'get_closest' API for template
-towards an API for the successor set
-fixing 'delete_trail()'
-more datacache integration work
-towards tracking paths and trails via datacache in wdht
-implementing trail timeout handling
-nicer indentation
-fix ftbfs
-more stuff
-more stuff
-simplify socket NULL check
handle case where type is neither ENCRYPTED nor AX
-fix sizeof() arg
-select random
use NBO for expiration time in xvine
-fix leak
-fixing cleanup issues
-fix doxygen
-fix doxygen
-fix FIXME
-pass trail paths to callbacks
-handle response message
towards using EdDSA-ECDHE instead of ECDSA-ECDHE combined cryptosystem
(API only)
-adding cmp functions for RSA public keys and sigs
ecdh-eddsa implementation now works
-check return values
-avoid theoretical use after free
-log more diagnostics
also compare peer identity
-mystery solved, remove diagnostic
-ignore
-gettext and libtoolize updates
-use faster Eddsa key gen
-missing config
-work around stupid GCC warning (#3791)
-indentation, doxygen
set inbound flag which should fix #3719 (assuming diagnostic is correct)
at the expense of having two session objects for what is pretty much the same
address; however, as the inbound one is not really validated and might not work
due to NAT, this actually does make sense
-include local_info in hello cmp
-this does not help #3719, likely diagnosis was wrong as well, undo
-adding a few extra checks for #3615
-no need to abort iterations on reconnect if they are not yet running
with respect to #3615: we might (theoretically) get a migration stop
before we finished the peerstore lookup; in that case, we would do a
GSF_push_start() twice (once via revive migration, and once from peerstore
lookup succeeding), and then on disconnect only stop once, keeping the
migration going for a disconnected peer. That still does not explain the bug
report, as the '*cp' in the report doesn't show BADFOOD, and we'd expect a
free'd cp in this case to be used for the transmission. Still, this is a
possible issue which can be fixed by ignoring migration stop messages until
after peerstore lookup succeeded
-also use NONE instead of INBOUND for WLAN
-fix testcase
fix assertion failure when two requests were concurrently submitted if
the first one was cancelled before receiving a reply
-better error reporting, use vars
-increase timeouts, some system really need longer
-use flag to disable key generation test for libgcrypt > 1.7
-fix assertion: channel could be NULL (especially if we're dealing with a
regex)
-do not print this garbage
-log warning if channel unready
-fix regex shutdown: clean up clients before DHT disconnect
-fix bad leftover peerid to string conversion
-stupid bug
-diagnose tunnel issues
-do listen to v6
-fix configs to comply with current version
-fix policy string generation
-more verbose
-adding integration test
-improving testcase
add logging and in particular checks to make sure no '.' wildcards are
during initial transitions of the DFA
do not generate wildcard '.' transitions in policy regex, be precise
-log conclusion of exit search
-doxygen, logging
-nicer logging
-adding missing const's and a comment
-fix cmp
-fix cmp
-explicitly initialize
-do pass local_address_info flag from/to ATS also for performance monitor
-eh, typo
-logging
-logging
-logging, doxygen
-fix channel initialization with regex involvment
-fix indent
notify cadet when we are done - for flow control and to get next packet
-comments, doxygen, indentation, style fixes, working on test_gns_vpn
-use UNIXONLY
-document subsystem aliases, add gns-intercept alias for GNS DNS
interception, support gns-intercept alias for GNS DNS interception
-dns_root option is now dead
-initialize GNS DNS interception in testcase, so that curl will actually
get GNS resolution
-check for NULL
-avoid qe double-free
-fix (harmless) use of uninit 'nick'
-doxygen, indent
-doxygen, indent
-fix logging
-exit helper when either direction breaks
-space
-indent
-indent, wait for helper termination on shutdown, do not eagerly create
tunnel
-more extensive permission check before running GNS test
-indentation
-note
adding --enable-taler-wallet configure option to build a reduced version
of libgnunetutil with only the parts needed by GNU Taler wallets
-typos
-install dependent headers as well
-typo
-fix ftbfs
-simplify logic, fix double free
also set permissions correctly if user 'gnunet' does not exist
-delay HTTP request 1s to ensure VPN is really up
-fix debug logic
-disable verbose
-avoid log
fixing #3799: only unindex if DB operations succeeded previously in the
first place, and when unindexing do proper shutdown handling
make libextractor actually optional, both for GNUnet and GNUnet-taler
builds
support MHD when cross-compiling, just assume version is OK (fixes #3805)
-leak
-fix unespected status message (#3799)
-fix #3799
-typo
better fix for #3805
more fixes on #3805
-noinst-ing non-functional binaries/libraries that just exist as
examples/templates
-run gnunet-testing with proper -c flag
-slightly better logging
-improve Python3 compatibility
-simplifying Cadet KX to directly use EdDSA keys for ECDHE
turn 'force_reply' into a bitfield for future extensions
-must avoid returning resolutions during initialization
-nicer logging
-return IP in dotted decimal if reverse lookup fails
-do not use included hellos in test
-do use loopback for test
-indentation, doxygen, be a bit pickier about return values to eliminate
possible causes of #3794
-doxygen
-bogus assertion
-assertion also does not hold, messy code
-eh, remove assert
-fix use of uninit memory
-if the rest of the code uses horrible casts, one more hardly matters
-adding additional error checking logic
-run unregister even if path is NULL, do check uniqueness
-uniqueness was simply false
-check next and prev are disjoint
-avoid concurrent modification trouble when handing peer disconnect
-minor fixes
fix use of deprecated MHD symbols
-remove dead/unused fenv.h include (as suggested by Daniel Golle)
check for __GLIBC__ before mallopting
#include byteswap, if detected by configure
-ack Daniel
-fix align.m4 macro based on Daniel Golle's report
-allow 'years' in time units, indentation and doxygen fixes
-doxygen
support gnurl's curl.h being in include/gnurl/ OR include/curl/
-doxygen fixes
fix #3862
-add key
downgrade gettext requirement
also check for existence of libpq-fe header
-add missing include, remove redundant includes
downgrade to gettext 0.18.3 for more sane availability
-fix indent
-fix indent
-build system fixes, gettexting
fix #3693: if PA is NULL, ignore message, we must be in the wrong state
-fix #3870
fix #3871
patch from Nicolas Fournier to add some _dup and _cmp functions for RSA
signatures and private keys
fix #3869: outdated FSF address
-adding ecc dlog support
-style fixes
-fix non-deterministic peerstore sync failure
-change default port, as 53 won't work without root anyway
do not add our virtual interfaces to our HELLO
slightly nicer ifname cmp
-fix -h string
DEFAULTSERVICES is dead, remove remains
-indentation fixes
raise low limits, as they were they are bound to cause problems with
reverse-proxy setups
automatically clean up left-over Unix domain socket files when trying to
bind (fixes #3723)
-no binding for daemons
-no binding for daemons
adding moderately expensive instrumentation to track down invariant
violation #3846 in CADET
committing version used to produce 3846:9432
this should fix #3846
-fix gnurl/curl detection warning
-fix ARM binding
-need to fully unregister from neighbours, as otherwise assertion fails
if identical connection is added
-fix
-downgrade version requirement further
-remove stupid assertion
-do not use header of LE <1.0 by also checking that the binary version
was acceptable
-adding missing 'const' decls, fixing indentation / doxygen a bit
minor modifications to reduce warnings
work around bogus cc warning -- for some C compilers...
-doxygen, indentation
-do not run test without sqlite
fix #3904, also simplify and clean up gnunet-publish shutdown logic in
general
-cleaning up test logic, modernizing calls, improving test documentation
-remove log
-use simpler, modern scheduler API
-use simpler, modern scheduler API
-use simpler, modern scheduler API
-fix indentation
-use simpler, modern scheduler API
eliminate dead GNUNET_SCHEDULER_add_continuation, rename
GNUNET_SCHEDULER_add_continuation_with_priority to
GNUNET_SCHEDULER_add_with_reason_and_priority for consistency
-bump year
-agreed
-ignore
-reduce autostarting for tests, reduce timeouts
extend datastore performance tests to cover scenario described in #3903
-- but no issue discovered
-report id of peer that is missbehaving
-log more details about failing addr
-improve indentation, reduce duplication of PIDs in core's neighbour map
mesh is long dead
-fix leak
-more correct logic
-indent
adding some assertions to ensure we only start one gnunet-publish
tolerate spurious SIGCHLD in gnunet-auto-share
applying patch from #3950 by jah
support RFC4043 and RFC 4398 mnemonics in conversion
fix double-free
-add benchmark for RSA crypto
starting ECC-variant, but for now just a straight copy
-dead
-redundant
adding bin_to_point and point_to_bin functions for GNUNET_CRYPTO_ecc API
add GNUNET_CRYPTO_ecc_pmul_mpi
-clarify
-towards ECC variant of SP
-fix computation
-test skeleton
-test skeleton -- disable for now
add testcase for SP logic
-simplify
-simplify
-fixing ECC calculation
-more data
-fix ftbfs
-fix
-fix
-also bench hash on small inputs
add ecc dlog benchmark
report skipped tests, doxygen fixes
stop crashing if sshd is not running on loopback
-properly verify size, fixes #3991
find network interfaces also using 'ip' command (#3911)
-bump year
improving comments and coding style for libgnunethello library
-improving comments and coding style for libgnunethello library
fix #3986
adding DEBUG logic for #3863
-adding more logging to diagnose transport transmission delays
major bugfix: is_ready must only be set to GNUNET_NO if we actually did
transmit something to transport, otherwise we cannot expect to get a SEND_OK
back
-fix NPE
-do not leave PPT timeout tasks behind
stylistic fixes
-removing spectacular dead code
enable clean up of NAT STUN state, prevents test_plugin_udp from
segfaulting
-fix testcase to handle SUCCESS status reply, only recv UDP after STUN
Request is out, etc.
-ensure result is +- 1
-simplify
-typo
add additional diagnostics to help localize #3971 invariant violation
-no printf
-newline, include error code
-more meaningful log msg
determine network scope for ATS even if we do not yet have a session and
only have an address
add capability to log per-service resource consumption via ARM
export compression/decompression functions
-fix template
add logic for strata compression
-fix npe
-fix FTBFS
-fix ptr ar
-make code nicer
-fix
-ignore
-make time API more intutitive
improve doxygen
add constant for the delay after which we log warnings
-fix indent
use constant for overdue warnings
-signed cmp, otherwise -1 will not trigger
redefining core message timeout semantics and core-core-API messaging to
address (hypothetical?) stalls which may explain #3863. This kind-of breaks
the API in that the timeout now has a different semantic -- it's an advisory as
to when the app would like the message transmitted, the transmission will no
longer actually time out. However, that was the documented semantics before,
just not the actual behavior. Most applications didn't rely on it, and tests
still pass, so I didn't break too much...
-NULL builder on disconnect
trying to fix #4003
-fix destroy_ai
-ignore 2nd FRIEND hello when patching
-update DST hello
-use INFO, not WARN
-fix doxygen warnings
-bump warn threshold
-fix use after free
-ignores, indent
-do not export set_incoming_quota past file, it is not used beyond the
scope
-add option to ignore SetQuotaMessage (but still generate them), and if
the option is not set, really do use the quotas consistently
-be harsher about transport service sending messages out of line, also at
the same time check in transport service to only transmit messages if in the
right state
add logging to watch for transmission delays within transport service
fix misscalculation of per-session flow delays and apply flow delays
properly per message instead of per session
log plugin involved in slow transmissions
combing through transport ATS logic, documenting and cleaning code
-dead function in header
-typo
-mirror check on disconnect
-doxygen
-note dead field
rename 'struct Session' to 'struct GNUNET_ATS_Session' to satisfy naming
conventions
add 'strength' argument to address suggestion request for ATS
-change topology to use ATS connectivity API instead of deprecated
transport try_connect
-track hello offer handle while it is active
-doxygen
use ATS connectivity suggestion instead of deprecated
GNUNET_TRANSPORT_try_connect
-stop active offer before starting another one
-cosmetics, and preparations for try_connect change
-fix NPE
-only schedule select once per fragmented message
-use LOG instead of GNUNET_log
-fix reporting of IP address
-more logging, avoid duplicate re-scheduling
-do not suggest twice
-use UINT32_MAX to mean disconnect, for real
-we use 4 chars, not 8 for pIDs
-smoother calculation of flow delays, base it on the current message not
the previous one
-log expected transmission time on enqueue, together with queue size
-log ACK delay vs. last frag transmit
rename all_known_peers to all_connected_peers
transition from TRANSPORT_try_connect to modern ATS connectivity API
-install entry in DHT
-handle NULL for session (i.e. already gone/down)
-fix log call
move -C option from gnunet-transport to gnunet-ats; also, use
blacklisting API to implement gnunet-transport -D
-remove deprecated try_disconnect API
-fix
-fix
-indent
-bogus log
-simplify log, eliminates bogus warning and we didn't need this info
anyway
-adding Florian's IBF sim code
-randomize ps each time
-fix indentation, fix bug that caused us to try to merge requests twice
check against all pending requests with same hash, not just first; this
is a true multihashmap
-indent
-reduce calls in get_latest()
-use hash in prd query to reduce memory consumption of CP request_map
-doxygen, simplify
potential fix for #4024
-another case
-better error reporting
-check transport-side for connect-to-self
-truncate on write
-truncate HELLO
never pass addresses for ourselves to ATS
-dox
-indent
-dox
-ignore PING/PONG from self
-fix
-fix udp transmission, avoid scheduling when unready
-document quota, use higher quota for test
-simplify
-indentation, doxygen, etc.
-extra logging for #4025
prepare testbed for #3675 by having an option to establish connections to
ATS
-be less verbose
stop using deprecated TRANSPORT_try_connect() API, use modern ATS
connectivity suggestions instead
-logging, indenting
-doxygen
-indentation
-pot-update
-ignore
-fix SEGV
log rekeying operation, do not log distracting garbage for #4019
-fix #4025
-fix FTBFS
-dead
-fix doxygen
distinguish PING failure after rekey from other failures (#4019) when
logging
-do not segv on controller crash
-doxygen, indent
fixing #4035
-stop ignoring the option; #4009 should be fixed by other changes
-preparations for replacement of try_connect call
replacing deprecated transport_try_connect call with ATS call
preparations for replacing try_connect from gnunet-transport-profiler.
replacing try_connect in gnunet-transport-profiler.
-code cleanup
-fix ftbfs
-doxfix
-do check
r36622 was supposed to set to YES, not NO
-typo-doxygen
-fix compiler warning
-properly say if test was skipped
-dead:
-removing last uses of TRANSPORT_try_connect API
removing definitively dead try_connect-related logic (#3675)
removing try_connect-related logic that really should no longer be needed
(#3675)
-some code cleanup (indentation, redundant branches, etc.)
-more DCE after try_connect refactoring
-indenting
-update and correct man page (#4056)
fix crash #4053
fix #4054
more PATH_MAX issues (#4054)
typo
-fixing memory leaks and shutdown issues related to NAT
-fix indent
-removing bogus empty lines
preparations for fixing #4065
new French translation
add crc8
-fix (C) notices
force 1 to be 64-bit integer, due to shift by up to 63 bits
-fix double-free
-fix, this option is in gnunet-fs only
-fix #4151
-clarify man page
fix crash bug with -n
creating libgnunetpq library
-fix
make some more use of libgnunetpq
-document -d option
-env is dead
-bugfix, adding comments, fixing indent
-fix build issue
-experimental is by dir, not only for malicious op
lock solver during shutdown to terminate faster
-normal during shutdown, reduce log level
adding library for basic JSON conversions
-fix build issues
-reindent
fixing symbol naming and coding convention issues
extend GNUNET_OS-API to allow re-use of os_installation logic for
programs with different libs, paths, binaries and environment variables
-null out ptrs on init
add function for string parameters
add logic to parse JSON uploads
actually add new source...
-fix indent
simplify PQ API by removing NULL as a possible acceptable value
-fix leak
expand GNUNET_OS_ProjectData API to also enable de-duplcation of logic
for --help
expand GNUNET_OS_ProjectData API to also enable de-duplcation of logic
for --help
-fix resolver logic bug
fix NPE
fixfix
add copyright headers, fix includes
killing untested, unmaintained, unused sensor/sensordashboard code
fix ftbfs
small API change: do no longer pass rarely needed
GNUNET_SCHEDULER_TaskContext to all scheduler tasks; instead, allow the
relatively few tasks that need it to obtain the context via
GNUNET_SCHEDULER_get_task_context()
-fix indent
-re-running pogen
add option to set configuration file to be used by default by
libgnunetutil routines
support overriding GNUNET_DEFAULT_USER_CONFIG_FILE
fix ftbfs
adding libgnunetcurl
-use const and free json where it was allocated
removing dead code
remove dead/ugly code
remove dead/ugly code
implementing new scheduler shutdown semantics
API update to fix #4479
add implementation of scheduler integration of curl context
fix API design issue
fix #4478
add support for string results from PQ
detect format string bugs
add -w option to gnunet-config
fixing compiler warnings
fix ftbfs
fix compiler warnings
fix compiler warnings
fix compiler warnings
use 'static' where applicable
fix compiler warnings
fix #4493
rps.conf is generated from rps.conf.in
only run integration tests if libgnurl/libcurl was available
starting libgnunetmy
forgot file
rename.sh GNUNET_CRYPTO_rsa_BlindingKey to GNUNET_CRYPTO_RsaBlindingKey
following naming conventions
extending MY api to cover result extraction
define GNUNET_JSON_from_data_auto
fixing #4483: optimize blinding key storage/transmission
fix #4536
-indent fixes
sketch envisioned API
fix #4545: create directory for log file if it does not exist
only run tests if they are enabled
fix #4546
Fix for #4553
invert option order for tg
-indent
refactoring my API
-fix misc issues
-fix & vs &&
check for existing task
fix memroy leak
fix indentation, bad stack allocation of buf
partial refactoring, will cause FTBFS, to be completed ASAP
convert conversation_api_call.c
adapt peerstore API to new MQ API
-fixing part of Martin's FTBFS
fix RPS to match new MQ API -- and to check message format
fix consensus to match new MQ API -- and correct CONCLUDE_DONE to be
fixed-size
update secretsharing to new MQ API
-indentation and comments
redefine GNUNET_MQ_queue_for_connection_client to capture client handle
-fix typo
refactoring core API to use new MQ lib
use new connecT API
avoid changing API for now
-fix the fix
-fix gnunet-revocation
-assert on task
-use new connecT API
-incorrect comment
proper shutdown
avoid assertion on NULL mq
avoid deprecated API, use new connecT API
-fix FTBFS
-force c99
use c99
-fix off-by-1
convering nse_api.c to new MQ API
-tolerate init CB being NULL from the beginning
convering revocation_api.c to new MQ API
-indenting and doxygen
avoid excessive logging in config parsing
fix ATS testcase
-indentation fixes
-fix ftbfs
-fix testcase to manage with new shutdown semantics
towards fixing mysql plugin
update statistics API to use new MQ API style, also get rid of timeout
argument
convert namecache to new MQ API
-dce
rework identity API to use new MQ API
convert monitor API to use MQ
-fix statistics termination issue
refactoring ARM api to use new MQ
adapt to new MYSQL API
convert DHT API to new MQ API
convert perf API to new MQ API
convert perf API to new MQ API
migrate first half of regex API to MQ lib
-fix ftbfs
-fix format string type issue
converting core monitor to MQ
-try to fix stats issue
do not crash for #4588
use new MQ API in regex_api_search
update peerinfo API to use MQ
fixing compiler warnings...
-fix ftbfs
converting peerinfo-notify to new MQ ApI
converting peerinfo-notify to new MQ ApI
move testbed logger to its own directory
adapting testbed-logger to MQ API
migrate nat_test to MQ
-doxygen
migrate nat_auto to MQ
bump ARM lib version
fix over-allocation in datastore API
-handle NULL results
refactoring datastore API to use MQ API, also fixing misc. bugs in new
mysql backend
-housekeeping
-housekeeping
-build my
-make log level less severe
-sane log level
-fix #4582
convert namestore_api to MQ
convert namestore_api_monitor to MQ
-minor refactoring, add missing check
convert dns api to MQ
-fix bad check
-cosmetics
add GNUNET_MQ_send_copy
refactoring gns_api to use MQ API
-doxygen
first steps towards MQ
-doxygen
remove global variable for barriers, move into controller
-fix leak
-missing newline
-fix arg order
fix potential NPE on shutdown
-fix queueing issue
add function to obtain mq length
-misc fixes to DHT tests
-fix ftbfs
-fix testcase
-do not crash if disable_try_connect is set
-ignore
handle failures more gracefully in test
-handle disconnect case more nicely
add timeouts to test
add timeouts to test
move functions unrelated to TESTBED_BarrierWaitHandle out of
testbed_api_barriers.c
move functions unrelated to TESTBED_BarrierWaitHandle out of
testbed_api_barriers.c
-simplify barriers logic, make sure cb is called on all errors
avoid NPE
convert testbed_api_barriers to new MQ API
convert testbed_api to MQ (not perfect, but working)
-use more of MQ API
-doxygen, error messages
-convert resolver_api to MQ
-only trigger check config if we actually need it
-do not start unnecessary services
-ensure stats queues do not grow too big
-add extra checks to catch shutdown bug
-convert backoff test to new MQ API
convert fs_list_indexed to MQ API
-fix format string
hacky fix for cadet issues
hacky fix for 4588
-disconnects DO happen
-def before use
allow empty/NULL context message
-doxygen
-fix ftbfs if extractor.h present but libextractor.so missing
-fix ftbfs if extractor.h present but libextractor.so missing
rewrite scalarproduct_api to use MQ
-doxygen
-indicate skip if tests are not used
avoid ARM warning
-reindent
-fix testcase shutdown
-handle namestore disconnect properly
-handle namestore disconnect properly
-fix use after free on shutdown
convert unindex to MQ
-just indenting fixes
convert download API to MQ
convert search to MQ
convert fs publish to MQ
report skip if DB does not allow table creation
update tests to use new MQ API
update tests to use new MQ API
update remaining tests to use new MQ API
transitioning VPN API to MQ
indentation, moving API towards MQ-style
-reworking DV to use new MQ API
fix type -- caught by new MQ logic
converting blacklist API to MQ
-fix misc issues
-fix build issues
-separate mysql plugin build from sqlite, FTBFS so not active for now
fix dlog API for mteich
misc fixes
misc minor fixes
-typo
converting GNUNET_TRANSPORT_monitor_peers implementation to MQ
removing GNUNET_TRANSPORT_monitor_validation-API
converting monitor plugin functionality to MQ
converting GNUNET_TRANSPORT_address_to_string to MQ
remove dead monitor validation test
-skeletons for transport-ng
-be more careful, ppl, please
towards factoring out HELLO handling from transport_api.c
add GNUNET_MQ_set_handlers_closure API
-draft of next version of transport-core API
migrate transport_core API to MQ
-update ignore files
-fix shifting issues
-init cfg
-indent
-lower log level
-fix loglevel
-avoid calling memcpy() with NULL argument, even if len is 0
-fix fix
fix sanitization issues
fix packing issue
get transport_api_core to compile
-ignore meaningless timeout status
-ignore meaningless timeout status
-make handle static
-cancel reconnect task
-properly cancel offer_hello handlers
-fix cancellations in test
-fix 0-termination
-indenting
th is dead in this test
-indenting
make namestore API less brittle/sublte to use
pack network structs
fix comment
-start to refactor testcases for sanity
-towards nicer transport-testing lib
-adapt disconnect test to new testing lib
more minor fixes
remove redundant test
remove redundant test
clean up switch tests
fold bidirectional test into main test_ats_api
-use new testing API in montior_peers test
modify manipulation_send_tcp test to use new transport-testing lib
removing dead test
modify manipulation_recv_tcp test to use new transport-testing lib
modify manipulation_cfg test to use new transport-testing lib
adapt test_transport_api_reliability to new testing API
more transport testing hacking
-notes on issues
make more use of testing lib
towards having sending in transport-testing API (not yet flexible enough)
towards having sending in transport-testing API (not yet flexible enough)
-skip mysql tests properly if mysql is not configured correctly
generalize send API for transport tests
migrate more tests to new transmission API
migrate more tests to new transmission API
migrate more tests to new transmission API
migrate more tests to new transmission API
migrate more tests to new transmission API
more refactoring of tests for new send API
fix http server shutdown sequence
-fix http fix
convert last test to new send api
-ignore
-ignore
switching transport tests to new MQ-based transport API
-preparations
-preparatory steps for transport API change
-ensure clean DNS shutdown, fix DNS-STUN termination issues
-unregister nat on error, indentation, DCE
fix testcase shutdown
fix overflow/underflow handling in tracker to properly handle large
bandwidths
change time multiplication/division API to long long to avoid accidental
range reduction
-avoid overflow in calculation for very large values
-indentation, logging, style issues
-converting CORE service to new transport MQ API
-unset flag after transmission
-dead code elimination
migrate gnunet-transport to new API
migrate gnunet-transport-profiler to new API
convert tests to new transport API
-ignore
-dead code elimination
-indentation
starting point for MQ-core API
-migrating ATS tests to new transport (and core) API; core tests disabled
as new core API doesn't exist yet
-convert to new transport API
-convert to new transport API
-dce/fix ftbfs
-finally able to remove old transport API transmission logic
implementing new TRANSPORT_hello_get API
use different enum for HELLO address classes, deploy for DHT
conver to new hello_get API
converting core tests to hello_get API
convert to new hello_get API
convert to new hello_get API
converting transport tests to hello_get API
remove now obsolete get_hello-API
-bump year'
draft of new core API MQ-based implementation
enable setting per-envelope or per-queue transmission options with MQ API
starting to convert core tests to new MQ API
adapting next test to new MQ API
adapting next test to new MQ API
removing now unnecessary test
minor test cleanup
convert another test to new API
re-enable core tests now that CORE_connecT API is implemented
-convert NSE to new core MQ API
convering more services to new core MQ API
converting FS to new MQ-based core API
-ignore
-migrate to new core MQ API
-adapting xdht and wdht to new core MQ API
-fix ftbfs
-somehow, somebody changed the key...
-fix warning
fix #4609
-finish dv conversion to new core MQ API
-migrate testbed to new MQ API
-fix core mq cancel operation, need to use discard, not send_cancel as
env was never send
-convert quota test to MQ API
-fix build
-update PO
eliminate constantly hashing PIDs by storing and caching the result
-proposal for new service API
use non-constant ports for CADET with scalarproduct
-fix more shift errors
-convert vpn/exit/pt to use new CADET ports
-note simplification possibility
-do not cancel on disconnect
-change log level to DEBUG
-converting API to use new CADET ports
-converting API to use new CADET ports
-disable nse pow in FS tests
-indeed, fixed
-fix bug introduced when STAR topology was added, STAR must only be once
in the topology_strings[]-array
-use new_array where appropriate, fix indentation
fix star topology generation
-indentation, doxygen
-slightly better FS port conversion, more to be done _much_ later
-slightly better port setup, for RPS I don't quite see a better way yet
-missing include/ changes from last patches
-fix testcase shutdown
-fix testcases, ensure shutdown works properly
fix use-after-free issue
migrating set to new CADET port API - tests now fail due to CADET issues
-fix ftbfs
-fix ftbfs
-fix shutdown issue
-fix NPE
-fix NPE
fix exit CADET port binding for services, remove redundant service/port
descriptor transmission from packets
fix #4620
fix #4620 - return value part
-remove dead code
-fix unaligned access
-support hours/hour/minute as relative time units
-nicer indentation
-fix indent
ensure binding multiple ports to the same service descriptor works
(again) with the correct mapping of ports
-fix cov warning
fix consensus memory leaks
fix double free, bad use of strtok
-fix possible NPE
putting CADET on a diet: proper alignment for AX KX, kill obsolete OTR
legacy code
-fix leak
-fixing idiocy
fix memory leak and unnecessary allocations
fix missing 0-termination
-fix use after free
-fix memory leak
fix MAX_PAYLOAD definition, thereby fixing wrong type for GNUNET_log
-missing NOT
-fix leak
-proper checks for inbound messages being well-formed
-removing pointless NULL cmp
-document new API
-document new API
skeleton for service reimplementation
-starting with service_new implementation
-starting with service_new implementation data structures
-starting with service_new implementation data structures
-starting with service_new implementation data structures
-starting with service_new main logic
-starting with service_new main logic
-starting with service_new accept logic
-starting with new service MQ logic
-towards a new MST for the new service
-recover lost service_main from SVN 37845
-misc fixes to test:
fix and report version check
gns needs dht, and gns does not run as user gnunet
fix dht kill issue
-fix bad indent
kill dead code
realigning CADET messages and bumping message types to avoid interactions
with incompatible old versions complicating diagnostics
fix #4653
getting service_new.c to compile, albeit it is not yet complete
ack florian
more work on new MST and service logic
adapting test_service to new service API, fixing misc issues found by test
-modifying tests to use new service MQ API, implementing more of service
MQ API
misc minor fixes for new service MQ API, implementing resolver using new
service API
towards adding mq destruction notification
-implementing server-independent NC logic
header for new NC logic
converting nse to new service API
convert statistics service to new service MQ API
converting nse to new service api
converting scalarproduct alice-ecc to new service API
converting scalarproduct ecc-bob to new service api
-fix minor issues
convert non-ecc scalarproducts to new service API
converting core service to new service API
init resolver in new service api
migrating transport service to new MQ API
-make test more robust
migrating more services to new service API
-fix queue length counting bug
-fix use after free
-fix mysql fix
converting datastore to new MQ API
handle case of missing handler properly
migrate peerstore to new service MQ API
convert peerinfo to new service MQ API
fix #4672
use the correct type for dummy messages
migrate ARM to new service API
-must not expand BINARY as filename
-fix compiler warning
fix NPE
want himBHexpansion, but not path-prefixing
theoretically, file names might be longer than 1024 bytes, use MAX_PATH
minor stylistic issues found while hunting #4619
fix #4680
-do not crash with NPE if bind fails
refuse to run if binds fail
-do not log EPIPE
refactor DHT for new service API
porting xdht to new service API, major code de-duplication effort
-migrating wdht to new service API
init connect map on connect
convert mockup service to new service API
do not start unnecessary services while testing DHT
set flag to know that client did call init, thereby avoiding assertion
failure on code that expected initialized client
fix client init handling
fixes relating to intercepting DNS queries over IPv6
-cleaning up implementation of typemap signalling
-fix include
-fix leak
fix #4663
fix #4383
porting DNS to new service API
-eliminate dead code
migrating fs to new service API
fix #4619
fix #4703
use new MST API in helper
sketch new service start/stop API as needed for testbed
do not try to intercept link local DNS traffic, as we cannot properly
re-inject it intot the kernel's IP stack
skip test if link-local IPv6 DNS server is used
extending process API to allow obtaining status code in blocking fashion
avoid triggering dropped message 333 warning
migrating testbed to new service API
migrate VPN to new service API
proper bail out if plugin loading fails
use gnunetcheck for the DB
check rvalue for mysql_prepare
make sure testcase stays in bounds
fix looping over results
logging
deprecate USE_SSL for USE_TLS, rename in code
converting conversation to new service API
converting dv to new service API
-work around MHD_USE_TLS not being in old MHD versions
minor indentation issues
new simplified, but incomplete, client logic
revert to MHD_USE_SSL for now
fix build for Taler wallet
use new MQ in_flight API to support cancel in new service/client
implementations
misc fixes to new client impl
-fix misc minor bugs
improve test shutdown and indentation
activating client_new implementation, seems to mostly work fine, or
better than the old one
test needs larger timeout to pass
convert namecache to new service ApI
migrate template to new service API
eliminate last residue of old service api in ARM
eliminate last residue of old service api in transport
new service API will no longer export this function
simplify gnunet-nse to avoid deprecated CLIENT_service_test
simplify gnunet-namestore / namestore_rest_plugin to avoid deprecated
CLIENT_service_test
avoid using dying type for function
avoid use of CLIENT_service_test which is dying
move to new client API: remove old client API
design for new NAT service API
define message types for new NAT service
start defining IPC messages for NAT
specing out the NAT service IPC messages'
new NAT lib client api skeleton
put logging into place to diagnose #4737
working towards new NAT client library implementation
more work on new nat library
more work on new nat library
towards implementing autoconfig part of NAT lib
finish (untested) new NAT client library
-missing commit
-missing commit
add template for NAT service
-more work towards NAT service
get rid of ChangeLog generation from svn, start new ChangeLog for Git
Christian Rupp (3):
<none>
<none>
Commit: Christian Rupp Descripton: File for constants for the new sat
protocol
Christophe Genevey Metat (56):
result for libgnunetmy
GNUNET extract result libgnunetmy
libgnunetmy query helper
libgnunetmy begin result helper
fix result and query helper
test case for mysql
fix warning compilation libgnunetmy
function result helper rewritten
begin to rewritten test case
create table for test case fixed
fix select request
fixing query insert
fixing insert query
start to written extract_result
start to fix extract result
continue to fix extract result
start to fix extract vsize
start to fix extract vsize
written function cleanup
fix memory leak
finish to fix memory leak
finish to fix memory leak
lcov test function
plugin datastore mysql
plugin datastore mysql
plugin datastore mysql
plugin datastore mysql
plugin datastore mysql
continue to fix execute select
continue to fix execute select
fix plugin datastore mysql
fix plugin datastore mysql
adapt gnunet-regex-simulation-profiler to libgnunetmy
adapt gnunet-regex-simulation-profiler to libgnunetmy
add function conv param string
add function conv param string
add function conv param string
add function conv param string
delete mysql function
psycstore mysql plugin
fix leak memory
plugin psycstore_mysql
fix Makefile
start to fix psycstore mysql
start to fix PSYCstore mysql
start to fix psycstore mysql
add configuration file
start to fix psycstore mysql
fix mysql_prepare function
fix mysql_prepare function
start to fix fragment_get function
fix fragment_store function
fix fragment_get function
fix fragment_row function
fix message_add_flags function
fix fragment_select function
Claudiu Olteanu (22):
Adding a helper file for bluetooth plugin :
gnunet-helper-transport-bluetooth.c
Implementing SDP protocol
Adding a copy of wlan plugin used for bluetooth plugin
Integrating bluetooth plugin
Integrating bluetooth in the configuration file
Fixing dependencies for bluetooth plugin
Fixed some bugs
Fixing the function called to convert a string address to a binary address
skiping the broadcast messages
Implementing broadcast functionality; Resolving the 'security block'
errors; Handling select and send errors
Extending the testcases to use bluetooth
Fixing the comments
Fixing the comments
Adding W32 functionality
Undefining "interface" for W32.
defying NS_BTH
solving some binding errors for W32
Removing the bluetooth tests on W32
cleaning up the code
Removing all the tests for the bluetooth transport plugin.
Removing some log messages
Fixed the overflow problems reported in issue 3350
Daniel Golle (25):
namecache: fix testcase config for namecache-plugin-flat
namecache: fix NULL-pointer dereference in namecache-flat
pq: fix errornous include
identity-provider: install into lib instead of plugins
gnunet_json_lib: include local instead of installed gnunet_util_lib.h
dns: add support for skipping the routing setup
exit: fully skip routing setup if no interface is specified
exit: update comment for changed helper parameter
curl: fix building with gnurl/curl.h
curl_lib: fix building with gnurl/curl.h
curl_lib: fix gnunet_util_lib.h include path
curl: fix linking against libgnurl/libcurl
transport_tcp: make use of GNUNET_HELLO_address_check_option
gnunet-service-xdht_neighbours: fix typo
psycstore: add postgres database plugin
psycstore: fix puckish copy&paste in postgres plugin
psycstore: add postgres section to test configuration
psycstore: sync postgres plugin with mysql, fix indention
gns: fix format string
psycstore: postgres: remove size modifier from BYTEA fields
namestore: postgres: use IF NOT EXISTS to avoid error messages
datacache: postgres: use IF NOT EXISTS to avoid error messages
datastore: postgres: use IF NOT EXISTS to avoid error messages
psycstore: postgres: introduce LENGTH check for pub_key field
psycstore: postgres: add LENGTH checks for signature and purpose
David Barksdale (130):
Remove use of log2(). FreeBSD doesn't have it and why use floating point
when you don't have to?
FreeBSD and OpenBSD don't have fdatasync.
FreeBSD has sockaddr_in.sin_len, and sockaddr_in6.sin6_len.
Remove configure checks for guile and dialog libraries. I assume we also
don't need config.scm anymore either.
All the world is not /bin/bash
Terminate options arrays.
Use ntohs where appropriate.
Remove generated files, remove unneeded commands in bootstrap script, and
nuke libltdl when bootstrapping.
Typo causing compile error on FreeBSD.
Fix incompatible pointer type.
Modified configure.ac to check for the python interpreter so that we can
run test scripts with it. Also removed gnunet-daemon-migration which is a
generated file.
Move python test scripts to .py.in and sed in the correct python path.
Set the oft-forgotten sin_len.
Fix mpz_randomize to always update the "state" of rnd, and optimize the
code that assumes that mpz_randomize can return a number with more bits than
requested. This change breaks kblock key generation again.
Fix arguments to GNUNET_CRYPTO_hash call.
Off by one in mpz_randomize.
Remove redundant checks for gmp.h and iconv.h.
Need stdint.h man.
Remove the 2 padding bytes (that are currently filled with junk) at the
end of the tcp/udp/http transport addresses.
The NULL test clearly needs to be before any dereferences.
Fix NULL pointer check.
FreeBSD 7 does not have log2.
malloc.h is depricated and doesn't exist on FreeBSD 7.
Removing generated file mkinstalldirs.
Removed byte that's not actually part of the ciphertext, test still
passes.
Fix logic error in HKDF and provide test case for regressions.
Fix "# friends connected" counter.
Fixed udp_plugin_address_pretty_printer and gcrypt HMAC configure test.
Shifting more than the size of the integer is undefined.
Unused variable.
Use change_host_trust correctly.
Fix inaccurate comment.
Fix NULL dereference.
Add --with-zlib=[DIR] configure option.
Removing missing source files.
Make libidn optional.
Generalize condition for missing random/srandom.
Compile on platforms without netinet/ip.h. Furthur fixes to come.
Removed GNUNET_DISK_get_blocks_available since it's not used and it would
be hard to port to emscripten.
This should be a more portable GNUNET_DISK_file_get_identifiers.
Fix windows build, statfs is broken there.
Fixed GNUNET_CLIENT_service_test on FreeBSD.
Fix non-windows non-IPv6 build.
Make --with-included-ltdl work.
Removed GNUNET_OS_set_process_priority since it's not used.
Another fix for --with-included-ltdl.
Make building gnunet-testing lib optional.
Fix incompatability with musl libc.
Run contrib/pogen.sh as part of bootstrap.
Fix issue 3154: gnunet_conversation_service.h doesn't get installed
Improve search result presentation for directories.
Load the correct ATS plugin when MODE is unconfigured.
stats can be NULL if [statistics] DISABLE=YES
Use configured location of libgnurl.
Fix simple copy-paste error.
Typo in comment
Add CORS support to HTTP transport
Add CORS support to hostlist server
Fix pkgconfig/gnunetsocial.pc.in
Support a wider range of http transport clients.
Missed mesh->cadet rename in fs.conf.in
Fix hostlist server CORS headers
Add HTTP transport tests with XHR client emulation
Simplify HTTP server session cleanup even more
Fix transport blacklist tests
Don't break, this is expected from XHR clients
Use MHD_USE_SUSPEND_RESUME, fixes #3460
Require libmicrohttpd >= 0.9.32
Fix hostlist server CORS headers
Further fixes for issue #3544
Add gnunet.io hostlist server
Fix gnunet-ecc vanity gen
Implement asynchronous peerstore plugin API
Fix for systems with NULL defined as long
Fix 'STRINGS_data_to_string size check and comment
Remove superflous heap allocation
Remove work-around for bad NULL definition
Add a test to reproduce issue #3554
Check return value of GNUNET_STATISTICS_get
Workaround emscripten bug in returning int64_t
Comment on ugly API design choice
Use configured stability factor, not constant
Use configured proportionality factor
Don't timeout XHR clients so quickly.
Convert datastore plugin API to asynchronous
We require 0.9.42 for MHD_HTTP_NOT_ACCEPTABLE
dos2unix
Fix installation of sensors when DESTDIR is used
NULL terminate core_handlers[]
Avoid trying to disconnect a neighbour twice
Change NPE into GNUNET_break (0)
Check in the right place for libpq-fe.h
fix gnurl comparison
Move libgnunetpsycutil.la to psycstore
Remove this GNUNET_htonl_signed nonsense
TUN_IFNAME is the tunnel interface, not EXIT_IFNAME
Match struct IPv4/6TcpAddress with UDP versions
Fix segfault in handle_core_disconnect
This probably fixes #3944
The empty path is invalid
Fix make -j
Fix typo in dv
Remove unused variable
Typo in gnunet-publish man page
gnunet-transport shutting down too early
Clean up gnunet_test_cosks_ssh_garbage
Update README to match configure.ac
Invalid Gnunet URI
Don't match clients without handlers or callbacks
Fix UAF in asynchronous datastore plugins
Fix file path in gnunet-identity man page
Fix typo
Comparing EXTRACTOR_MetaType to -1 is always false
Use statement exprs instead of local function
Fix comment
Restore Lenstra protection for gcrypt < 1.6.4
xmalloc already zeroes the memory
I think this is a bit more readable
Add checks to GNUNET_PEER_resolve2
Trying to track down a segfault in find_content()
Fix UAF detected by asan
Only one transmit_content can go at a time
cadet_api.c uses GNUNET_MQ_impl_current
Update po files
Don't mind me, I'm just going to put this function back together
Remove ?: operator to avoid confusion
This can't be NULL
Move type check after initialization to make compiler happy
Use GNUNET_NAT_StatusCode and GNUNET_NAT_status2string
Avoid double const
David Brodski (83):
First try of the wlan-helper
Makefile.am fixed, but to compile libpcap is still needed.
Error Message for libpcap was wrong
little mistake
Some changes and new files. I will not have internet till 7. September so
if there are errors just revert the changes.
removed comments with errors in doxygen
Queues to send data should work now, fragmentation code is not finished
Added wlan data protocol defs
Sending of data over wlan added; fragmentation added
Retransmit and timed retransmit added cleanup at timeout added disconnect
added wlan header changed to specs of ieee
changed to better random number generator
init of some variable fixed
wlan: seperate fragment and session queue wlan: schedule for timeouts
used dll: insert at tail dll: insert before element
wlan receive implemented some doxigen, some cleanup of data structs
fragments are now put together to form a message
first part of the dummy for wlan plugin tests.
Dummy for testing
Keys
Errors fixed
Errors fixed
tests
<none>
some doxygen entries and debug messages
wlan helper should work now for loopback, test configurations changed
accordingly
Some errors fixed, most of them where buffer-position errors.
Some errors fixed, new debug code.
two errors fixed in wlan
Memory and assertion errors fixed
close helper at eof
Fixed many errors, send and receive with acks works now
Spelling errors corrected and some minor fixes
messages are split correctly, test works now
More then one in-going and out-going packet possible now, bitfields for
fragments implemented
Fixed warning and mac to string function
doxigen and order of mac fields
an error and some changes for the hardware
First test of wlan driver, sends beacon every 2 seconds
Some tests for wlan
CodeSonar errors fixed
Some fixes
reliability test and fragmentation are working now
Added GNUNET_CONTAINER_node_get_cost
Big cleanup and heap used for fragments send and with the timeouts
changed name to fit better
changed name to fit better
Cleanup and working version with hardware
More cleanup, some errors, message in timeout implemented
shutdown fixes
Fixes compile error
Removed the 2 seconds wait
Some bugfixes, udp packets can now be send over wlan, unrel test added,
clean up
Fixed makefile to make dummy helper and unrel test
Small fixes and detection (backlog) of double received messages
one wlan card is now usable with more than one peer, wlan header format
changed so be careful and do not mix protocol versions. Some cleanup and
formating.
Removed MTU assert, added assert for strange error with scheduler
Changed test to reflect smaller mtu
Changed wlan plugin, now uses fragmentation lib
Cleanup of log messages and timeouts implemented
Big cleanup, less source files
Doxigen fixes
msg dump on error if needed
Cleanup
<none>
Error in test for filename
Fixed wlan beacon, added llc protocol for better compatibility.
Test comments
Some Fixes
Cleanup, fixed pretty_printer, fixed session not created when beacon
received
Bugfixes and statistics added
Some asserts
more statistics, gnunet_constants used
Error message handling changed
bugfix
wireshark lua file, at the moment only for wlan
Bugfix where no data was send. Added restart of helper process when not
working anymore.
Bugfix
More protocols implemented
Some fixes with scheduler
added send test
added send test
Documantation and small fixes.
svn:ignore for eclipse folders and gnunet_wlan_sender
Made it impossible to start/stop the helper more then ones, should fix
#1873
Fabian Oehlmann (80):
ats_ril: started ats reinforcement learning solver
added GAS_ril_addresses_preference_feedback() handle
ats_ril: solver continued
ats_ril: solver further continued
ats_ril: - q and sarsa updates - added first actions - agent interacts
with environment - untested
ats_ril: debug messages added
ats_ril: indent corrected
ats_ril: debugged to run first tests
ats_ril: several fixes
ats_ril: uncommented unused function
ats_ril: added inc and dec actions
ats_ril: address administration changed
ats_ril: fixed trying to suggest address, when none is left
ats_ril: fix: set address only active if address request for peer active
ats_ril: fixed memory error
ats_ril: bw_changed_cb not called in get_preferred_address anymore
ats_ril: fix bug in address indexing when changing address
ats_ril: -some comments removed
ats_ril: - bug fixed in address_add() - small reward
function implementation
ats_ril: -fixed property indexing -comment cleanup
ats_ril: made compatible to plugin change
ats_ril: - removed some redundantly saved plugin environment attributes
formatting
-malloc fix for matrix W -null-pointer fix
tracking of assigned bandwidth corrected
split reward calculation
fix documentation errors
- bulk lock - epoch triggered after changes of environment
- corrected discount for continuous smdp
added adaptive step-intervals
configuration parameter read fix
fix memory leak
fix eligibility traces
added utilization
comments
test_api_conf changed
-integer underflow fix -bw_changed_cb invocation separate -benchmark info
calls
removed debug statement
better experience initialization
globalized networks state
added preferences to state vector
global/local reward ratio made adjustable
- variables renamed - no suggestion if not at least min_bw bandwidth is
theoretically available
extended state feature vector
- bugfixes - changed reward calculation
test variables
removed task_pending attribute testing reward function changes
added convergence test
debug fa
removed unused function
added file-output for plots
added simulation
fix matrix init
normalized feature vector
added egalitarian utility goal
check config parameters for validity and set default values
changed function approximation to radial basis functions
added option of softmax action selection strategy
eligibility traces fix
minor softmax eligibility traces fix
rbf network
config parameter rename
made replacing traces available
minor cleanup
keep addresses inuse
-fix extended address suggest api callback call
- state-feature fix - minor refactorings
minor action-selection improvement
- minor renamings - eligibility trace fix
several fixes and comments
reward function made differentiable
refactoring
-eligibility trace fixes
higher initialization of bandwidth
transport reliabilities checken
-revert of address_bw addition
-minor renamings -default parameters set to better values -exploration
decay added
-fix merge mistake
- better variable names - fast start - utilization in utility
- fixed warning about missing return values
Florian Dold (298):
added a command line tool for running and controlling services for testing
fix
fixed indentation
fix
changed how gnunet-testing-run-service communicates over stdin/stdout
added JAVAPORT to peerinfo config, fix in gnunet-testing-run-service.c
changed @UNIXONLY@ to @JAVAPORT@
added declarations for the consensus api
started implementing consensus api and service
fixed doxygen
dummy consensus service
deleted duplicate message id definition
fixed copying of wrong memory address
consensus api, consensus service (local), peer driver and ibf sketch
added consensus message types
implemented the invertible bloom filter
collision detection for IBF, timing for test tool
implemented the modified consensus api, started implementing p2p protocol
for consensus
- gnunet-consensus now profiling tool - work on service implementation,
not working yet
-public api
implemented value exchange, various fixes
-fixme reminder
-fix
-fix
fix
-log messages
- improved ibf decoding algorithm - strata estimator now fits in one
message - work on consensus, not quite working yet
consensus now implemented with primitive conclusion group selection
fixed consensus for two peers, added log-rounds, started implementing
freeze, multiple peers still buggy
fix for for multi-peer consensus, non-power-of-two consensus now works
fixed consensus multi-peer communication, memory leaks, various bugs
fix
fix
fix
added consensus log-round simulation, work on consensus service, still
problems with dv test case
fixed compilation error in chat
added skeleton for gnunet set
started implementing set api, draft for mq
fix
partial docs
fix
fix
api fix
started implementing union operation for set
forgot to commit new files
fixed compilation errors: undefined reference to `GNUNET_ARM_[...]'
reverted changes to makefile, was just a problem with autotools on my
system
implemented GNUNET_CRYPTO_get_host_identity
work on gnunet-set, isolated bug in stream
forgot to add c file
fix
implemented most parts of the set service
fix / documentation
fix/docs
missing config value
test cases for mq, set works
removed unnecessary malloc
- moved MQ to util - MQ support for stream - set api - starting to use
set for consensus
- conclude for SET - consensus with SET
- opaque mq structs - mq for mesh - faster hashing for IBFs - mesh
replaces stream in set - new set profiler (work in progress)
commented out calls to functions not yet implemented
mesh bug for bart
testing lib missing in mesh Makefile.am
documented corner case of dll remove
doxygen
- fixed tunnel destruction/cleanup - peers can connect before the listener
- context msg size constant added ---This line, and those below, will be
ignored--
- fixed tunnel context - moved logic out of specific operations
doxygen
doxygen
comment
fixed duplicate tunnel creation
test case for single mesh handle
- initialized all fields of mesh's tunnel destroy message
- full message struct initialization
- set service working - set profiler
fixed test case
- queue destruction
- set autostart
- fixed malloc size
-missing return
-fixed memory leak
- cancel uncommited set operation correctly
- test for correct uncommited set op destruction
- handle cyclic IBFs and SEs correctly
- pass set service disconnect to client
- set iterate declaration
- comment
- set iteration
- docs
- better debug logs, ibf tweak
-send elements to client
- extract context msg correctly
- modulo in shuffle
- return, don't break
-consensus fixes
- msg type added
- logging
strange typo in dv
- format string
- doc for peerinfo
- docs
- external iterator for multihashmap
- listener re-connects transparently - bugs
- fixed memory leak
- doxygen comment clarification
- fixed consensus for >2 peers
- test for external iterator - const added to result parameter of 'next'
- test memory leak
- correct iteraion in multihashmap - iterate over elements with client
ack in set
- show received elements in consensus profiler
- fixed profiler - less INFO, more DEBUG
- rename gnunet-consensus to gnunet-consensus-profiler ---This line, and
those below, will be ignored--
- proper shutdown sequence for set makes consensus happy
- delayed requests correctly when in 'begin' round
- disable nse in the profiler config - handle ghost peers correctly in
exchange round
- IBF alpha adjusted - it's not an error to insert elements twice ---This
line, and those below, will be ignored--
- reset backoff for set listener - increase alpha
- bug in argument parsing
-dce
-dce
- profiler can be verbose
- fixed generation issues
-message not needed anymore
merged 'gnunet-testing-run-service' into 'gnunet-testing'
superflous printf
doxygen typo
consistent output with gnunet-testing-run-service
removed unnecessary printf again
- initialize set type in test case - handle 1-peer consensus correcly -
don't store peers in consensus api - updated consensus test case
- testbed port is used by java
- doxygen
use AC_LANG_SOURCE in order to satisfy AC_LANG_CONFTEST
correcly quote AC_LANG_SOURCE to silence warning
don't use the deprecated, redundant form of AM_INIT_AUTOMAKE
- replace deprecated INCLUDES with AM_CPPFLAGS
- mem leak
- initialize element type in test
MQ for CORE
- comment clarification
missing file
make identity port available for java
- doc clarification (identity manages egos, not identities)
- make transport and consensus ports java-ports
- fixed assertion due to task not testing for shutdown as a reason
- some of the missing set union functionality implemented
- secretsharing api proposal
fixed include guard
clarifications
typedef for decryption result callback
typo
separated threshold from other crypto parameters
- implemented missing set functionality - secretsharing api changes
skeleton for secretsharing
- Added the '-E' option to gnunet-ecc, which prints reference results of
ecc operations
fixed a bug where a set handle was used after free
- check for null in mq/client when disconnecting listeners
- fixed handling of incorrect channel port
- fixed use after free due to uncanceled task
- log as debug when ibf fails to decode, instead of warning
- work on secretsharing
- missing message type
- fix use after free in set union (#3178)
- fix #3185 and #3184 in set
- fixed use after free in set (#3188)
- key generation for secretsharing - gnunet-ecc -E also prints hex
- destroy channel after timeout
make sure incoming request is not destroyed twice
- profiler actually added to svn - test for signature segfault with
secretsharing - rudimentary implementation of cooperative decryption -
consensus uses absolute time instead of relative - working DKG without zero
knowledge proofs
- remove #define for elgamal_q
- removing superfluous test cast, my fault
- htons => htonl
free
missing secretsharing_common.c
- correctly adjust buffer when printing MPIs - fix confusion between
paillier and elgamal field elements
doxygen
- make mpi scan/print public - secretsharing key generation and
decryption fixed
- paillier
- remove adjust
- type error
- paillier implementation
- document string termination
- add missing crypto_paillier.c
- gnunet-secretsharing-profiler.c should not have exec flag set
- doxygen
- paillier implementation and test
paillier constants / structs from util
remove paillier_create in secretsharing
remove paillier_encrypt in secretsharing
remove paillier_decrypt in secretsharing
use paillier public key container in secretsharing
use paillier private key container in secretsharing
remove unnecessary mpi conversion
- test paillier with random plaintext
- encrypt even if no homomorphic operations are possible - parse
ciphertext container correctly
- consensus and secretsharing have both start time and deadline
- GNS port is a javaport
- delete unused message type
- fix operation cancellation in SET
- correct handling of timeouts in consensus - fixed segfault in
secretsharing profiler
- syntax error
- fixed wrong round comparison code
- don't kill clients if the set operation has already been destroyed
- fixed wrong crypto in secretsharing - added zero knowledge proofs
(except the one for fair encryption) and other verifications
- secretsharing api test for single peer - fixed wrong endianess
conversion
add missing ifdefs in header files
Rename SearchMessage from regex to RegexSearchMessage so that it is
unique in a total c-file
Rename uLong to GNUNET_uLong from crypto_crc so that it doesn't collide
with the different uLong from zconf of zlib
Add missing stub definitions of structs to the corresponding header files
Rename name_map and plugins from gns so that they are unique in a total
c-file
missing struct definition in gnunet_scalarproduct_service.h
missing ifndefs in headers
missing struct definition in header
- call operation_done on testbed operations in profiler - destroy
consensus handles in secretsharing service correctly
superfluous allocation
fix another memory leak in paillier
fix lots of memory leaks in secretsharing key generation
fix lots of memory leaks in secretsharing decryption
don't use valgrind in secretsharing profiler by default
add missing cancel implementation for MQ
fix bug in mq destroy
timeout for set tests
set parent queue to NULL before discarding envelopes when destroying a
message queue
generate proof of fair encryption
full proof of fair encryption
secretsharing cleanup
work in progress: fix set bug, implement lazy copy
missing changes to headers
add test case for 'GNUNET_SET_copy_lazy', fix bugs
fix inverted condition in set intersection
work on CONSENSUS and SET
set api changes
SET service: accurate results for symmetric mode
remove macro, fix indentation
fix undefined behavior in GNUNET_array_grow
- consensus test conf
- another fix to generation handling and lazy copying
work on consensus and set
consensus
include element type in hash
consensus work in progress
towards handling byz. faults correctly
fix contestation handling / blacklisting
set statistics
statistics in set profiler, temp. fix for salt
consensus
fix consensus logic bug
More refined evil behaviour.
Statistics log file for consensus profiler.
Make resource diagnostics output more useful.
statistics
More statistics.
Also install set profiler.
Set statistics
Proper salting for set union.
Increase maximum IBF size a bit.
Handle larger IBFs correctly.
Remove logging.
Disable salting because it's still buggy.
Make HAVE_REST depend on both mhd and jansson.
Add support for $GNUNET_BASE_CONFIG
Make ARM understand the TYPE option for services
Do $-expansion for OPTIONS of simple services
Remove MSEC from option description.
Implement early stopping.
Fix ibf key salting.
Only compile plugin when we have sqlite.
be more lenient in the set api
Output error if both gnurl and curl are missing
Do not include platform.h in headers from include/
Allow expansion of [arm]BINARY.
Allow GNUNET_BASE_CONFIG to be customized.
Missing change in include file.
Revert 37060
Remove format string abuse with sane code.
Use dollar expansion helper instead of custom code
don't free -- expand_dollar does realloc
expand dollars in BINARY for service type 'simple'
log error when timed out
adapt set test case to new scheduler semantics
log
fix compiler warnings
docs for service api
more docs for service api
setup code for new service api
setup code for new service api
make service_new (almost) compile
don't build contrib/ when compiling for GNU Taler
don't even check for postgres when building for taler, since it breaks
the build
quoting, since test breaks (under emscripten) when variable is empty
string
fix configure.ac for emscripten
Add evict functionalty to mq cancellation.
Rename MQ_evict to MQ_evacuate and make API nicer
fix comment about mq allocation
implement impl_in_flight API for MQ, replacing evacuation
fix comment
turn if into assert
oops ... inverted logic
Gabor X Toth (254):
util/server: removed unused var
configurable hello expiration
moved hello uri functions to hello lib
hello lib header fix
psyc apis
psycstore api docs
-
social api docs
psyc apis: use more doxygen syntax, added some missing docs
multicast message types and api docs
multicast: messages types; function & callback for ping/pong messages;
separate join/leave callbacks
multicast: ping/pong docs
PSYC APIs: added missing args and functions, more consistent naming
PSYC/social: header for method callbacks
PLACE record type for gns & namestore
psyc api: no state events, those are handled in social
multicast: message fragmentation
psyc(store): group generation only needed when storing the message
PSYC/social: use an Environment for setting variables / state operations
PSYC API: use master/slave/channel terminology
env: doc; spaces
PSYC: move history & state functions to the common channel API
social: use the TransmitReadyNotify callback from the connection lib for
transmissions
PSYC: historic message flag
psyc/social docs
local doxygen
social: finish callback for history lesson
PSYC/multicast: message parameters for join/part requests
untabify
multicast/psyc: use start/stop naming for master/origin
multicast/psyc: no part cb needed
multicast: group_member_remove()
multicast: message_cb and last_fragment_id for origin_start(); env lib:
indicating state reset
psyc: use relays for join decision in this api too
psycstore: state_modify()
social: home_leave()
multicast: additional args for member_start, origin_start, origin_to_all;
social: use private key struct instead of file
multicast/psyc: join cb for members/slaves too
multicast: no state_delta
social: use full pubkey as place/nym id; multicast/psyc/store: identify
members by their pubkey here too
psyc/store: added effective_since msg ID for specifying when a membership
change went into effect; psyc/social: added state_reset/clear_objects flag for
message sending functions
psyc/social: transmit/announcement flags; multicast: re-added sig. purpose
psyc/social: slave transmit & guest talk flags
psycstore: added state_hash_update() and counters_get(); psyc: async
state_get*(); multicast: membership test & replay params
social api: use ego from identity service
moved policies from multicast to psyc layer
social: home_advertise() - added arg to specify peers for the entry
multicast/psyc: transmit callbacks; social: password for home_advertise()
social: place_enter args
social: place_enter2 arg fix
social: place_look args
multicast: replay functions, args, callbacks, return codes
multicast/psyc: added admitted_since arg for join_decision(); social:
added nym arg for the method cb
multicast: added replay_end(), returning replay handle from
join_decision(); removed admitted/effective since where not needed; social:
keep_active flag instead of away
multicast/psyc/psycstore api: fix typos / type names
identity: doc fixes
psycstore service skeleton
psycstore: sqlite plugin
PSYCstore SQLite backend; API fixes/enhancements
multicast: renamed replay() to replay_response(), doc fixes
psycstore: doc fixes
multicast: doc fix
multicast: revert packed struct syntax
PSYCstore service and API implementation
psycstore tests config
colorit.conf
psycstore: fixing issues reported by static code analysis
psycstore: use GNUNET_assert()
psyc service skeleton
psyc service skeleton
psyc service skeleton
env lib
psycstore: single counters function for master & slave
psyc service: start/stop, join/part, message transmission: lib -> psyc ->
mcast; psyc API: stop/resume transmission
multicast API: stop/resume transmission
multicast: simplistic origin_to_all() code to be able to test the psyc
service
PSYCstore: result_code arg for CountersCallback
\
PSYC: master msg transmission
logread
psycstore test home
logread warning fix
log msg and comment fixes
logread: ipc message monitoring
psyc: handling messages from multicast and passing them to clients;
pause/resume fixes
psyc: ipc messages
env: added head() and shift()
psyc: ipc messages, notify callback for modifiers, tests
psyc: message handler checks
ssh key
logread: more generic message type regexp
PSYC: implement slave to master requests, tests, fixes, reorg
PSYC: in-order delivery of fragments; tests for large messages
PSYC: message types, documentation
PSYC: some code cleanup
logread: use global matching for colouring
add GNUNET_CONTAINER_heap_peek2()
psyc: in-order message delivery
multicast: send messages between client lib & service
multicast, psyc: client connections, join requests
psyc, multicast: join decision
logread-ipc colouring
psyc, multicast: fixes, comments
psyc, multicast: join decision, tests
add social service
multicast: run testing peer
undef _FORTIFY_SOURCE before defining it, in case it's already set
social: separate host/guest/place API; add missing method cb params;
improve docs
client manager: higher level client API with a transmission queue and
message handler registration
PSYC util lib: receiving/transmitting/logging PSYC messages
psyc, multicast: reorg code, use new client manager & psyc util lib
PSYC(store), Multicast: use ECDSA slave/member keys; PSYC: add
GNUNET_PSYC_message_create()
rename logread tools
Revert "- persuade social to at least compile"
Revert "- convince gcc to compile"
social: implement enter/leave/messaging; psyc: improvements and fixes
social: fix linking
social: guest_talk test; multicast: fix member_to_origin_resume()
multicast,psyc,social: remove DEBUG #defines from tests to keep `make
check` working
psyc: fix timing-related test failure
psyc,social: fix linking for tests
test_psycstore: use ecdsa keys where appropriate
test_plugin_psycstore: use ecdsa keys where appropriate
social,psycstore: fix tests linking/debug
social: fix test debug
multicast: use cadet to contact remote peers
multicast: send join decision to remote peer
psycstore: add option to perform membership test when retrieving fragment
or message
psyc: multicast callbacks for membership test and fragment/message replay
psyc: membership store
client manager: documentation
psyc, psycstore: retrieve state and history
psyc, psycstore: operation result max error size
doc fix
logread: no buffering
add GNUNET_(hton|ntoh).*_signed() functions
client_manager: add API for async operations
psyc/social: request history & state from psycstore; more documentation,
tests, cleanup
psycstore: doc formatting
psyc/store: apply state modifiers
psyc/social: get state from psycstore
social: watch modifiers
psyc: add STATE_MODIFY flag during transmit
multicast, psyc, psycstore, client_manager fixes
social: farewell
social: advertise; farewell fix
social: guest_enter_by_name(); zone_add_pkey(); join_msg fixes
multicast tests
psycstore: add fragment_limit arg for message_get
multicast: replay
multicast: doc formatting
psyc fixes
multicast: removed membership test callback as clients check that already
during join & replay
multicast: removed replay cancellation as responses are limited
multicast: remove membership test & replay cancel message types
multicast/psyc/social: message acks & scheduling
free fix
social: store/load entered places & notify clients about them
social: API changes for application connections: store/load app
subscriptions to places
multicast test: now compiles in non-C99 mode
psyc/social: local join flag; social service: leave place, save _file
social: compile fix
doxygen config
doxygen: group/module definitions (part 1)
doxygen: group/module definitions (part 2)
doc
psycutil reorg: message, env, slicer
doxygen: add documentation links
social: set/clear msg proc flags
psyc/multicast: var names
psycutil: slicer callback args
social: relay msgs; psyc: slicer api; multicast: reset fragment_offset
doc
social: save file
social api: some missing bits
social: api, doc
indent
social: relay + save doc
social: AppPlaceState doc
social test fix
social: todo
gnunet-social cli
social: app connected callback
social cli
social cli
social: fix compiler warnings
social: cli options
social cli
social api: app_disconnect callback
social: place load/save
social cli: transmit fix, sig handlers
social cli: --host-assign option
psyc/social: transmit fix
social cli docs
debug msgs
fixes
fixes
fixes
social cli: fix shutdown
multicast,psyc,social: use CORE_connecT()
multicast: switch to MQ
social fixes
op: lib for async operations
psycstore: switch to MQ
multicast: switch to MQ
psyc, social: switch to MQ
GNUNET_MQ_get_last_envelope()
multicast, psyc, social: wait till last message sent before disconnect
multicast: use CADET port hash derived from group key
social: debug msgs
psyc, social
GNUNET_MQ_get_last_envelope: return tail or current
test_psyc: log level
psycutil: check for paused state in got_ack()
multicast: close cadet port
rm client_manager
multicast: towards multipeer test
psyc,psycstore,social fixes
psycstore/mysql: check return values
rm client_manager
psycstore-mysql: remove filename config option
+
multicast,psyc,social: remove core dependencny
MQ doc
multicast: switch to SERVICE API
multicast: cleanup unused message structs
psyc: switch to SERVICE API
social: switch to SERVICE API
multicast,psyc: switch to SERVICE API
psycstore: switch to SERVICE API
psyc,psycstore,social: switch to SERVICE API
postgresql cflags
nix package definition for development
nix package definition for development
nix package definition for development
nix: debug build, ccache
psycstore: mysql stmt fix, logging for tests
log: add \n
psycstore/postgres: fix return value checks, use template1 db for tests
psycstore: fix msg handlers
mysql: statement_prepare: log query on error
psycstore: fix limit
psycstore: conf
my: fix extract_string
psycstore: fix mysql
psycstore: postgres fixes
pq: fix crash in clean_varsize_blob() after extracting NULL value
psycstore: postgres fixes
psycstore: fix postgres
psycstore
gnunet-dev.nix
gnunet-dev.nix
gnunet-dev.nix
gnunet-dev.nix: add jansson
Heikki Lindholm (25):
update e-mail
OS X is a BSD here
OS X version of MSG_NOSIGNAL
don't repeat setsockopt on every send
cosmetics
for better POSIX compliance
autoreconf should fare better, esp. where libtoolize = glibtoolize, but
needs recent version of autotools
try to take rescheduled transmit into account
remove ws
fix
fix linking
standards compliance
lowly sparc64 couldn't cope with 5 minutes
libtool to the rescue!
observed on ppc64
fix format types
fixme
clarity
fix test on darwin, freebsd
fix compile on os x
HOST_NAME_MAX doesn't exist on freebsd or darwin
better: sysconf seems to be the preferred way
static until proven otherwise
rehash since gethostname is used in resolver as well
defining stuff in headers confuses the linker
Jeff Burdges (36):
I'm unsure if I've correctly tested it in the odd build enviroment for
which it was create.
Fix of last commit for non-GNU make, I think.
Grant myself buildbot access
Patch from Daniel Golle <address@hidden> on Sun 26 Jul 2015:
Statment added in r36048 terminated by a . instead of a ;
Add support for building connection proxies.
New to notify that we're done with the handshake
Export proxy methods
Initial SOCKS5 proxy code. This will not yet work properly.
Build socks.c and call it from client.c It still won't work yet, now
debugging.
This appears to be dead code. I'm not really sure however! Can someone
please tell me if this would break any testing situation and/or revert it?
Fixed one serious bug, working on another. Still very broken.
Add test code, but it's still segfaulting.
Woot! It might actually work now, although I should tweak the test a bit
to make sure.
Add socks lines to the makefile, but do not yet activate
Fix the SOCKS proxy test and activate it.
Make SOCKS test conditional on an SSH binary
Use SSH more safely from configure
Implement a Full Domain Hash (FDH) for RSA signatures and blind signatures
Updated global symbols for FDH
Fix paramater description
Release outside of helper function
Avoid exporting mpi_print helper function
Clarify a few paramaters and names
Authors, (C), and C++ comments
Copyright years
Fix cut & paste error
Use a uniform random number mod an RSA composites for both the blinding
factor and the full domain hash.
Testcases for KDF mod n
Verify that GCD(m,n) != 1 when n is an RSA modulus
Abstract out previous GCD(m,n)=1 commit into a single function
Rework the error handling for gcd(r,n) != 1 so the Taler wallet can see
errors.
Fix perf_crypto_rsa.c after various changes
Fix a few errors in socks.c
Not reason, but the macro exists so I'll use it I guess.
Removed duplicate protection agaisnt Lenstra
Ji Lu (30):
bio testcase harness~
bio testcase harness
bio testcase harness
bio test
bio test fixed
complete Test_time.c
complete Test_time.c
complete test program
test time.c
please check if there are some bugs inside
please check if there are some bugs inside
89.1%
fixed test_bio.c
98% test_bio.c
fixed all null problems
fixed all null problems
tiny thing
FIXME
some log problem and 55bytes mem lack not fixed
<none>
<none>
<none>
<none>
pending fragmentation
pending testcase for fragementation
fixed ==
to be continued
to be continued
passed test
final version from ji lu
Julien Morvan (9):
Add AppArmor profiles
Add AppArmor abstractions(gnunet-common, gnunet-libaudio)
<none>
.unix change to .sock
<none>
<none>
<none>
<none>
libextractor
Julius Bünger (333):
before deleting PeerList
Got rid of SList
Completely got rid of SList
Got rid of compile errors and warnings
fixed error
fixed errors
Fixed errors and cleaned up
Cleaned up
Added documenting comments
now using API calls instead of own fkts
fixed some issues
minor fixes
fixed compiler warning in tests
_malloc() -> _new()
removed unnecessary uncommented code
fixed usage of get_rand_peer()
reading alpha, beta from .conf
fixed place of resizing gossip_list
using get_peer_context()
initialising gossip_list correctly
added seeding function to the api
added message type for seeding
fixed doxygen
make the test run without internet connection
added log message
fixed struct
fixed config
removed unneeded api call
moved sampler functionality in file of its own
- modified Makefile accordingly
changes to rps.h
moved part of _resize() from _sampler.c to -rps.c
now sending push,pull to at least one peer/round
changed logging + cosmetic changes
fixed compiler warning
removed unwanted string from log msgs
different fixes and cosmetic changes
got rid of touch_*()
fixed index of for-loop
checking msg sizes, cosmetic changes
check if id is already in current push/pull_list
randomised scheduling of next round, cosmetic changes
small fix in random scheduling
moved computation of request rate out of sampler
renaming
updated .po-files
implemented seeding
fixing memory leaks
restructured the sampler
removed unneeded global
fixed indentation
resizing lists implemented, fixed type error
added todo
send empty push/pull-request messages
fixed wrong deletion
corrected message size check
added space between fkt names and args
clean up channels to other peers correctly
cleaning up disconnected peers
schedule some requests for later
fixed message size
fixed integer sizes and _get_rand_peer()
simplified removal of peers from peermap
check if peer we open a channel to is live
undo last commit
removed unneeded vars
undo last commit
trying to remove get_peer_task
optical changes, small fix, draft for notification
restructured test, testing seed
updated .po - files
update .po - files
optical changes
restructured service and sampler
clean up ntfy_tmt_rdy handles on removal of peer
make sure fkt is called with right argument
keep track of whom we sent pull requests to (avoid duplicate replys)
don't read threshholds from .conf
modified TODOs
improved selection of peers for pushing
moved the client request rate estimation in fkt of its own
code cleanup
mostly taking care about max size we are allowed to send
taking care of send msg size in the api
fixed doxygen
fixed _seed() funktion
optical changes
fixed wrong cancellation
fixed size problems and removal of channel
fixing size problems
fixed wrong size check
- modified .po - files
- fixed double free problem
- trying to fix issue of not sending push/pulls to peers
- fixed error in peer selection, never grow smaller than 3
- change in POTFILES.in
- trying to fix coverity defects
- fixed resource leakages
distinct samplers for client and Brahms protocol
- change in POTFILES.in
- macro for getting/setting peer flags
- restructured in_arr() and rem_from_list()
- restructured check whether peer is live
- changed criteria for putting peer in pull list
- restructured peer removal
- don't put our own id in the algorithm
- add additional check before peer removal
- leave peer in sampler on peer disconnect
- changed behaviour on peer disconnect
- fixed behaviour of init_peer_cb()
- don't add own id to pull_list
- renamed flag
- got rid of unneeded code
- cancle own connection when peer is not in gossip list
- fix wrong check
- make sure there is an established channel for each peer in gossip list
- check for validity of peers
- fixed confusion of push and pull
- fixed slow startup of service
- fixed uninitialised memory, logic
- fixed logic
- cleaned test code
- fixed memory and coding issues
- fixed conflicting int sizes
- fixed rounding
- fixed conflicting int type
baseline for test with malicious peers
- sending act malicious message
- sending act malicious message
- several minor fixes
- get rid of unnecessary code execution
- renaming
- added malicous round execution primitive
-fixed issue with ntfy_tmt_rdy ()
-scheduling malicious do_round ()
- s/#if/#ifdef/
-optical changes
correct handling of act_malicious messages from client
-corrected type
-corrected peer_is_live() checking
-more verbose logging
-only turn portion of peers malicious
-further implemented mal peer, restructured code
-small fixes, logging
-additional assertions, fix
-malicious peer type 1 sending pushes
-mal peer type 1 pulls
-refined mal peer type 1
-malicious push handling
-mal peer handle pull request
-mal peer pull reply handling
-cleaning up
-restructured code
-unified testing
-fixed handling of client requests
-restructured handling peers from 'external sources'
-fixed types
-further refined tests
-several fixes in api
-renaming
-logging
-fixes
-fixes
-logging
-doxygen
-updated TODOs
mal type 3 (combined 1 + 2)
-restructured tests
compile test to profiler
-writing important sampler info to file
-fix sampler
-fixed Makefile.am rule for profiler
-replaced own util fkt with GNUNET_ one
-cont write internals to file
-further logging in malicious execution
-optical changes
-tried to fix removal of peers
-additional logging in sampler
-small fix
-taking also care of type mal_type 3 in api
-restructured test
-optical change + small fix
-restructured churn
-restructured retrieval of information about peer
-restructured testing
-small changes in test_rps.conf
-fixed coverity Uninitialized pointer read
-better splitted brahms sampler and modified one
-fix
-fixed naming
-restructured the code to faciliate viewing the internals
modified sampler to keep elements after deletion for evaluation
-writing the gossip list (view) to file
-tried to fix channel cleanups
-comment out annoying debug message
-changed permissions for created file
-fix coverity
restructured test
-got some {} for free
learn about peers from peerinfo
-renamed (gossip_list -> view) and converted it to peermap, fixes
-fix hanling of pull replies
-removed unnecessary data structure (pending pull list)
-logging and comments
-restructured cleaning of known peers
-fixed comparing of peer ids
-fixed wrong pointer usage
-restructured sending of push/pull request
-fixed typos
-clean logging dir before running profiler
-fixed missing link to library
-logging
-fix coverity 0003864
log missing config
-fixed wrong check
-removed unneeded peermap_put
-fixed inline docu
-fixed duplicate ntfy_tmt_rdy
-cleaned resizing of sampler
-logging
-prevent harmless warning
-fixed indexing
-fixed early destruction of channel
-removed unnecessary complex code
-split up sampler and sampler element
get perfect sample of each sampler element
-keep peers malicious
-fixed cadet usage
get indices for evaluation with diehard
-logging
-fixed emptiness check
-fixed assertion
-stop passing static peermap
-fixes
-fixes and improvements
-fix push receive
cancellation of request and according test improvements
-improved management of peer information
-improved testing
-fixes
-logging and fix
-fix malicious peer
-use hashmap instead of array
-renaming
-mark ops as done + don't start connect all clients
-test seeding many peers
-fix (create peermap), logging
-logging messages: going to send, sent, received
-rename and fix char pointer
-keep track of messages passed to mq
implemented CLI
-fix: prevent division by zero
added a custom list/peer-/hashmap structure for use in rps
-substitute array-structures push_list, pull_list with custom list
structure
added helper for handling the "view" in rps
-fixed 4076
-fix of issues in helper for "view"
-add tests for helpers "view" and "peers"
-added test of _change_len() to test of "view"
-fixed _clear() in peers "helper"
-using "view" helper in rps service
-doxygen
-fixed bug in removal of peer in "view"
-rename _peers.* -> _custommap.*
-removed obsolete functions
rps: restructured service. moved peer-related code to
gnunet-service-rps_peers
-rps: doxygen
-rps: restructure gnunet-service-rps_peers
-rps: add parentheses in macros
rps: add test for gnunet-service-rps_peers
-rps: deleted redundant function call
-rps: proper setting of "online" flag
-rps: doxygen
-rps: restructure service internals
rps: keep track of valid peers in peermap
-rps: logging
rps: store valid peer ids in file
rps: add missing rps.conf
-rps: logging
-rps: try to avoid leaking memory
-rps: fixed storing valid peers
rps: use stored peers at startup
add missing options to rps.conf.in
-rps: rename and doxygen
-rps fix: check valid -> check online
-rps: removing unused counter
-rps: check knowledge of peer
-rps: only insert online peers in the view
-rps: logging
-rps: open channel when inserting peer in view
-rps: merge duplicate functions
-rps api: actually call check_reply()
-rps service: doxygen
-rps service: correct check of peer status
-rps _peers: restructure of status check
-rps _peers: add disable option to peer storage
-rps test _peers: disable storage of valid peers
-rps api: delete leftover struct
-rps service: correct check of peer status
-rps fix: change way of online notification
-rps fix: do not ask cadet for peers
-rps peers: logging
-rps test: fix directory specification
-fix rps service: handler for check_live message
-rps service: improved handling of errors
-rps sampler: remove useless code
-rps doxygen
-rps: proper destruction of sampler elements
-rps test: try to close file in case of failure
-rps service: remove unnecessary var = NULL;
-rps service: correcting scope for variable
-rps _peers: prevent leaking of storage
-rps _peers: prevent leaking storage
-rps _custommap: more assertions (keep coverity happy?)
-rps test: add assertion (keep coverity happy?)
-rps service: add assert (make coverity happy?)
-rps service: prevent division by zero
-rps: check return values
-rps service: replace memcpy -> memmove due to overlap
rps: added test for sampler element
rps sampler_elem test: add to Makefile
-rps service: proper removal of clients at shutdown
-rps _peers: logging
try to convert rps to new service api
-rps: proper cleaning/removing of peers
-rps: fix #0004678
-fix rps service: call sampler_request_cancel () on destruction of handle
-rps api: remove superfluous call to check_repy()
-fix rps service: handle client service types correctly
-fix rps service: memcpy() -> memmove()
-fix rps service: prevent calling _request_cancel() twice
-fix rps service: _continue() instead of _drop()
-rps api: resend requests after reconnect
LRN (312):
Make gns port non-unix-only
A small fix for W32cat
W32: safer process termination
Missing file
Demote WARNING to INFO, preserve errno
Safer handling of corner-cases in w32 select
Handle peek errors more appropriately in w32 select
W32: correct handling of crazy W32 process quirks
Fix transport makefile
Really try to fix compilation on Windows
Fix arm-lsocks protocol violation on W32
Use test tgz instead of COPYING for FS tests
Fix dns2gns dependencies
DOS-pathsep-compatible filename extraction
Fix hostkey loading for old testing framework
Use Python batteries in fs_rec test
pythonize test_dht_tools
Fix w32nsp headers
Use pipe control for helper processes
Do not run install commands from makefiles directly
Fix a typo
Fix dns service dependencies
Don't run unix test on W32
Roll back r23388, fix it differently
Don't access t when t is NULL
More dont-access-t-when-it-is-NULL fixes
changes to address notification strike back
Fix fs test config
fix http client ldflags
Make poisoning configurable. Enable it on debug builds by default
Fix https client ldflags
No valgrind in HEAD, please
Remove duplicate cleanups
Fix gns dependencies
Remove non-existing files from POTFILES
Use native srcdir for #includes
Add ldflags to gnunetgns_common
W32-specific platform stuff is now pure C
Fix a typo
Missing declarations
Short timeout for starting local services
Try to fix W32 builds
Fix 0-terminator test in lockmanager
Implement non-inheritable sockets for W32
Create directory for hostkey
bratao: W32: Optimize ..._select() implementation
Fix naming, HEAD fails to compile here
Allow using libunistring for string conversion
W32: Fix get_path_from_module_filename()
Missing include
Just use unistring
Restore warnings, fix test compilation
Fix ENABLE_NLS usage
Debug for dollar-expansion
More configuration debugging
W32: optimize GNUNET_NETWORK_fdset_add
W32: Remove unnecessary 0 and -1 checks
Use strrchr() instead of legacy rindex()
twopeer test depends on testbed
twopeer test depends on dhttest library
link testbed after dhttest (depends on testbed)
Typo
Use full, 3-argument FPRINTF invocation
Use full, 3-argument invocation of FRPINTF (again)
Update gauger C binding for W32
Fix W32 gauger C bindings - quote arguments
Update w32nsp to use 'gads' instead of 'gnunet'
Use full, 3-argument fprintf invocation
Hardcode svn revision in compiled binaries
Be more specific about UDP read failures
Should be compilable without libglpk again
Don't crash when address suggestor fails
Fix configuration parsing
Unnecessary argument
Fix percentization of non-ASCII strings
Fix default hostlists
Resolve port conflict between ats and sysmon
Dependencies in consensus makefiles
Re-enable signature check warning
Replace EPV with a warning
Fix FPRINTF invocation
Fix regex dependencies
mesh_block_lib.c is gone
Don't send more than 64k from fs helper
Fix a buffer overflow
Fix a memory leak in MD serializer
Discard large metadata items first
W32: Don't leak stdin/stdout when spawning
Make pipe ends detachable, fix W32 corner-cases
Fix file close to work on pipe ends correctly
More flexible native->GNUnet_handle file opening
Minimal W32 version to XP, use uniqueness flag
Use passed-by-fd pipe to control chilren
Make w32cat pipe-controllable
Oops
Temporary fix to make test_os_start_process pass again
Another highly temporary fix to make os_start_process pass
Try to fix process spawning
Detect bad control pipes earlier
Make pipe sanity check work on W32 too
It's on stack...
Fix git version detection in configure (D'oh...)
Don't generate namespace key files in the root dir...
Asynchronous namespace creation. With a test.
NS test: Use different URIs for different phases
Silence a warning
Implement and use GNUNET_CRYPTO_rsa_get_public_key_hash
Implement and use GNUNET_FS_namespace_get_public_key_hash
Be able to only-read-not-create RSA keys
Be able to open existing namespaces
GNUNET_FS_namespace_rename
Make 2nd and 3rd arguments of GNUNET_FS_namespace_rename constant pointers
Add error messages to GNUNET_FS_uri_sks_create
Implement GNUNET_FS_namespace_insert_advertisement_into_metadata
Accept time strings where number and unit are not separated by a space
Increase GNUNET_CRYPTO_ECC_MAX_PUBLIC_KEY_LENGTH
Correct testbed dependencies
Pseudonym-removal function
Correctly destroy OS_Process (cosmetic fix)
All-encompassing ARM update
Forgot to add the file...
Receive_done in monitor hanlding
Use correct argument for receive_done
Fix arm service message handler declaration
Modify one of the tests to check service list
Continue processing after listing services
Do log messages with wrong size
Use correct size for messages. Again. D'oh.
Correct %s in some cases
Docfixes
A buildslave patch for our W32 slaves
disconnect -> disconnect_and_free rename
Simplify ARM alloc/connect
Fix warnings and fix/improve logging in network.c
Fix the buildslave patch
A temporary test for buildslave process killing
ARGH. Automake and its special variables...
Update w32 buildslave patch to 0.8.6p1, and FINALLY FIX THE DAMN THING
Remove temporary W32 buildslave test
Don't wcscmp a NULL pointer
Use a W32 constant in a pure W32 program
Update plibc header
Use safer functions in w32 watchdog
Better mingw-w64 compatibility for win.c
Fix mingw-w64 warnings in w32nsp
Try to fix fs block plugin deps
Mingw-w64 compatibility for VPN helper
Mingw-w64 compatibility for W32 exit helper
Fix a warning - include ws2 before windows.h
Better fix for w32nsp warnings
Fix chat to compile and link successfully
Correctly forget task ID once task runs
Don't start accept task twice
Fix test_os_startprocess
Adjust test_arm_api after API changes in r27231
Try to fix mesh dependencies
Make gnunet-datastore dependent on libgnunetdatastore
Rename WAIT_TIMEOUT (clashes with windows.h)
Fix invocation of just-built tools
Modernize plibc, use MinGW ANSI STDIO
Try to fix a warning
Fix the lack of memrchr on W32
Make w32 gns helper compile again
Make exponential backoff test compile again
Link mesh tests to mesh lib, duh
Fix perf_ats to work on W32
Remove short hashes from W32 GNS helper
Link gns to vpn - always
Fix the use of s/S field types with ANSI-compliant wprintfs
No valgrind by default, please
Fix integration configs: INCLUDE -> INLINE
Add forward declaration for tunnel_send_queued_data
W32 compatibility for test_testbed_api_barriers
Fix main test function arguments for perf_ats
Remove generated config file
Fix the fix
Fix memory allocation in set service
Fix timing problems in *_select() on W32
Fix a typo
Remove static psycstore.conf
Make psycstore.conf generatable
Disable test_testbed_api_barriers on W32
Don't cancel if we didn't start watching anything in the first place
Don't go into infinite loop once only uncancellable head is left
Fix the endless loop fix - don't miss pre-last item
Put missing UNIXONLY into set conf
'No valgrind by default' strikes again
Add proper includes to test_gnunet_prefix
Change console inheritance on W32
Remove the config file (no longer generated)
Non-generated config file
More comments to make the testcase easier to understand later. Increase
code verbosity to make it obvious what is being tested.
Fix connection_watchdog to work on W32
Shift the includes (include platform.h earlier)
Add missing blackslash
Add missing include
Update plibc header
Tweak the stat-calling code
Replace OFF_T with off_t
Update make GNUNET_STRINGS_fancy_time_to_absolute return local time on W32
Update time conversion functions docs, test converting abs UTC time to
string and back
Precussive maintenance on env/multicast/psyc
Sanity check
Misc makefile fixes
Fix to build on W32 again
Fix psycstore tests to compile
Make it compile again
Skip the testChangeOwner test on W32
Fix a typo
Change the tmp dir after .conf
Don't put back separator that wasn't there originally
Define _XOPEN_SOURCE to an integer, don't leave it empty
Link namesotre to libgnunetgnsrecord too
Beautify the version script
Add doublequotes; also, r for 'revision'
Fix a typo in util.conf
Link dv transport plugin to libgnunetats (for logging)
Pythonize the revocation test
Fix Python interpreter in revocation test
missing file
Skip gnunet-config-dependent tests when gnunet-config is missing
Add hostname to psycstore.conf (required on W32)
Add hostname and port to psyc.conf (required on W32)
Save the prefix value on W32, it does not survive unsetting
Change gnunet-config detection in gns tests
Skip tests properly
Dup the envvar before unsetting, everywhere
Update W32 GNS helper and NSP
Remove test home at the beginning and at the VERY end
Add a dummy gnunet-secretsharing-profiler.c until fdold gets around to
add the real one
Non-uniform keepalive timeout for different plugins
Clean up dns resolver code a bit
Clean up W32 NSP tools, increase verbosity
Update for (gn|c)url 7.40.0
Use more-or-equal (some machines are fast enough)
malloc -> new
Don't free opt cfg if it's NULL
W32 NSP: Don't use 64-bit types, link to libgcc statically
Fix w32nsp exports
Correctly send trailing empty header
0-long message payload is not valid
Fix 0-terminator check, logging
Support normal socket (non-NSP) name lookups in resolver (for testing)
Fudge includes to prevent GUID leaks, make non-exported functions static,
use mingw-w64-provided definitions for NSP
Increase NSP debug logging verbosity greatly
Fix debug logging in w32nsp to not to use libintl
Keep requests in a list, cancel them before disconnecting from GNS
Remove unused macros, use a different error code for early bailouts
Silence warnings in W32 NSP
Make GNUNET_STRINGS_absolute_time_to_string() return utf8-encoded string
Use newer plibc, adjust the code as needed
Fix gnunet-config check in gnunet-gns-import, don't add records twice
Add compilable gnunet-gns-import
A docfix and a style fix
Don't hardcode stat sizes
W32-compatible integartion tests, try to handle rmtree()ing of read-only
files
value__ is uint32_t, use %lu to print
Remove CREATE_NO_WINDOW flag, allow children to get our console
Add console hackery for gnunet-conversation on W32
Add GStreamer-based implementation of conversation helpers
Remove the W32 timezone hack to match r31771
Fix consensus test to compile
Fix W32 process spawning to use CREATE_NO_WINDOW only when there's no
console
Don't transmit partial packets
Link dv transport to libgnunethello
Wrap opus in ogg container
More W32 resolver workarounds
Add hostname to conversation config
Check that we're not scheduling destroy twice
Fix vpn helper argparsing logic
Fix scalarproduct argument parsing (don't scan past 0-terminator)
Check for libopus to be sufficiently new
GST-based backend does not require opus at compile time
Fix dependencies
Revert "use -Wl on -no-undefined as it is a linker option:"
Disabling secretsharing until someone fixes it
Jury-rig perf_ats to compile again. MWachs, fix this
Fix TESTS_ENVIRONMENT setting (allow GNUNET_PREFIX to be overriden, allow
TESTS_ENVIRONMENT to be overrided (use AM_TESTS_ENVIRONMENT))
Add required HOSTNAME to test config
Fix a typo
Fix consensus test to compile (whether it makes sense or not is up to
dold)
Consensus does not really depend on mesh, does it
Clean up transport connection in mesh
Update peerinfo tool test after r32025 changed the output
Warn about not building hostlist daemon
Update plibc from svn
Include winternl.h, needed for testbed
Fix ats-tests makefile
Change dependencies and linking for ats-tests
Fix barrier test dependencies (W32)
libidn is needed unconditionally for libgnunetdnsparser
USERNAME is a placeholder, not a variable
Don't crash on NULL lookup_name
Don't use the 'interface' name, it conflicts with definitions in W32
headers
There's no GNUNET_LDFLAGS, there's only GN_LIB_FLAGS
Also exclude .c libtool wrappers from pogen
Grothoff's massive indentation and documentation fixes in util/network.c
Grothoff's W32 select slist elimination and prettification
Rename DEFAULT_QUALITY to MLP_DEFAULT_QUALITY to avoid conflicting with
W32 GDI
Add support for selecting on W32 events
Add GNUNET_NETWORK_fdset_handle_set_first
Fix the use of (GNUNET_)strdup and (GNUNET_)free
W32: Also consider ERROR_INVALID_HANDLE when checking 'file' type
W32: Remove copy-pasted stray increment
Include initguid.h instead of just defining INITGUID
Move the initguid.h before gnunet_w32nsp_lib.h for it to take effect
Update bundled plibc header (don't redef S_I* constants), fix mkfifo
prototype
Wait on actual HANDLE, not the structure that contains it
Add missing const to the argument of the readiness checkers
Update plibc header from upstream
Don't shadow the system() function
Fix a typo in perf_crypto_asymmetric
Lin Tong (4):
function coverage & line coverage finished
delete print message
Please check L128 - L133
function coverage 100%
Marcello Stanisci (2):
typo
polishing the english
Markus Teich (7):
Add script to generate monkeys expression db.
it is 2014 now…
updated tutorial, please check and rebuild the pdf if you have latex
installed. (i don't)
fixed error in c tutorial.tex. The pdf still needs to be built by s.o.
with pdflatex and checked for errors!
gnunetutil: add 2d and 3d allocation including tests
libgnunetutil: fix doc and indentation for realloc
libgnunetutil: add file, line debug info to multidimensional array
allocators
Martin Schanzenbach (655):
-hello world
Added preliminary API and stubs for GNS
- added some message types, dns code
- added makefile, borrowed lots of code from dht for client handling
- commit after up, ty!
- removed add, other feedback
namestore stub api added fixes to namestore api
-gns service with dns/namestore api added
-minor changes
-more fixes
Better logging modified ns stub
-added dht handle, fixes
-dht put, some serialization
-sketched dht resolution, fixes
-tiny fix, comments
-remove pointless client handler, added more comments
-big change to resolver
-some dht stuff
-helper funcs, fixes
-new comments for resolver
-new api
-make it compile against new api, added parser utils borrowed from
dnsparser
-fixes
-process dht lookup
-more dht
-ns continuation
-cleanup
-fix
-dnsparser stuff, still incomplete
-fixes, implemented new stub
-can finally resolve alice.gnunet as well as www.bob.gnunet (no
signatures)
-cleanup
-fix
-test and config
-first try on blocks
-cleanup and bugfixes
- dht lookup stop, doc update to signatures
-periodic put fix, namestore stub fix
-doxygen, more block plugin
-fix stub, add better task handling to dht update zone task
- use cstruct
-block compiles and added to makefile
-adjusted dht block layout
-more dht block layout fixes
-try to get keys form config
-fixes
-some more config handling for tests
-fixes, added loading external trusted keys
-added test, lots of bugfixes
-fix duh
-fixes
-fix
-cleanup, fixes
-integrated block plugin
-see previous
-cleanup,fixes,logging
-fix
-fix xquery
-use common serialization, defines
-modified test
-add dht timeout
-do sth meaningful on timeout (ie no such name)
-fixes
-remove stub from makefile
-add proxy
-fix proxy
-new test, fixes, config
-fix
-new test, fixes
-see prev
-fix
-fix tests
-new test. fixes
-fixes
-dns exit
-visible error log
-log
-no script
-remove test functions
-fix
-fix
-fixes
-rewritten resolver, hits some namestore bugs
-fix
-started some gns service api
-PKEY is key no hash
-added some shorten stuff
-fixes, shortening start
-new resolver (#3), for shortening
-more shortening
-disable tests for now
-fix
-shortening in cli app (in progress)
-more generic record resolution, prepare for api lookup
-fix some leaks
-leaks, make valgrind happy
-new shorten test, fixes
-GNS service api change, replaced complicated buggy code
-leak
-lookup now works in test, removed dns hijack from defaults
-fixes
-fixes, leaks
-some cleanups.. renaming
-ignore
-fixes, added get_authority api
-new test
-add get_athority test, fixes
-bugfixes, tests
-bugfixes, debug, new test
-fix zero termination
-revert
-large cleanup and bugfixes
-resolver bugfix, gnunet-gns command line features
-fix
-t'works
-proxy magic, interceptor fix
-cleanup
-werror fix
-fix
-fix client receive, zonefile loading
-fix
-readd and fix zone iteration
-fix leaks
-fix config parsing for tests
-fixes
-fixes, block plugin change
-fixes
-fixes
-fix
-cleanup
-fix tests
-fixes
-fixes
-pseu caching + test
-fix
-prepare record postprocesssing
-mx,cname,ns postprocessing + test
-fix
-fix mx records postprocessing
-less malloc more stack
-test for zkey resolver
-fixes, interceptor now speaks zkey
-new short hashes, new short hash cmp
-fix build
-import pkey config option
-fixes, auto_import_pkey imlpl
-fix mxrr
-fix soa
-timeouts, parallel lookups
-doc, new config options
-parallel background lookups
-cli tool fix, bugfix
-doxygen
-add test
-add correct file
-doxygen
-fix
-MX
-soa
-coverity
-fix
-fixed, cleanup
-make debug output more readable
-fix
-doxygen
-fix zkey shortening
-fix, will failon werrorr
-fix warning
-add zkey shortening
-add bg lookups for shorten
-first zkey then .+, catch exception for they are not critical
-less dht record lookups
-normalize input names. important for DHT queries
-fix
-fixes
-fix
-add leho to proxy
-utf8 tolower/upper
-properly disconnect from dns
-fix
-fixes, no deflate/partial content in proxy
-fixes
-lots of fixes, new dht test
-fix
-cleanup
-add record type any, pending flag
-add records as pending
-fix, add config
-fixes, cleanup
-remove messy test
-array of fixes, finally finished test
-ign
-fix
-config cleanup and fixes
-cleanup
-fix
-fix
-proxy work
-fix
- -debug
- max name len
-tests
-fixes
-any type
-use precomp keys
-more precomp keys
-more keys
-more keys
-use hostkey as zonekey for tests
-performance tests
-ign
-renamed config options, tuned periodic puts
-vars
-fixes
-start nss
-license
-coverity
- #2313
-do not hijack dns by default -iterate over all authoritative zones
-nss build files
-preliminary nss plugin AKA works for me
-needed for nss plugin
-cleanup
-new per zone api
-per user lookups done
-fixes
-fix raw output for nss
-new proxy
-cleanup
-add mhd to .gnunet requests
- added curl
-fixes
- make mhd and curl play nice thread wise
-start replacement code
-raw for shorten
-add gns glue
- more
-fix
cleanup
-comments
-add new proxy code
-more
-more
-fixes, socks done
-fix, mhd start
-added accessors for network handle
-basic mhd working
-parse host
-build proxy only when mhd present
-fix warning
-mhd curl working, needs love
-doxygen
-fixes
-fix response buffer handling
-regex, buggy
-add GNS, fix mhd response creation bugs
-fix
-beginnings of ssl
-fix cleanup
-fixes, add leho
-leho, proxy ssl
-on the fly cert generation
- s/openssl/gnutls needs testing
- cleanup
-zkey re
-fix
-fix
-zkey, fix gnutls import
-fixes
-coverity, cleanup
-unix domain sockets
-add connection limit
-read sockfile from cfg
new API for shorten key
-cleanup
-add cached flag for performance
-test fixed, modified shorten
-verbose
-doxy
Script for CA generation. Making it easy with config option
-fixes
-fix
autoimport CA
-fixes
-lots of fixes
-fix, doxy
-fix html detection
-make CA info better, connection limit mhd
-big shorten algorithm change, hopefully fixed conflict
- big shorten algorithm rewrite - API change for lookup - tests fixed -
introduces 3 zone system: root, private and shortened
-doxy
Add VPN resolution
-add vpn record type to namestore
-1variableperline
-NS delegation WIP
-more NS delegation
-ign
-proper data layout vpn record
-fix
properly handle CNAME as specified in rfc1034#section-3.6.2
-fix cname, more ns/vpn delegation
-ign
-pregen keys for test
-add CNAME handling for GNS
-ns tests, fixes
-ign
-zone revocation
-bugfixes
-ign
-fixed record flags in tests. proper expiration from NS missing
-cleanup
-fixes
-coverity
-doxy
-fix cnames #2448
Added proper SRV handling and test
-make record data parsing easier for namestore,gns with dnsparser structs
-test cleanup
-test for working system resolver
-added vpn gns test, fixes
-fix??
-fix.. that took ages to find
-test now working
-fix tests
-fix warning
-add rp filter fix
-ignore
-add rp filter fix to dns helper
-remove rp filter fix from test
-fix order
-fix
-fix
-fix bigendian
-add af check
-fix
-big commit, make new ns and valgrind happy
-fix
-add ptr.+
-fix
-fix inconsisten record naming
-onlu linux
-fix 2447
-coverity, fix shorten test
-fix
-fix
-pregen hostkey
-add key to makefile
-fix
-fix
-fix
-hopefully make sparc happy
-check for ipt
-doxy
-fix
-fix non GNS TLD lookup
-towards cookies... broke resolver
-fix
-fix notfound handling
-add some statistics to gns
-fix
-fix
-start better zone iteration scheduling
-fix
-do not shorten in bg lookup
-correctly document cache only flag
-fix 0 record zone iteration
-improve change in record number on zone iteration
-fix
-fix
-update test conf
-fix resover memleak, more proxy cookies
-cleanup
-cleanup
-cleanup
-add cookie support
-fix request handling
-fix
-fix
-fix site of origin refs
-fix
-add uri values to curl
-fix compile
-api love
-more api
-more new api
-fix
-large proxy rewrite
-broken
-fix
-get working
-rip timeout issue
-fixes, cleanup
-fix bo
-start post
-urlencode post working, performance, replace location hdr val
-fix
-fix
-leak
-fix shorten
-fix ports
-fixes
-fix
-add test
-silly
-fix
-fix
-fix
-this should work
-mhd use no listen socket
-fixes, check for mhd version
-fix
- add check to test, code cleanup
-some notes for dns2gns
-coverity
-fix escaping of url
-leak
-leak
-move record defs out of dnsparser, add tlsa rr
-tlsa rr added to namestore
-tlsa rr added to namelist
-proxy: add GNS header
-handle race
-add httppost, put, fixes
-leaks
-actually add CORS header
-add header
-dns2gns
-fix
-update gns manual
-fix gns cli hang, fix help
-fix hang for good
-fix NetworkRecord/GNSRecordBlock madness
-not needed
-bugfix dnsstub, add dns support to gns gateway
-doxygen
-change vpn wire data format
-attempt at zoneinfo page
-tmp fix
-real tmp fix
-real fix
-fix
-fix threepeer test
-really fix test
-fix to api
doxy
-coverity
-less verbose
-fix record expiration in test
-fix dht flags
-tests
-fix
-ignore
-ignore
-test
-test
-towards .gads
-bugsbugsbugs
-fixes
-fix
-fix
-fix
-fix
-disable for now
leak
- apex is allowed
-fix
-fix
-fix
-grml fcfs
-short hash output func
-fixes
-doxy
-cleanup
-add gns proxy man page
-cleanup
-i'm back
-begin new test cases
-fix gns not returning anything of value
-more script tests
-cname test, warn: bug:endless recursion
-mx test, mx type not recognized
-more tests
-more tests
-fix name assembly for ns delegation
-dht test
-add dht test
-cleanup
-cleanup
-pseu shorten test, fails
-more tests
dht zone put interval should never be longer than minimim relative
expiration time of a record
-fix
expiration test
-add test for resolving refreshed record, fails
-more test coverage, more bugs found
-ignore
-soa test
-revocation
-remove old tests
-quick put test
-fix
-fix
-clea namecache in test
-fix
fix reverse DNS lookup, prepare GNS reverse lookup
-no reverse GNS
fix only cache variable long/short
- fix segv when remote closes connection
REST Service
- fixes, intendation
-fixes
-fix
-fix
-add identity REST, fixes
-add ego creation
-add delete REST
-bugfixes, change plugin call
-fixes, add set ego for service
-towards JSON API compat
-more json api
-change API
-fix, debug
-fix
-start a lib for REST and json:api
-more rest handling
-more REST
-fix mem corruption
-add rename
-fix
-minor put vs patch
-gns rest
-minor
-PUT since MHD cannot PATCH; add config
-doxy
- fix
- namestore rest
-json decrefs
-make static
- add CORS logic
- give control over CORS methods to plugin
-fix
- clean up gns rest api
- Identity token support
- flat namestore plugin (experimental)
-add flat namecache
-fix
- fix
- fix
-fix
-fix
- update
- update
- revert
- move rest plugins into rest directory where they belong
- revert plugin move. Add new identity-token
- see last commit
- fix install path for plugins
- token issuing
- purge token creation from plugin
- issue token including attributes
- fix invalid free in token rest, fix build bug, move rest up in
dependancy list
- fixes
- fix leaks, add token list
- fixes, store issued tokens in namestore
- add identity attribute processing
- implicitly enable rest functions in experimental if jansson.h present
-fix build
- fix flat plugin
- Add expiration and renewal parameters - Remove debug output as errors -
Fix buggy error output
-add identity token service
- fixes
- update; add attribute updates
-update
- update flat plugin to namestore flat plugin status. Add config filename.
- add flat plugin config filename path
- Add ecdsa ecdh functions - Update identity token to encrypted protocol
- minus leaks
- encrypted token
- verify ecdh signature in token code
- Refactor
- update token handling
- Start commenting. Major refactor
- Finish refactoring
- More heavy refactoring. Probably lots of broken things to see here.
- update
- fix build, fix bugs
-add config
- refactor
- update issue handling, fix memleaks
- fixes, modify api
-remove debug output
-minor
- fix
- sscanf type fix
- fix
- Refactor jsonapi into separate module
- missed file; bugfix
- add jsonapi tests
- rework rest/jsonapi API; bugfixes
- fix string format
Update jsonapi to current specs, refactor
- build rest before jsonapi
- remove HAVE_REST, remove json reference from rest
- refactor jsonpi utils, add test
- ignore
- add test, bugfix
- jsonapi build fix, add peerstore flat (EXP)
- add peerstore tests
-fix
-fix
- Fix #4532
-oops, remove printf
-fix #4541
- revert to r37134. Broken build
- fix rest plugin
- fix more rest and jsonapi
-fix
-fix
- fix warnings
- fix warnings
-fix format warning
- fix coverity
-coverity
-coverity
-coverity
-coverity
-coverity
-fix fixmes
move to MQ api
-draft API
-fix format string
-doxygen
-update AUTHORS
- fix coverity
- add persistent connections
-ignore
- Remove content length header in responses
-fix various bugs
-fix BADF and add encoding filter
-fix ipv6; dont remove Expect header from curl
- fixing a variety of bugs including POSTing data, content-length
handling and improving connection handling
-only check cert once
-only check ssl cert once
- move to new MQ API
-new service API
-new service API
-fix
- Add reverse resolution with limited functionality
-add tests
-add reverse autoadd; with test
-ignore
-fix
-fix
-fix
-fix
Matthias Wachs (4193):
<none>
Added code to test gnunet's dns resolver service. The test code compares
gnunet's results to the system's own forward and reverse name resolution. Using
dns root server as a targets to get a static 1:1 mapping
source code changed to meet coding conventions
<none>
<none>
Added type GNUNET_MESSAGE_TYPE_HOSTLIST_ADVERTISEMENT
config files for learning testcase
<none>
testcase for hostlist learning
Initial changes to implement hostlist learning
<none>
<none>
hostlist client gets notified when hostlist advertisments arrive
added -a argument to start advertisement
Modifications so hostlist client and server will notified when new peers
connect
<none>
<none>
added code to build the advertisement message
<none>
added function GNUNET_RESOLVER_local_hostname_get to resolve local
hostname
<none>
<none>
<none>
<none>
Complete code to send hostlist advertisements to new connected peers
Correct handling of daemon options to en/disable learning and advertising
added methods to load and write hostfiles
added some code for hostlist management
<none>
<none>
update a hostlist if already existing in database
<none>
And now for something completely different..complete implementation
changed from hashmap to dll
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
removed peer from hostlist structure
added safety check in writing and saving hostlists to stop uncontrolled
writing
Hostlists are now saved to disk regularly
hostlist client is now alternating between preconfigured servers and
learned hostlists when downloading
<none>
<none>
<none>
<none>
<none>
added code to update hostlist statistics
hostlist file now gets deleted if learning is disabled
<none>
<none>
<none>
<none>
<none>
<none>
<none>
creating a correct testcase
<none>
<none>
changed definition MAX_URL_LENGTH to gnunet_daemon_hostlist.h
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
replaced local_hostname_get with local_fqdn_get
replaced local_hostname_get with local_fqdn_get
<none>
<none>
partly working: hostlists are downloaded before added to datastore
<none>
<none>
<none>
<none>
<none>
<none>
- Cleaning up and restructuring - Fixing download tasks to shutdown
cleanly
<none>
<none>
Disabled debugging and verbose to suppress logging messages
Changed implementation of hostlist updating from regular polling and
getting triggered by the connect handler to triggering by peerinfo notification
API
<none>
<none>
testcase with one peer is passing
Download limit for client implemented
- Implemented download limit on client side - Disabled verbose
hostlist client and server now use the same limit for transfers:
MAX_BYTES_PER_HOSTLISTS = 500000
<none>
changed to work better on slow machines
<none>
<none>
Added configuration option to set hostname used for advertisements Fixes
problems on systems with misconfigured dns resolution for local hostname
The external dns name used can now be set in the config file By default
EXTERNAL_DNS_NAME it is set to localhost to prevent the testcase from failing
on systems with incorrect name resolution
<none>
initial preparations before implementing http transport plugin
<none>
Added documentation for defines
<none>
<none>
<none>
<none>
<none>
<none>
clean shutdown
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
fixed crashing durign shutdown on freebsd7
<none>
<none>
<none>
<none>
<none>
<none>
<none>
rewrite from scratch
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
Added MHD_OPTION_NOTIFY_COMPLETED Handler
<none>
<none>
<none>
<none>
<none>
Canceling pending STATISTIC_gets prevents testcase from segfaulting (I
hope)
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
cleaning up the testcase
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
changes to make codesonar and coverity happy
added support for multiple messages in a transfer
<none>
<none>
Added MessageTokenizer functionality
<none>
<none>
Now using multihashmap to store sessions
<none>
outbound messages are now stored in a dll
<none>
This line, and those below, will be ignored--
plugin changed to be loadable multiple times
<none>
<none>
<none>
<none>
<none>
Now supporting chunked transfers with paused curl handles
<none>
<none>
<none>
<none>
Added timeout_watchdog, a small helper tool for the buildbots
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
Fixed compile warnings under windows
Fixed compile warnings under windows
Fixed compile warnings under windows
Fixing compile error due to wrong type
adding test_plugin_transport_http test
<none>
<none>
clean up
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
Checking for libcurl >= 7.18.0 curl_easy_pause is used in
transport_plugin_http
<none>
<none>
<none>
<none>
Fixed "Bad file descriptor" error after plugin is unloaded by canceling
previously scheduled curl tasks
<none>
<none>
<none>
FIX BUG: Adding invalid curl PUT handle to multihandle causing error
not skipping loopback devices
Cleaned up code Rewritten session selection Improved session section by
skipping session lookups Added binding to interface
<none>
<none>
added configuration value to en-/disable ipv6 usage
completed code for binding to a specified port
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
FIX: no curl scheduling if no curl handles are running, because this
causes segfault on sparc machine
<none>
<none>
<none>
<none>
<none>
<none>
<none>
initial checkin of https transport
Added shell script to create https certificates
added test configuration files for https
<none>
<none>
<none>
changed https pretty_printer to https://
fixed certificate creation script
<none>
Improved tls certificate creation in https plugin
improved test order
<none>
<none>
removed piping stderr to /dev/null
changed required libcurl version to >= 7.20.1 this is a preliminary
change since there is no newer version for my system final required version
will be >= 7.21.0
changed debug output
<none>
<none>
<none>
<none>
<none>
<none>
Removing private key file and certificate before and after every run to
test key and cert generation
- debug output decreased - private keys and certs removed after all
testcases
changed https cipher from aes256 to rc4-128 SHA1 -> 3x faster crypto_init
string can be configured in the config file for testing purposes
removed debug out
no debug output
Fixed Codesonar Bug 10128.60649: Leak
Fixed Codesonar 10128.70650 Leak
Fixing bugs
fixing bugs
fixing bugs
fixing bugs
<none>
make windows client work
fixing little stuff that made this not work
Added MHD_EXTERNAL_LOGGER to log mhd messages with GNUnet logger
<none>
<none>
just a little script
FIX: bugs found by klocwork
<none>
Increased timeout Just a test if it fixes the freebsd problems
fixed: klocwork bug #305
fixed bugs found by Klocwork
fixed bugs found by klocwork
changed MHD Buffer size and enabled TCP_NO_DELAY for curl
added prefixes for strace and callgrind
missing newline
<none>
doxygen documentation
fixed doxygen documentation
Improved documentation
Improved documentation
removed unused variable
<none>
IN PROGRESS: fixing bsd bug
Smallers steps to keep plugin running
<none>
Hopefully fixed BSD bug
MHD debug messages switched off
Merging http and https plugin ... first steps
merging http and https: next steps
<none>
<none>
HTTPS testcases pass
clean up
<none>
HTTP & HTTPS plugin source is now merged and boths plugins are build from
plugin_transport_http.c
https plugin code merged into plugin_transport_http.c -> old file can be
removed
added curl logger to integrate curl messages with gnunet log
<none>
Should now compile under windows: - removed #include <arpa/inet.h> -
MINGW32: alloca() is defined in malloc.h
Added some information when test fails
<none>
BUG 0001602: A patch to fix process spawning with redirected std streams
0001602: A patch to fix process spawning with redirected std streams
0001602: A patch to fix process spawning with redirected std streams
0001602: A patch to fix process spawning with redirected std streams
0001602: A patch to fix process spawning with redirected std streams
0001602: A patch to fix process spawning with redirected std streams
0001602: A patch to fix process spawning with redirected std streams
<none>
<none>
<none>
Removed memory leak in curl_logger
removed unused variable
removed malloc from curl_logger
changed order: build hello before block
Removed includes for windows compatibility
Introduced limited per connection queue size
Definition of ATS costs
Added cost_report to plugin environment for ATS
fixed copy and paste error
Added note about converting dBm to power P
correct and better to understand
uups...never commit things before having coffee
added basic definition of quality information
<none>
<none>
<none>
Bug 0001607: Need to $-expand HOSTLISTFILE
stricter suggested address checking
removed output
no unit given
no unit given
implemented inbound quota regulation
<none>
Removed IPv6 linklocal check This check is not needed, since no linklocal
addresses are added to plugins address list
<none>
removed not needed check
fixed bug related to Codesonar 10818.81116
<none>
<none>
minor changes to debug output
First steps to get GNUnet running under OpenIndiana: include needed for
network.c
removed additional >
testcase to test quota compliance
<none>
<none>
<none>
<none>
<none>
<none>
test now works
<none>
<none>
<none>
comment out because of compile error
<none>
<none>
<none>
<none>
changed bps (bits per second) to Bps (bytes per second)
more fixes...
FIX: value (now.value - av->last_update__.value) is always 0
Fixed comparison to get correct timeout values th->timeout is absolut
value, duration was relative, so comparison was never true
some changes bandwidth carriage
Changed GNUNET_TIME_Absolute and GNUNET_TIME_Relative to allow safe
comparisons between time values
Refactoring GNUNET_time structs
Refactoring gnunet_time
Refactoring GNUNET_time
Refactoring gnunet_time
Refactoring gnunet time
Refactoring gnunet time
Refactoring gnunet time
Refactoring gnunet time
Refactoring gnunet time
Refactoring gnunet time
Refactoring gnunet _time
Refactoring gnunet time
Refactoring gnunet time
Refactoring gnunet time
Refactoring gnunet time
fixed
fixed problems with GNUNET_time_relative
Changed struct GNUNET_TIME_RelativeNBO
<none>
safey: sorry...please check the file...this makefile does not work if I
don't comment the two lines out
<none>
<none>
Adding support for other plugins
<none>
Adding quota compliance test for http
quota compliance testing for https
<none>
and quota compliance for udp
fixed bugs caused by eclipse refactoring
fixed eclipse refactoring bugs
removal proposed by LRN
fixed eclipse refactoring bugs
fixed eclipse refactoring bugs
<none>
Added testcases for asymmetric quotas
more asymmetric tests
a lot of new testcases
new: increasing quotas
new: stopping test when maximum transmission rate is hit
<none>
<none>
Initial code base for core level quota compliance test
<none>
<none>
removed assertion since you can not blame hostlist-client how often core
dis/connect handlers are called instead code just ensures that the value is >= 0
reverting changes
improved output
removed sender_constant testcase does not work on transport level
test has to be more flexible with lower quota value
<none>
config files for core quota compliance test
<none>
limiting send rate
<none>
<none>
added statistics counter for discarded messages
added statistics to get information about discarded messages
did not compile
fix: correct plugin name
added check to prevent testcase to exit with success if throughput is 0
test if a longer measurement interval results in better results for
smaller quota values
Next try to fix the tcp reliability test problem This config fixes the
failing at least on my system
added new perf counter
fixed coverity bug #10022
fixed coverity bug #10042
test takes too long
Added core api changes
Basic problem fixed: Line 477: Timeout for
GNUNET_CORE_peer_request_connect with 1 sec. too short to connect to other peer
rewritten test...not yet finished
More changes
<none>
<none>
added tests with limited sender and receiver quotas
Configuration files for asymmetric tests
Quota compliance testcases should ready to be commited...
Renaming MSIZE to MESSAGESIZE since MSIZE seems to be in use on FreeBSD
Systems
changed confs to use seperate ports
different directories for each testcase
clean port assignment
<none>
<none>
<none>
Adding soft quota
Added size check for unix path configs
Returns now GNUNET_SYSERR
Activated progress bar
<none>
improved bandwidth distribution and outbound quota compliance
removed break to prevent testcases dumping core
<none>
<none>
build nat for windows
Transmitting ATS information to transport api
added ATS information to InboundMessage
changed api to support ats structs
changes to transport_api
added api changes to testcases
added transport api changes to core
fixed transport api
added multi plugin transport api testcase
<none>
<none>
<none>
fixed: - off-by-one in msg size - changed 0 to
GNUNET_TRANSPORT_ATS_ARRAY_TERMINATOR
<none>
<none>
Working version of transport service with ATS integration all tests
passing on my machine
<none>
<none>
this is a merged version of revision 13866 and my latestest changes
without the old ats changes all changes from r13826 not made from me are
included
<none>
<none>
<none>
one line to break it all...
fixed recommited off-by one bringing core_api to fail
sparcbot requires longer timeout
<none>
transport api now passes ATS data to clients
- modified transport plugin interface to support ATS - modified transport
plugins to support ATS
modified reiceive call
added counter to avoid single measurements failing to break the whole test
delete this strange file
<none>
just to remove "unused variable" error
removing "unused variable" error
<none>
<none>
fix
added fix
added fix
fix bug from klocwork
fix: function and variable with same name
klocwork id 751: fixed memory leak
doxygen fixes
safety check
wlan changes
fixes
<none>
<none>
disabled wlan test
<none>
<none>
<none>
doxygen
doxygen
doxygen
doxygen
<none>
changed reliability to cancel notify_transmit_ready
no comment!
fixed: "Error in communication with PEERINFO service" printed
Fixed failing test and discrepancy between documentation and implemented
functionality:
not needed
Changed peerinfo api to distinguish between last element and timeout
cleaned up config
Fixed failing on sparcbot:
doxygen
initial changes to get the nat test to run
added ayush's ssh key
adding ats framework
ats testcase
doxygen
unix domains sockets are now preferred
debug and timeout
ats
disable ats debug
changes to configure.ac
<none>
more ATS
clean up and changes
<none>
removed debug
<none>
<none>
ats changes
check for glpk >= 4.32
error -> warn
ats
reduce exec frequency
bug hunting
ats
msg
ayush's new pubkey
the new formulation including feasibility constraints
more changes
and executing the MLP
more changes for ats
<none>
now it is really time to leave
more ats
define for systems without glpk
run only when glpk
<none>
<none>
- mlp presolver is crappy: 0 for all variables is an integer optimal
solution -> presolving with simplex solver -> benefit: glpk version check is
not needed anymore and jaunty-arm should be happy
bound had wring direction...MLP now works as far as it is implemented
optimazitaion constrain 6 works
constraint 8
Constraint 9 implemented MLP is solvable and afaik well defined
more changes
new testcase for ats based on testing framework
<none>
<none>
latest changes
missing conf
script to execute test with different number of peers
cleanup
<none>
changes
-verbose
starting to import ressource costs
-debug
<none>
moved define
commented out dev variables
<none>
implemented: quality metrics
fix: numerical problems with final bw assignment check
did not compile
<none>
gauger
changes
<none>
clean up
loading ressources
<none>
implemented: cost
<none>
back to double
always remember: GNUNET_i2s is not reentrant!!!!
<none>
<none>
changes
now with history
clean up and non functional improvements
ADDED: Dump mlp and solution to disc
<none>
improved mlp reoptimization
changes
<none>
<none>
<none>
<none>
<none>
finished: rebuilding problem only when peers (dis)connect starting:
modififying matrix when qm and cr changes
optimized code
quality and ressource updates
many changes...improvements
<none>
improvements for better benchmarking
benchmarking script + config files
<none>
changing default presolver to mlp builtin
the mlp presolver does not provide an initial optimal solution so we have
to use simplex solver to provide this optimal solution
test update function to ensure number of changes
<none>
<none>
setting ats intervals to productive use settings
<none>
updated testconf
not needed
fixed
<none>
SHUTDOWN
SHUTDOWN
SHUTDOWN
SHUTDOWN
shutdown
timeout
scheduler
<none>
fixed memory leak
fixed bug
new glpk performance benchmark
improvements
<none>
<none>
<none>
<none>
<none>
<none>
recalculate ats when sessions end
<none>
<none>
removed valgrind
don't prevent other testcases
<none>
broke on freebsd
<none>
bug hunting
surpressed identity problem
<none>
<none>
bug fixed
<none>
changed confs for second mesh addr
this test is just not working
disable mesh autostart
fixed segfault
fixed bug breaking transport
fixing assertion failure
bug fix: ipv6 did not check bind address
more fixes in address handling
Another fix for specific bind addresses
chvc fix
fixed: transport disconnects when inbound sessions exists
ALLOW_NAT is not even used in tcp plugin
for the nat we need the client and server?
<none>
adding wlan testcases
updated confs
<none>
<none>
changed confs
changed confs
changed confs
changed confs
changed confs
also uses testing framework
disabling mesh
changes to testing
USE_LOCALADDR not needed in configs, is set by testing framework
not needed
not needed
NEW: local addresses are filtered
the "we have the time but not the people" way of regression testing
Removing this lines breaks the transport tests on Freebsd Failing tests
begin with rev 14256
no unix on freebsd
unix on freebsd
testcase
bug on bsd
fix: unlinking unix domain sockets on freebsd
now also works on freebsd
renaming
changed gauger calls
freebsd related changes to unix
fix format
should fix hostkey problem on efikamx
fixing codesonar 11754.107630
fixing codesonar bugs
removed output
disable mesh
wlan tests fail
<none>
<none>
llu prob
this code made no sense in this context
reducing debug msgs
minor changes
shutdown issuess
<none>
<none>
fixes 0-length a2s assertion
removed http
fixing perf_gnunet_service_fs_p2p_trust
fix for mantis bug 1686
<none>
Bug as described in
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=590765
<none>
removing glpk output
<none>
<none>
<none>
<none>
statistics should not be disabled
fix segfault
<none>
disabling ats for release
<none>
use different ports
simple switch to disable ats
<none>
improvements
fix
updated _api_, _api_reliability_ and api_unrel... confs
clean up in transport tests
changes for wlan
confs
changes to tests
<none>
fix resolver api
mem leak
<none>
<none>
<none>
<none>
<none>
fixing: assertion failed at transport_api_new.c:1277
fixing assertion transport_api_new.c:379
patch for time out bug
<none>
restrict number of connections
fixing memory leak mantis #1700
memory leak
bug hunting
memory leak
fixing "minor changes" and repair things that obviously never worked
getting plugins to work again
http now supports minor changes
fixing memory leak
removing http & https plugin testcase the testcases do not work with the
asynchronous address notifications and imho we need a more general plugin
testcase coming with the transport testing lib
disabling output
need more checks
this brings api_multi to pass but binding to localhost causses
error...more investigation needed!
remove output
pre-lunch commit of nat implementation for http(s)
<none>
<none>
<none>
<none>
this is no reason to break: this happens during shutdown, when the -
plugins are unloaded - the addresses are freed - and the NAT_unregister does
the callback to remove addresses but the plugin adn addresses are already freed
plugin clean up after failing assertion fixed
changes to address mgmt
old loop made plugins inaccessible
clean up
fixes mantis 1726: wrong number of addresses was returned when BINDTO is
used
<none>
remove http/s from experimental
<none>
configuration templates to be used with @INLINE@
<none>
<none>
fixed: do not disconnect from neighbour if an address validations fails
but other validations are still pending
<none>
putting ats to external file... first steps
ats
new and modified testcases
changes to ats
conf files
ats related changes
ats updates & peer disconnects
splitting transport service and ats
more changes to ATS it's running, minor changes have to completed
<none>
peers with inbound connections have to validate these immediately
debug output
additional assertion to find segfault
use after free
<none>
printer: print ip if dns timeout occurs
return ip address for timed out reverse lookups
remove reverse dns timeout code
some docu
addiional checks
<none>
<none>
new pretty printer
scheduling order is important
<none>
Mantis 1736
old address notification
removing values not needed
<none>
<none>
<none>
aaded new function to iterate over section entries
preliminary commit to prevent more conflicts
<none>
<none>
<none>
renaming ats files
removing breaks
reverting changes
<none>
<none>
<none>
<none>
<none>
ats data configuration
related to mantis bug 1744
plane hacking
transport testing testcase
cb wrapper for connecting peers
version with connecting peer working
connecting peers fully implemented ... time for PIIZZA!
documentation
test_transport_api ported to use new testing lib
fixing memory leaks
reliability uses testing lib
<none>
<none>
WLAN does not compile Disabled until David fixes the issue
indent corrupted the file
enabling wlan in makefile, since byteorder.h issue is fixed
improved shutdown
improved shutdown
<none>
fixing segfault caused by missing hashmap init
new test to check transport start and shutdown
the test
<none>
minor changes
fix: - correct message size - correct strlen
one char to break it all...
Extended ATS and transport service to store session IDs to support
inbound sessions
transport service sends wrong message type to transport_api transport api
requests GNUNET_MESSAGE_TYPE_TRANSPORT_RECV for payload messages
disabled indent in define section
fixing thrown break during shutdown
changes to server lib
fixed assertion in clients_handle_start():390
not needed anymore
- handle_pong used wrong address size - TODO: GNUNET_ATS_address_update
crashes transport service
fundamentals for an ats test
removing unique id code
tests now use GNUNET_OS_check_helper_binary to find helper executables
gnunet-nat-server was not correctly installed
check if gnunet-nat-server is correctly installed
reenabling gauger support with automatic testname detection
ported unreliability to transport testing
general intendation some progress entertainment in
transport/test_transport_api_disconnect.c
<none>
fixed bug https://www.gnunet.org/bugs/view.php?id=1762
migrated multiaddress to testing framework + progress entertainment
moving code to generate cfg name used in every test to testing lib
utility functions
latest changes
using new function + additional safety checks
porting all tests using testing lib
minor fixes
fixing floating point execption
ensure neighbour map is existing
fix shutdown order to prevent segfault
separate transport api debug messages from transport debugging
re-enabling ats in validation
fixing issue with unix domain sockets: updating sessions with plugins not
using sessions failed if (session == NULL) addresses where just overwritten
debugging
small utility to connect running peers
<none>
fixing it one more time!
udp now working with new transport service udp plugin should not call
transmit_continuation, if NULL
<none>
gnunet-service-transport_neighbours.c requires ats_count > 0, at lest \0
terminator is required
not including \0-terminator is different from typical semantic more
import: functions using this iterator already use it this way!
correct usage of ats \0 termination
adding statistics about connected peers
fixing gnunet-transport-list-connections step 1: wrong messagetype just
discards request
porting gnunet-transport-list-connections to binary address
<none>
renaming and fixing
merging ats arrays is working but needs an additional testcase
fix in address update + testcase
moving ats test from transport dir to separate dir
moving ats tests to ../ats-test
basics for a testcase with a limited number of open files
canceling wrong task
throttling down to ~5MB/s
no hard fail in case out transmit_ready timeout instead shut peers
correctly down
fix: ats data
fix
http plugin revisited
fixing coverity bugs
more functionality
pre-launch commit
implemented disconnect notifications for outbound connections
server side connection accepting and merging
fix
error msg
<none>
<none>
client sending & receiving
transmitting data
sending and receiving implemented
fixes
update buildbots first...
fixing https://www.gnunet.org/bugs/view.php?id=1800
terminated sessions were never removed from ATS
max_bps was never used
increasing timeout
test case for udp and constant packet size
binding to specific address
removing incoherent address formats cleaning up logging
NAT is reading the [nat] section to figure out if to use ipv6 :-(
fixes
fixed: - session matching - curl send handle suspending
test_transport_api_http now working!
can now send a message
working version
mem leak
renaming plugin
<none>
<none>
client bw throttling
comments and clean up
remove limited address reporting
- renaming test_transport_api_multiaddress to test_transport_api_timeout
- adding new timeout tests for http and https
unused var
disabled until configure.ac is working
setting default connection timeout depending on MHD version
changing to constant
disable verbose output
removing quota set to max_bps cfg setting
fixing list-connections to show correct peer id extending
neighbour-iterate to return plugin, addr and addresslen
added assertion to find https://www.gnunet.org/bugs/view.php?id=1808
check if transport client is still connected before sending SEND_OK
fixing #1801
some diagnoses
beautification
putting session handling in function
just to be sure curl is not causing the disconnects
added functionality: immediate server rescheduling
re-enabling timeouts
fixing #17134
<none>
bugfixes timeout handling for other transmission direction
fix for 1819: when HELLO message arrives, peer's public key is added to
peerinfo even no addresses could be validated
fixing bug #1812
correct configuration for the nat testcase
implemented: client-only mode
config files for http nat test
http https nat tests
removing experimental from http
correct cfg file for tcp_nat test
more beautification
<none>
<none>
http/https nat tests
remove logging
<none>
<none>
<none>
scheduling issue
error handling if server could not be started
log2 does not exist in math.h on freebsd
compile error
logging
logging
<none>
changed get addresses functionality and added IPv6 test
<none>
mem corruption
<none>
<none>
<none>
ipv6 test for efikamx done
transport-testing is not stateless anymore -> fixed assertion failure
transport_api.c:1348
double testing done
assertions
naming
fixes
modifiying transport service
quota changes
fixed notification order
rescheduling time
immediate scheduling
unix can send with force-reliable bi-directional
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
<none>
safety check: session address comparison
address expiration
<none>
fixing mantis #1825
more expiration
<none>
safety check
<none>
compile
<none>
removign ats from neighbour
disconnect on zero bandwidth notification
<none>
<none>
ats service
missing
<none>
<none>
<none>
test + fix
<none>
<none>
<none>
<none>
<none>
remove addresses on client disconnect
no merging just replacing
issue: zero-length plugin and addr fixing: message size parsing address
comparison
more bw to get the reliability tests passing
updating configs for ats
<none>
disabling consumption test and verbose
more confs
increasign sockets for limited test reverting neighbours: nice idea...not
so easy...but ats is more important
new quota compliance
asymmetric quota test
<none>
<none>
ugh
crappy cfg
more crappy cfgs
<none>
all tests use generated configs and pass on my system
crash on 32 bit
unused
fix fail reason
timeout too short for slow machines
<none>
additional info
analysis: when easy_handles are removed and destroyed, on some systems
curl_multihandle informs about that with return code "NO ERROR" -> not a bug
unused variable
coverity fixes
coverity fixes
fixing mantis #18773
fixing mantis #1833
order of sending ack and notification
transmitting flow control information between peers
concept for new completely session based udp plugin old plugin required
too many dirty hacks to pass required data around
mantis #1844
typo
<none>
fixing 1841
hmac issues
work in progress - neighbour management based on a FSM - 3-way connect
handshake
implemented: send connect send connect_ack receive connect_ack -> mark
peer as connected
<none>
<none>
3 way handshake works fixing now sending and quotas
<none>
latest changes:
test_transport_api_tcp passes!
<none>
more changes:
quota propagation
<none>
<none>
message types
<none>
implemented ATS_suggest timeout
canceling is important!
I do not like the idea of guessing the responsiveness a service
<none>
<none>
new test modified makefile
fixing bug
return value
one commit to ruin the day: give transport to the buildbots
this is a place to not remove the ats task... we are willing to wait
keep alive task scheduling
moving transport protocol section
<none>
state and unused var
complete state reset functionality
connection not up bug
moving files from dev to final location
<none>
fixing transport_api.c:484 assertion
just update quotas and do not switch addresses
udp broadcast
<none>
UDP broadcast finished but not really tested
fixing https://www.gnunet.org/bugs/view.php?id=1867
<none>
<none>
+ prevent bug https://www.gnunet.org/bugs/view.php?id=1868 + changes
shutdown order to send DISCONNECT Messages
broke the build
disable local addresses
remov break_op
transport template defaults
do not validate own hello
<none>
improved and centralized helper path detection
fixing assertion tcp_plugin:976
additional checks due to mantis 0001868
starting dummy seems to work, so we can disable output
<none>
extending ats api to inform about addresses in use
fixing mantis 1878
workaround for: https://gnunet.org/bugs/view.php?id=1882
https://gnunet.org/bugs/view.php?id=1884
new test to test bi-directional connection attempts
<none>
transport notifies about addresses in use
dummy for address in use first I have to find the bugs in scheduling api,
then I can add the logic here
trying to fix https://ng.gnunet.org/bugs/view.php?id=1888
remove output
ATS suggests sessions that we in between terminated from transport This
causes:
cancel address suggests
cancel suggests
implementing fast_reconnect state
bidirectional connects
checking quotas
receive set_quota messages
removing quota test output
handle PONG only if expected
latency measurement with neighbour_keep_alive_task
<none>
this break is not needed anymore, since we measure latency in neighbours
improved fast_reconnect improved continuation management (no peer
references)
mem leak
+ moved outbound quota setting to separate function + fixed validation
error: race condition between address switching and disconnect
<none>
session pointers can rarely be identical due to timing issues
just return on invalid session
update address on receive
ATS address updatign and deletion
loaded the same plugin multiple times instead of multiple plugins
fixing memory leak #1895
<none>
<none>
related to mantis 1905
debug messages need to print correct address format
<none>
<none>
<none>
<none>
check for natted peer address
neighbour calls continuation on SYSERR
<none>
clreaning up conf files
<none>
fixing mantis #1906
transport-testing uses precomputed hostkeys to improve speed especially
on slow systems
fix for mantis 1897
<none>
speed up reliability
<none>
api change for network interface iterations to support broadcast address
and network mask
send broadcasts on every interface
remove log message
pritpals key
port was missing, udp broadcast for ipv4 is now complete
udp now supports broadcasts using ipv6 multicast
distinguish between beacons
<none>
now using default socket and sending on default port
removing broadcast port
different plugins with and session==0
transport testing: peer restart functionality
<none>
remember
do not validate own address
fixing reconnect bug
fix for mantis 1928 in combination with rev 18254
tests for mantis #1028
fixing unsupported messages in core
Adding UDP IPv6 peer discovery support using the GNUnet IPv6 multicast
address with site-local scope FF05::13B assigned by IANA
fix shutdown order + disable broadcast in testcase
fixing bus error on sparc, mantis bug 1909
fixing mantis #1970 bug: ats only deleted active addresses, but should
delete all addresses with session removed and addrlen 0
- hunting a bug
first steps to transport_api cleanup compiles but no guarante to work!
api clean up next steps
- latest changes for refactoring: iterate sends disassembled hello-address
- finale commit for the api change
moving ats quotas to template files
some doxygen + refactoring of GST_validation_set_address_use
- doxygen
timestamp not supported on fedora core 8
adding GNUnet endian operations
-fix
testing cmd line tool
fixed docu
<none>
<none>
added hostkey generation functionality
space after backslash
basics for integration tests
added option to specify hostkey file
modified to create cfg with default without template
step by step
-more
added quiet mode just printing the value
test now checks if boths transports are connecting
first test ready
added core sessions
disconnect test
clique
including integration tests in checks
WAN/LAN detection
IPv4 check implemented TODO: IPv6
WAN/LAN detection Mantis 1991
- remove outpu
move code from service to api fix bug add test
LRN's patch argument order
fixing crash 0002007 adding network information to addresses
added ATS addresstype information to unix
address type in WLAN
changes: changed order of startup since ats is now required for plugins
transport provides ATS handles for plugins network detection for tcp
missing
<none>
removing ats functions from plugins, instead provide callback function
WAN/LAN for HTTP/S
adding detection for AF_UNIX
wan/lan detection in plugins
-fix
-const char
Python version required to run tests is >= 2.6
added python remark to hacking This line, and those below, will be
ignored --
improved peer management
peer get automatically stopped
added improved check management
improved test framework
improved tests
added support for interupting a test
<none>
new test: connect after peer restart?
improved statisc output
adding nat clique test
fix
improvements and fixes
<none>
checked mantis #0002016 and added additional assertion
additional network type for WLAN
added LRN's patch from mantis bug 1998
fixes for mantis #1988 and cleanup
<none>
fixing some timeouts
- only when VERBOSE
a new nat disconnect test
- mauricios ssh keys
first changes for new plugin api
Improvement in reconnect: first disconnect, then destroy neighbours
fix for mantis #2008
fix for mantis 1959 compare addresses and only mark address when
addresses match
- remove whitespace
fixed assertion: wrong return value
fixed assertion in gnunet-service resolver improved address printing for
IPv6 addresses, especially if reverse lookup is not successful
- minor change
fix compile error
include plugin in gnunet-transport output
missing file
-compile error on freebsd
added section to solve vpn conflicts
fix: wrong binary and duplicate unix path
adding glpk to the makefile
mlp configuration
fix compile errors
- fix docu
- fix wrong documentation
- fundamentals for mlp implementation
-fix
- mlp
- fix
- more mlp
- even more mlp
- more mlp
- adding mlp functions to addresses
- fix vpn sock error messages
- more
- more ats
- added missing file
- improved multi instance functionality problem creation
- fix
- fix test
- configuration for mlp coefficients
- implemented: quality metrics
- minor
- more mlp
- removing not existing files
- improved error logging
- changes
- adding constraint handling
- constraint 1 & 3
- constraint min conne
- fix mem leak
- functionality for "for all peers" constraints
- more
- more changes
- changes
- disabling code : debugging was required
- problem matrix index starts with 1 not 0
- using DBL_MAX as VERY_BIG_VALUE makes problem unscalable
- fixed problem with b_min double conversion BIG_M is now UINT32_MAX
since assigned bandwidth cannot be bigger
- basic constraints working correct
- more working constraints
- latest changes
- more code
- quality updates
- fixing compile error
- use dummy msg type intended for testing
fixing: mantis 0002098: transport did not propagate session to ats
fixing mantis 2098: ats did not lookup addresses correctly ats overwrote
existing session when updating addresses
coverity 10048
coverity 10054
- clang
clang: mem access if plugin_name_length == 0
- clangs
- clang
- clang
- added martins keys
- new get_session functions
- more new code
- test code
- missing
- cppcheck
cppcheck
implemented session based sending in transport service (coexisting with
old code)
- more changes
fixing mantis 2101
- session sending in tcp
- fix
- changes
- changes in udp
-first half of write select implementation
fix memory leaks
complete select write implementation
removing retry code removing unused structs removing UDP Address structs
fixed PrettyPrinter (printed UDP addresses???)
implemented sessions
- debugging msgs
- fix
- debug msgs
- new udp implementation: start and stop
- changes
- changes
- more changes
- sending and getting sessions
variable message size
- latest changes
fix assertion
fixing const api and add check to address
adding verbose message
ifconfig parsing Shum's patch bugs fixed in patch: - IPv4 loopback
address was not included: added line 179 - stack allocated strings were not
zeroed out, so last value was used if value was not included in current line -
IPv4 netmask was passed as broadcast address (patch line 81) - IPv4 netmask was
passed as IPv6 netmask, caused invalid address conversion (patch line 113)
- enable output to test on Os x
- fix
session based sending for wlan
- implemented andtested timeout mechanism
- clean up the receive switch case
- rename
- cosnt
- changes
added assertion (which is successfully failing ;-) )
- fixed 2128: Blacklisting makes transport service crash
- basic bl test
fix segfault: api never saved callback and callback_cls
fix 2 crashes in blacklisting api - client transmit handle was not sent
to NULL after sending - BlacklistMessage was not checked for NULL
improved blacklisting test
fixing and testing blacklisting api and service
- working with fragmentation now fine tuning
- minor
- minor
- changes
- just a comment to clearify
removing backup ATS code
adding rewritten udp plugin
removing old send code from neighbours
fix access before null check
- fix for freebsd
- minor fix
new sending in validation
removing legacy send functions from plugins and renaming new send function
- fixing coverity errrors
changes in includes
- doxygen errors
- coverity errors
- remove ERROR
- remove log msg
- fix logging to remove error messages
- remove vpn sock warnings
- nicer error msg
- fix
- improved logging
- unused variable
- fix EAGAIN issue
fix for Assertion failed at gnunet-service-ats_addresses.c:587
- cleanup while debugging
- fixing assertion in gnunet_service_ats_addresses
- clean up pending nat sessions
fix test for peer restart
fix for fast reconnect: send ack after fast reconnect even when we are
already connected since other peer waits for it
- fix for Illegal state transition from `S_FAST_RECONNECT' to
`S_CONNECT_RECV' in line 2594
- changes to integration
- peer id in statistics
- pexpect
- new test
- fix for FAST_RECONNECT related timeout issue
- test
- logging, testing and result extraction
- changes to ats
- removing break used for debugging
- changes to ats
- changes
- choosing BIG_M and default quota is crucial to get a solvable
constraint system (hours of debugging!)
- latest changes
- latest changes
- fixed definition
- basic files for namestore
- remve whitespace
- adding namestore api test
- some func
- basics for the api
- missing file
- fix
- missing file
- changes
- changes
- changess
- fix
- fix
fix coverity 10138
- coverity "issues"
- separated sending for ipv4 and v6
- fix
- plugin (un)loading
- more code
fix to use correct queue
fix for mantis bug 0002154: change order of calls: - discard unsend
messages - call transmit send continuation - call session_end
averaging fast changing quality values
fix: ats suggested address for unknown plugin
fix to the 100% CPU load problem
- new test
splitted ipv4 and ipv6 socket select scheduling removed looping for write
select improved ipv4/v6 en/disabling session management
improved rescheduling improved recv error handling
not an error: plugin can return NULL (e.g. for IPv6 address when does not
support IPv6)
- tests
- operation id for api
- changes
- changes
- communication
- missing file
- lookup API send recv
- disable chat
- fix
- coverity
- fix doxygen
adding error messages
- put record
- improved nat handling
- correct port
- removing old nat server code
- more communication
- more code
enabling libglpk detection
- send START first
- more zone iteration code
- more zone iteration
- more
- record serialization + test
- using de/serialization functionality
- more changes
- put test
- lookup test
- changes
nametore api change
- adding sig purpose and test for lookup up
- checks
- run tests
serialize privat key
- namestore
- and of course a missing file
- message renaming
- comments
- changes
- at least compiling
- memleak
- test pass (not on sparc)
- removing complete
- more changes: sign & verify working and tested
- change signing func
- added signature check
- more changes
- memfix
- fixes
- added signature check
- namestore.conf not installed
- fixes
- reset db before test
- create test
- LRN's patch 2191
- use record_cmp
- LRN's patch: return value of nat client
- LRN's patch
- fix for mantis 2189
fix for mantis 2189
<none>
- add
- expiration update
- return value
- fix for 2192
- lookup specific types
- doxygen
- changes
- changes
- segfault
- reverting the change: we need the hash code to do the dht lookup
- step 1) compiles again
- fixes for put and lookup
- fixes
- fix
- checkin nbo conversion
- fix unused variables
- zone iteration stuff
- remove dbg message
- complete test
- fix in api - added second zone to test
- final zone iteration code
- hostkey2
- extradist
- docu
namestore api change: include block expiration time in record create
- reverting api change
- fix remomve callboack order
- not required anymore
- fix
- blame martin
- coverity issues... not bugs... but coverity didn't like my assumptions
of memcpy length
- coverity
- mem leak
namestore manages zonekey files with private keys
- mem leak
- extradist and hashmap
- outbound signature creation
- remove valgrind
- additional hostkey for pubkey generation that test are not authoriative
for
- fixed lookup
- changes
- fix
- additional test
- signing during zone iteration
- mod test_case
- fix
- changes to signing verfifying: includes block expiration
- make gns compile
- unused
- use logskip
- file extension
- adding dir
- adding assertion for name
- fix
- additional check
- fix
- new program_run and run_2
- unixpathes
- improve documentation
- fixes
fix segfault
- fix
<none>
- doxygen fixes
- usage of short hashes
- tests pass
- includes
- new zonekey files
- test use new zonekey files
- minor changes
- fixed record update for unique records
- old value was only 8h
- changes to record remove
- remove name if no rec is left
<none>
- declaration
- null check
- fix
- coverity
- adding LL
- dozygen
- code deduplication
- dedup
- dedup
- adding LEHO record type
- LEHO record type
- rewritten zone iteration code for code deduplication and improved flag
handling - added mode for gnunet-namestore to list public records and to set
record to public during creation default: list all records, create private
record
- fixing offset
- added "never for expiration"
- mantis 2207 + docu
- implementented string to address
- udp error message
- remove err msg
- debugging code for mantis 2216
- error msg
- log from ats
- throtteled fast reconnect
<none>
- remove err msg - added error msg for mantis 2214
- assertion causing crash, see mantis 0002244
- set initially number of tcp sessions to 0
fixing 0002249: report only new addresses
- fix for debugging code
- fix ipv6 address format for tcp pretty printer
- remove http
- typos
- "fixing" mantis 0002211
- additional debug message for mantis 2214
- for mantis 2215
- changes to the scripts
- changes
- changes
- changes
- do no run the test_integration_connection_value.py
- configuration for a normal gnunet peer
- changing conf and script for tcp only
- changing log functions
- more changes
- new test: connect on restart without bootstrap server
- clean up keys
- connection_watchdog tool
- consistency check
- added check against statistics
- mod
- doxygen
- fix for session counting: nat_wait sessions decreased session counter
but did increase it
- rename
- rename
- fix
- added check for low level connections
- remove msg
- shorten log messages
- added plugin functionality
- switch to netstat for compatiblity
- send recv test
- fixes
- api docu
- docu
- cmd line arg
- fix
- fix
- fixed string to address parsing
- off by one
- wrong NBO conversion: port is always zero!
- should be unsigned int
- fix
- confs
- bug
- improved configuration and statistics handling
- do not parse back files .conf~ but only files with extension .conf
- changes
- tmp commit
- update
- mem leak and changes
- changes
- changes
- latest changes
- doxygen
- hunting a bug
- fix
- rm debug code
fixing bug 2272: added functionality for watch to notifz about fresh
created entries with value 0
- fixes
fix 2273
- remove statistics use return values
- improve
- fixes
- added benchmarking for updates
- fix
- doxygen
- fixing mantis 2282
- fix for 2262
- double free
- assertions and checks for mantis 2283
- minor fixes while checking implementation
- transport receives client_start messages after all subsystems were
shutdown ... setting option to NONE
- compile error
- coverity bugs
- fix for mantis 2273
- fix for http
- fix
- continous watch for statistics
- fix
- from time to time core likes to monologize
- some statistics for the plugin
- some stats
- fixes
- invalid read when socket is already bound
- new confs for unix
- section for unix
- fix
-stat counter for http
- fix
working string toaddress
- fix
- missing PLUGIN section for tcp tests
- default unix ports
- breaks not required anymore
- fix for bug #0002291
- minor
- refactored sending code
fix 2297
- fix
- fixed tcp_string_to_address
- fix
- improved rescheduling
- unused var
- conf fixes
- fix
- coverity
- coverity
- fix for 0002248
fix for mantis 2320
- for make dist
- for make dist
- workaround for mantis 0002325
- fix for 601 assertion
- disable unix
- cleanup ports
- braodcast fix
- fix crashes
- memleak
fix for mantis 2330#c5818 GNUNET_SERVER_notification_context_destroy does
not cancel transmit_ready
- fixes
- fix memory leak mantis #0002334
- unused var
- fix for 2336
- added additional assertions
- confs
- memory leak when fragmented message is not acked
- more statistics
fixing WLAN
- fix crash
- fixing stat values
- msg
- increase timeout for sparcbot
- stats
- fix
- stats
- fix for rev 0002348
fix for mantis 2346
- only tcp
fix for mantis 2355
fix memleak
fix for mantis 2356
- improved fix for 2356
- fix for mantis 2350
- fix compile error
- compile error
- some debug messages
- fix typo
- fix for 0002364
- fixing double free mantis 0002363
- error msg
- suggestion throtteling as described in 2362
- suggest backoff reset
- changes
LRN's patch
- additional back off resets
- additional checks
- find invalid mem access
- fix for 0002371, 0002343: mst freed in callback
- adding return value to mst callback
- fix
- fix memory corruption, various mantis bugs: 0002374 0002368 0002367
0002369
- remove msg
<none>
session timeout for udp and tcp
- timeout for unix
- timeout for http/s
- LRN's patch
- missing
- missing
- fix
- preweekend fix
- fix
- fix for 0002383
- refactorind reverting last fix
- fix address length
- fixing memory corruption
- no rescheduling when mhd is scheduled to run asap
- remove msg
- adding timestamp for watch mode
speedup mechanism to manipulate gnunet time
- error message if fail
- base for a testing based connection stability test
remove unused variable
additional error message
- fix
- correct output
fix for 0002392
fix for 2395
- fix
monitor mode implemented correctly
- fix mem leak
- typos
core monitor mode
clean up
exclude notification about myself
- fix
- fix
- fix
- since session == NULL this is more clear
fix for segfault
fixing deletion of unknown addresses
- adding GNUNET_ATS_address_add functionality .. no changes to scheduling
API yet
- changes related to ATS_address_add + testcases for address deletion and
adding
- gnunet-ats cmd line tool basics
- changed behavior when receiving CONNECT messages ... inbound session
and NAT tests work again
- unused
- fix
- fixes
- fix
- compile error
- fix
-fix
- fix
- fix
- fix
- fix
- fixing assertion in scheduling_api.c:1222
- cleanup
- cleanup
new test deleting a session using the plugin_env_session_end way
(addr=NULL,addr_length=0) + session this fails atm and is a bug
- unused var
fix a major bug: delete session for unknown address (addr==NULL,
length==0)
- fix: addr is not NULLs
- fix memory leaks
- doxygen
- typo
clean up and documentation for namestor API
- coverity
- minor
- docu
- code cleanup round 2
- docu and cleanup
- commit often
- docu
- fix
- workaround for mantis 0002445
fix for mantis 0002462
fix for 0002463
- build order: first vpn then gns
- fixed session code
do not print error msg when we have no network connectivity
- mistake
improved error messages when no network connectivity and fancy interval
time
- typo
- change for mantis 0002445
fix for mantis 2445
- " undo"
fix for mantis 2445 & 2471
- some docu and new testing lib
- minor changes
- new test
workaround for mantis 2476
porting transport testing to new testing lib (mantis 2453)
- fix
- missing cases
- double checked all switch statements for missing cases
- change for 2477
- unused var
- link against math
- bug of the week
- fix for 2481 and 2472
- clean up testss
- compatibility code
- check did not fail over night with running peer, so it's not needed
anymore
- fix compile error
fixing assertion: cancel transmit ready on disconnect
- filter for expected message type
- additional checks
-m discard dummy messages
- fix confs
- updated confs
- can happen on disconnect
- fix
- implementation for mantis 0002485
- dozygen
- set start time
quota tests for wlan
- undo
undo
- fix segfault
- check if service is running
- improved session creation and listen only mode
- improved session creation and listen only mode
- improving client disconnect code to fix mantis 2486
improved server disconnect
- fixed result
- test for namestore
- fix
-m memory leak
- new rsa api
- new rsa api
- namestore uses new rsa api
- change
- cleanup
- changes
- some tests did not clean up
- issues with multiple services accessing one hostkey
memory leak
coverity 10104/3/2
- coverirt 10105
- fix
- coverirt 10116
- coverity 10080
- coverity 10055
- coverity
- coverity 10070
comment
transport fix
- fix compile error
fix
remove msg
callback for offer_hello
- fix
coverity
- coeverity
bug
new conf
change
improved url parsing
mod
- preprations for http plugin split
- test conf
key
libcurl check
changes
more
fix for https
changes
changes
changes
latest changes
changes
update template with latest api changes
changes to plugin test
<none>
- doc
plugin test
fix
fix
mem leak
improved security checks
more test
- fix compile errors
more code
mod
code
more
changes
update
changes to address notification
- fix
fix for wlan
changes
test
fix
docu
changes
code
<none>
more
change
more code
changes
fix
timeout
test conf updates
qutoa
asymmetric
tests
changes
changes
changes
changes
confs + coverity
changes
coverity 10270
coverity 10260
coverity 10259
not used
changes
- fix for unix
changes
- fix
changes
cleanup for http_client
makefile
cleanup
doxygen fix
config for new http/s plugins
external host
changes
a config file
- remove unused test
test porting
fix
- fix
test
- next test
fix
- more test
-next test
- fix memleaks
- next test
- next test
- memleaks
- next test
next one
- next test
- next
next
minor fixes
- some dbg
unaligned memory access
no valgrind
- fix
fix unload
shutdown
- finally the last test
fix
install testbed
use precomputed zonekeys
do not autostart: I have to copy zonefile
more
test working, but fails
- fix for 2nd cb
does not compile due to api changes ...
fix
- order
- msg
- fix
- bug hunting
- change
- change
- no valgrind
error message on windows
- mod
<none>
- mod
- fix
- memory consumption test
changes
- changes
- key
- log unix messages
- minor changes
- unaligned memory access
- bug
- formatting
- timeout parsing
- print key when put
demux everywhere option for CLI
client side PUT disconnect
- shutdown fix
- fix for connect
fix for "bad address " issue with wlan helper
- memory leak
-confs
- do not crash on shutdown
- refactored access_cb
- work in progress
- changes
- changes to http
- fix segfault mantis 0002550
- fixes for http reverse proxying
-output
strip external hostname prefix && improve url parsing output
move https cert location to servicehome
- url parsing
- compile error
- pausing, unpausing
- bug
no valgrind
- changes
- changes
prefix based plugin lookup for peerinfo tool
prefix based plugin lookup for transport
- working ... not needed anymore
- fix disconnect problem in put connection
- hunting bugs on buildbots
- hunting bugs on buildbots
changes
hunting bugs
hunting bugs
- removed debug output
- doxygen
dozygen
coverity bugs
coverity bugs
- coverity
changing type
- statistics values for all plugins
WLAN stats
UNIX stats
-fix
new statistics value to measure payload
rm old code
fix type issue
not required
undo
- fixes
-m coverity
documentation and preparation
changes
overhead reporting
fix for overhead measurement in unix,also fixes assertion in neighbours
1152
rm error msg
header
debugging
changes
stats + delay
changes
doxygen
overhead for dv
crash in test due to pending callbacks
fix
use precomputed hostkey for slow computers
improve code while debugging
clean up and fixes
minors
new statistic values and fixes
-remove
stat
-fix
new statistics
preliminary workaround for 0002549
- fix uninitialized value
coverity
- dangerous things for a friday: fragmented message timeout
- figure out http_client overhead
- measure overhead
doxygen errors
fix
- removal for fragmented messages
- fix
conf
timeout stats
- coverity
coverity
fix
basics
fix
- more
- changes but not complete
changes
gnunet-ats working and bug 0002593
report only active addresses used
gnunet-ats /w address printing support
move ats-tool due to transport dependency
add ats-tool to buildsystem
- remove cli
move to ats-tool
some docu
monitor mode
- docu gnunet-service-ats_addresses.h
- docu
doxygen
doxygen
workaround
use calculated delays /w wlan
fix unused variable
bug hunting
more changes
change
simplify retry mechanism
memory leak: is const
gnunet-statistics can retrieve stats from remote hosts without config
man
docu
- changes
unused code
error
fix
make check fails in root dir
coverity bugs 10360 && 10361
fix memory leak
bug hunting
hunting
- hunting
infocb can be NULL
- fix for mantis 0002618
make dist not working
doxygen error
docu
not a hard error, print a warning
fix
last cb NULL
- change peer preferences via cmd line
fix mem leak mantis 0002635
- clean up gnunet-transport (not finished)
not working yet
- changes
- changes
- working: -i -e, todo: -c, -m -s -b
- working: -i -e -C , todo: -m -s -b
changes
- gnunet-transport improvement done
-man page
fixed -b semantics: receive from all peers
changes to transport api
API change
docu
- "missing
implement mantis 0002419
doxygen
improve unix timeout behaviour
hunting bugs
use api correctly
fix mantis 0002653
- bug hunting
fix
- fix for mantis 2639
prefix support for testing
implementing mantis 0002193
-change to normalization
print ats information
print ats properties
- checking for resolver when trying to connect to remote host
transmit all ats properties
logging
omar ssh key
- verbosity
- preparation for list addresses changes
- changes to api (not finished)
- changes
changes ... more TBD
-changes
-changes
- changes
- changes
changes, almost done
dogygen
- first test for perf api test
- bug fix
- fix
- next stage
- stage 3
stage 3
test done
- fix
bug fixing: cannot drop messages
fix uninitialized mem
changes
changes
changes
- more changeS
- more changes
- move sysmon to experimental
add a test for glibtop
test
changes
compile error
network test
possbility a) I did not commit or b) some removed my commit?! Please
complain about it
type issues
installation test
- includes
- fix warning
- fixing 0002674
- mem leak debugging: monitor allocations in udp
changes
changes
changes
doxygen
- mem debug
changes
restructure
mem leak
changes
changes
docu
changes
doxygen
changes
fix for tests
- init test
commit required ... things can be broken
fix tests
some docu
remove unusued code
remove unusued code
changes
callback check
test rewrite
test rewrite
fix and new test
changes
changes
changes
minor
fix
fix
changes
changes
changes
changes
improvements
ats info
changes
tests
changes
remove old code
increase blocking on suggest in addresses, checking in solver
change
mod
mod
changes
- fix for session check and clean up
mem leak
mistake
mem leak in line 193configuration.c
one more time
fix
- no valgrind
- init
fix
fix
mgmt
more changes
clean up
coverity
coverity
remove hattrick check
test and change
changes
changes
rename
changes
- changes
changes
new test
- deletion
<none>
changes
- change
changes
- changes
changes
FIX: add quotas for all networks, transport tests should pass
msgs
adding function
fix break
fix coverity 10390
fixing confs: core tests should pass
fix coverity 10389
msg
add additional stage due to issues on powerbot: initial address
suggestion can take quite long during service startup
changes
fix: tcp did not return network type
updating addresses support changing networks
changes and test
- doxygen
changes
test: check network type
simplistic solver supports moving addresses between networks, test renamed
- centralized inc/dec
statistics for solver
new switch -q to print configured quotas
improved quota loading
docu
updated man page
fixes for make dist
doxygen
doxyen
doxygen
check for glib
glib check
new check
libgtop check
adding client reference
no valgrind
changes
- new test
unused code
- format
- preference mgmt
- format
- missing
fixing test
- doxygen
- wlan test quota below minimum bw
- improved minimum bandwidth check
- new test
- docu
- change
- change
- bew test: suggest alternative address if no bandwidth is available on
network change
preference value calculation
- change
- preference calc improved but not active for release
- remove url parsing breaks for invalid requests
- docu
fixing test by debugging (but still fails)
10 sec was too optimistic for mamasparc
- fix: SOCKTYPE not defined
debug code for mantis 0002726
enable preference calculation to finish it and debug
commit for mantis: 0002729 quota load: support "unlimited", fancy strings
and integer values (set by gnunet setup)
docu
global calculation
change to test to let updates settle
changes
changes
fix test
fix for broken tests
changes
fix for quota check
no suggest to cancel
no pending request
adding suggest handle to allow transport to check if we have pending
address request
fixing mantis 0002732
handle
changes
changes
changes
gtop integration
get more information
fixing 0002735: precision error when calculating quotas
install withouth libgtop
fix
fix
changes
changes
IPv6 support
time stamp
changes
fix for failing fs tests
timeout for address listing
improved timeout handling for address listing
end when no addresses returned
address parsing
test
address parsing
change
mem leak
changes
changes
print network type
changes
preference aging + improved quota recalculation due to pref change
improving test
fixes
changes to aging
print dmesg messages with human readable timestamps useful to check when
segfaults etc happened
wrong type causing invalid reads
changes
test
forgot
check code for 0002774
fix for 0002723
implemented 0002773
check for getrlimit
throttling validations
limit connection for tcp
transport metric manipulation
more metric
more
last commit for metric
delay sending implemented
traffic manipulation: receive delay
fixes
new test
docu
extradist -This line, and those below, will be ignored--
documentation
hotfix for curl vulnerability: libcurl SASL buffer overflow vulnerability
CVE-2013-0249
fix for 0002797
new config file
change in test
active test
modifications
mod
mod test
docu
change to WARNING
remove static function
docu
more docu
docu
more docu
docu
docu
docu
docu
more docu
auto generation note
changes
break
memory corruption
does not terminate on invalid uri
docu
docu
docu
test
refactoring and renaming
changes
documentation
changes
changes
changes
changes
changes
clean up of MLP_Handle
changes
refactoring
changes
result propagation
auto_solve
fix test
implementing restructured problem creation
changes
create set_value function
changes
simplyfiying code
simplify
create constrain function
modifications
changes
fixes
fixes
code cleanup + api extension
old code
changes
memory leak
changes to perf
fix for test
perf
performance analysis
changes
changes
new test
changes
quality update
quality update
switching networks
changes
implementing pong singature caching
validation fast start and throttling
error msg
disable message
changes
print warning only on failure
writing solutions and MPS dumps
test should not dump
changes
address specific names
fix cast
updating addresses
implemented ats info updates + initial address update (fixing issue of
execution time peaks)
perf test done /w update
disable output
doxygen
doxygen
changes to manipulation
manipulation recv side
changes
unused
clean up
clean up
unused
transport plugin api change: split of address receive and ats updates
dv plugin
performance monitor API + rename
test
pre lunch commit
commit
remove break until fixeremove break until fixedd
changes
changes
changes
changes
fixed delay manipulation
mod
changes
initial cb
test
return
ats manipulation
ats manipulation
removing ats
ats related changes in transport
ats related changes in core
changes
fix test
changes to connect_cb
fix
fix
api fix
change to core pai
fix core api changes
api change
fix core api
fixes
fix
fs using ats data
fix
fix
failing test
more check
fix
fixing ATS data
ftb
doxygen
doxygen
doxygen
ATS manipulation
ats api change
fix
unused
rename test
fix if cb == NULL
preliminary change
changes
test
test
fix
api change
fix
transport cfg based blacklisting
missing test
confs
modified test
fixes to blacklisting
additional test
fix
docu
docu
more docu
new friend only HELLO type GNUNET_MESSAGE_TYPE_FRIEND_HELLO
change type
support for FRIEND_HELLO messages
support for FRIEND_HELLO
some documentation
documentation
docu
docue
breaks tests
changes
fixed mlp
test
api change, fix and improvements
fix
delta support for disassemble
fixed network switching
ats change, done
docu
docu
docu fix
generic ats manipulation
changes
generic ats support
initial server side change for MITM ssl validation
uninitialized value causing segfault
use after free
changes for mantis 0002676
peerinfo API change + changes in code using api
implement peerinfo notification api
peerinfo get implementation
test
c tutorial
wrong struct
fixed client notifications
wrong log type
test
fix
print hello friend type
add uri info
fix
refresh hello fix
doxygen
perf test
doxygen
not required
fixing invalid hostkeys
additional hostkey
test fix for integration tests
fs crashing in integraton tests added check
fixed cleanup
fix
changes
cleanup
fix
changes
change
fix
fix
remove sysmon and disable libgtop check since it is not working on
planetlab anyway
disable verbose
additional struct
initial changes to support 2 hellos
store 2 hellos
work in progress
fixes
testing hellos shipped
multiple major fixes: - shipped hellos were never parsed - when hello
files was parsed from certain file, hello from this was ignored
fix
fix mem leak and counting valid addresses
fix
fix
fixed periodic cleanup
fixed friend flag
simplified hostfile reading
dump hello to file
changes
doxygen
fix?
Added a valid HELLO for my machine - peerinfo test should pass now
remove invalid HELLOs
new hello
surpress core message " Unsupported message of type 16"
prevent hostlist connects
uninitialized
changes thx to SAP
traffic manipulation cfg settings
fix for 0002881 + test
added warning
documentation
missing test
bound (IPv6) socket to wrong (IPv4) address
new ssh key
compile error in udp additional debug messages for transport connections
fix for manipulation, machine horribly died while testing
fixing udp bindto
improved messages
changes to binding port
test using autodetection
docu
test failing on pi
remove memory debug code
fixing memory leaks in udp
returning here is not a good idea since: - IPv6 socket can be opened but
will be not closed if IPv4 error is handled as hard error - Plugin will return
correctly with IPv6 socket open but never scheduled
fix
debug msg
already defined on OSX
more dbg msg
improved test
fix return value
adding experimentation daemon
testcase to connect daemons in a clique
execute ldconfig
request management
changes
manage requests
added statistics support
modifications
changes
message types
split up code to multiple files
documentation
use GSE_stat as extern
adding capabilities
remove warn message and add documentation
fix for 0002890
fix
one '&' to break it all...
changes
no comma
check experimental
capability bit shifting
changes
issuer
parsing
experiments
test experiments
extended test
pubkey
additional check
documentation
docu
docu
docu
new components
experimentation timing
scheduling
changes``
changes
changes
changes
no valgrind
changes
changes
reflow and docu
rename and cleanup due to writing docu
restructure
changes
rename
changes
docu
changes
doxygen
doxygen
changes
rename function
rename
adding generic normalization support for all solvers
changes
proportional solver uses normalizationy
<none>
changes
changes
changes
doxygen fixes
remove normalization from solvers, use in address only
renaming mlp solver for consistency
renamed
changes to test
missing file
fixes
fix
fix for crashing perf_ats_mlp
normalization docu
doxygen fixes
basics for quality normalization
Updating the tutorial section in the source and including an updated
version of the pdf about command line parsing.
- fixed info message in test - added data structure for property
normalization - added property averaging
quality normalization without checking if min and max are given
update for plugin template
implemented ats property normalization with callback to address to do:
recalculate min/max to prevent outdated max values have an impact on
normalization
fix testname fix array out of bounds for test addresses
fixed test
completed normalizing with min max handlin
fixing coverity 10444
fixing coverity 10470: Uninitialized scalar variable
workaround for coverty 10451: Unchecked return value
fixing coverity 10464: Resource leak
fixing coverity 10465: Resource leak
fixing coverity 10466: Resource leak
fixing coverity 10467 Resource leak
fix for 10472 Structurally dead code
reenable address suggestion in test
fixing coverity 10474 Unused pointer value
renaming properties and adding function to print properties
disable unavailable test
updated test
added option field in tcp plugin + fixed bug in port map callback: stack
allocated variables not zeroed out -> addresses not comparable!
added options field to addresses fixed stack allocation problem when
converting string to address
added options in header
fix crash
enabling certification verification in test adding cfg options for
certificate verification
remove log message
Implementing bug 0002910 - extended plugin to contain option field in
addresses Implementing bug 0002677 - adding ssl verification for external
hostname having - the server send the HTTP_OPTIONS_VERIFY_CERTIFICATE option
- client use this option to enable ssl verification
(CURLOPT_SSL_VERIFYHOST, 1L) is not suported anymore, have to use 2L
fix printing
printing fix
new cfg option to include port in external hostname
added: port can be automatically included in external hostname required
for testing when ports are modified automatically
options support for unix
fixing memory leak: unused hashmap fixing segfault by adding NULL check
added option field for WLAN addresses
updating template with address options
fixes in tests: scheduling, removing connect request:
doxygen fixes
doxygen fixes
fix address printing
fixing extradist in makefile
fixing size check in address to string
fixing size check in get_session
fixing stack corruption due to buffer overflow
fix: memory leak
fix: memory leak
fix failing tests: returned wrong address size
fix: - empty string if no url given - print port as short
fixed: invalid network scope for external hostname
fix: do not assert if plugin sends you invalid network type
fix network type for external hostnames in http_server
adding bluetooth network type
adding functions for bulkmode
added support for bulk operations
finished buld support
fixing log level
remove unused fields
function to get normalized properties in solver
changing solver api: remove address hashmap from functions and pass
instead with init
no valgrind
proportional solver uses normalized properties
removing breaks and output
new function to update a value in the matrix
name now reflects function
correct order of address deletion: remove, notify, free
multiple fixes: - read cfg option to dump problem and solution - do not
resolve if no requests or addresses - fixed address deletion - implemented
function to update single value - fixed preference changes - preference changes
now use normalized values
major change to solver api: split _update function since it combined 3
different functionalities proportional almost done, mlp to do
split up update function and using normalized values
updating tests to match api
removing debugging break
updating now works with bulk lock
documentation for new solver api in header
documentation + remove unused code
improved network switching for ats
improved value updating for unkown values
doxygen fix
doxygen documentation
more doxygen documentation
removing debugging breaks
fix segfault on missing type
fixing ats-tool output
fixing output + additional peer id check
do not use hellos included in gnunet with testcases
better just a warning
functionality to disable shipped hellos
new function for plugins to signal inbound session, have to extend plugins
if io is enabled and shipped hellos not, hellos from peers we created
should still be read
missing file
fixed adding addresses with updated networks
do not add or update if sessions are equal
new api function to get network for session
added function
get function implementation in all plugins
this add should not be required anymore
adding addresses with network
changed return type to enum added get_network to template
check implementation of api while loading plugn
code deduplication: move to function
move add function directly to transport
renaming function
including expiration time in output
add default operation when no argument is given
fix: infinite loop if option was given
modified validation to use new api call to set networktype
move error reporting to client, since now it's the plugins fault if
sessions are unknown
unkown session reporting
report session
session report in http server
fix for header
fixing uninitialized values
remove ,
fix for printing
fix:
fixed: - printing inbound connections - lookup for existing sessions -
notifying transport with session_start only for inbound sessions
added bt quota
fixed: printing of inbound session size of inbound sessions
fix: wrong address length
notification for inbound sessions
fixed: inbound address size reporting inbound sessions
adding debug message
printing inbound addresses
removing general address adding
removing unwanted break
log level
new function to check if session is known to ats required for two
reasons: we cannot update metrics for validation sessions since these are not
known to ats a remote peer can decide to use an inbound validation session to
transmit data, so we have to add the session to ats if we receive payload on a
session
notify ATS about a session (used for for validation) if we receive
payload from a remote peer on this session
improved network switching
enabling address check in validation
verified and improved address checking for external hostname and verify
option
fixed address validation for plugins with client server architecture
disabling error level message, adding stats instead
fix state machine
substring search is the root of all evil: peerinfo did load wrong plugin
if substring matches: used https for http address
automatically replace all server addresses with client addresses
fix if plugin is loaded only in stub mode
fix
important check
unused variable
fixing crash
String to print inbound connections
using inbound string
template update
new hellos
do not use forever
while running transport: valgrind showed memory leak due to not removed
resolution processes storing pretty printer requests and removing them after
timeout
improved printing
udp returned addresses with 0-port
broken build
removing breaks, adding stats
bug fix
memory leak
const was discarded
uninitialized variable
RFC new gns publishing
updated code
changes for experimentation
docu and additional queue
refactoring since names are too long
renaming stats and cfg
message parsing on receive
changes to scheduler
new unified communication mechanism
fixes
fixes for list queues
fixed stat counters reduced logging
no valgrind
fixed doxygen documentation
fix for crash on buildbots
duplicate header
basics for generic ats benchmark
confs and makefile
add testbed disable all connections in conf
intermediate commit
variable number of peer
initialization with ats
changes to perf_ats
connecting
connects only with valgrind?!
peers are connecting
core connections
ready to start with benchmarkign implementation
core missing for one test
some output
error message instead of assertion
new hellos for fulcrum and gnunet9
added error message
this hello causes self-connects
invalid for gnunet9
more info
added warning for self connect
valid hellos for gnunet9 and fulcrum
improved messages
improved messages
debugging messages due to mantis bug 2980
fix for resolver timeout issues
remove output
number for peers
remove dead assignments
remove dead assignments
dead assignment
remove dead assignments
printing size to eliminate dead assignment
fix logical error to assure buf is never NULL
fix test result
changing to fancy string and from 3 seconds to 30 seconds
fixing mantis 3002
make test resistant againt duplicate notifications
fix peer ids
fixing blacklist: peer ids for blacklisting have changed due to crypto
change, so blacklisting did not work
fixing bugs found by clang
fixing clang reports
fix clang report
more clang fixes
check in wrong line
test compiles again, but fail to get the stored record: in
sqlite:get_record_and_call_iterator:
remove docu
fixed crash: wrong cls in api
switched test to new api
log level
missing check
fixing memory leak caused by key derivation
reorder tests: namestore_common -> plugins -> api block test
applying new speak: proportional
supporting multi master/slave approach
serialized startup
latest changes to support multi master
fix typo
fix for all masters
simplify to make reader and clang happy
remove debugging break
ported postgres plugin to new API to get it to compile functionality was
not adapted
test if records are removed correctly
cannot link memrchr on OS X
planned end of test
basic communications
communication done
safety checks for arguments
fixed core configs
how did this test ever work without assigning core handles?
load balancing for partner selection
partner task
quotas + partner evaluation
missing config file
missing cfg file
fixing integration tests
fix
cleanup service home directories since left overs can break the test
fix second peer for clique test
more changes to configs
implemented feedback API, but not yet calling solver
rename to match api name
adding message type
ATS service calls solver with feedback
updated experimentation tests added scope time for feedback api
integration configuration cleanup
h_addr is a define in in netdb.h leads to stack corruption -> rename
variable
mhd is failing to bind both IPv4 and IPv6 -> restructured http plugin to
no fail if one fails
fix: wrong check
order of daemons
no debugging sleep for buildbots required any more...
check if MHD_DUAL_STACK is defined
new test directory for tests depending on testbed
removing tests moved to ../ats-tests
changes to configure.ac and Makefile to enable ats-tests
missing makefile.am
checks for coverity
fixing coverity issue
renaming according to new speak
more renaming from simplistic to proportional
updated configuration files
renaming delay to latency to map to preference type added preference
usage in test
only solve on request: - if we have a new request - if we have an address
for the peer
fixing memory leak
- rename test according to what is tested - added additional checks
return NULL if rd_count is 0
improved test
updated zone iteration test
updated test to iterate specific zone
docu and Makefile
zone iteration stop
zone cannot be NULL
updated tests
additional test for monitoring: check if existing records are returned
new test: store a label and update it
tests not required anymore due to removed functionality
removing tests: not required anymore
new test
- iterator NULL check - do no empty name if it did not exist before
modified test
surpress error messages
disabling tests not yet passing
fixing coverity issues
improved logging: messages with higher importance changed to info
- bulk deletion for all addresses - no break, since solve can be called
when addresses is empty due to autosolve parameter
better logging since important messages are info
important messages are info ATS timeout was breaking tests /w valgrind
fixed bug in address iteration
remove not required assignments changes to log level
fixing assertion in ATS about unknown address in use:
active address was not set to inactive if session is deleted
fixes: resolve on canceling address request
address was no set to inactive
new propagation check reformating lines
warning if inbound session does not have address length 0
plugin returns inbound session if outbound session is requiered
clarified docu and log level for ATS disconnect notifications
define 0-bandwidth
fixes for proportional solver: - do not send 0-bandwidth on address
switch since it will disconnect peer - do disconnect peers if no address is
available
add disconnect message
address suggestion cannot be done while deleting: it will suggest the
same address
indentation: replace \t with 2 spaces fix: do not disconnect (send 0
bandwith) on address switch
disable multicast
fix: do not return inbound sessions when transport asks for outbound
session
indent: tab to spaces
log level
fixing indent
fix: upd has to use the session with fragmentation context when receiving
ACKs
fix: fail to compile
speed up address suggestions
indent
fix indent
improved ats performance tests to support parallel transmissions
new config files
new test
doku and indent
fix memory leaks move connect operations
initialize partners during creation
cleanup logging blacklist sam and fulcrum from connecting
print progress
improved progress remove master information struct add ats partner
adding ats preference setting
reduce log level
added transport level benchmarking support
remove unused variable added basic files for logging
mixed up .c and .h finished makefile
collecting log functionality mv structs to perf_ats header
basic log file writing functionality
log handling
logging for master
slave logging
slave data extraction
logging done
stop logging only when started
basic gnuplot script creation
gnunplot script generation
added duration cmd line argument
added logging frequency
use one scheduling function
measuring application layer delay
ats parsing
logging changes
logging application layer RTT
improved column indexing
new test
implemented ATS perf info logging + push logging issue: what to do if no
values changed
soem logging improvements
find partner function fixed crash during shutdown
logging on demand + log bandwidth
improved logging
bandwidth gnuplot
updated configuration file: - moved topology information to template -
blacklist peer IDs match hostkey
added template
move testbed setup to template
updated test
updating tests
single aging task + indentation
init test: ok
testing api without solver
add address test
statistic based test
changes for ini
modifications to test
indent
new test: check duplicate
next test with statistics
next test: destroy address twice
next test
additional tests
new performance tests
next test
do not use monitor callback
fixed segfault + improved aging
fixed for loop
use 10 seconds
add log message
fixing indentation
performance monitoring test fails atm
log message
additional check to prevent double free
race condition between APIs due to different locations for the
client_receive call: performance and scheduling now work the same
additional check
this iteration should return all addresses, not only active
new test
modified termination condition
new solver test
updated configuration files
minor changes
new solver specific test
next solver specific test
next test
docu
fixed test
indentation
fixed test
modifications to addresses and proportional solver
disconnect peer if active address is deleted and no alternative can be
suggested
indent
disconnect peer if no address is available after deletion
next test
added new test: has timing issue, still fails
addign test
do not suggest address twice
minor fixes
solver has to suggest address
fixing shutdown
indent + exclude peer
exclude requested peer from propagation since addresses will send result
log message
sort tests by solver and comment ril out until working
fix use after free crash on client disconnect
fix compiler warning
changing defaults to not fail on pi bot: less slaves, longer timeout
new namestore CLI tests
next test
caching test
debug on freebsd
fix unix plugin
fixed tests
fixing test
fix type
fixing test compile
blacklisting cfg files
fix warning
fix warning
fix warning
fix tests
*sigh* ...
fix
- fixing udp address handling for inbound sessions - fixing some
indentation
fix for incoming addresses: incoming addresses are only marked with
address length 0 if the address cannot be used to initiate a connection to his
address. Therefor only for tcp and http(s) incoming connections an address
length of 0 is reported whereas for other plugins (unix, udp, wlan) the
incoming address can be used to initiate a connection to this address.
fix for failing _init test We had a race condition: When ats is included
in default services it can happen that it is started before the test can issue
the statistics wathc request and therefore the test will not be notified about
the initialization of the value with 0
improved documentation
indentation and inmproved documentation for feedback
ats solver are now implemented as plugins
fixes to makefile
documentation and fix for const warning
renaming
documentation for solver api
applying LRN's patch to fix startup issue
fix type
fix indexing
fixes
fixes
fix resource leak and if0 unused function
fix more leaks
fixing resource leaks
fix leak
basic utilization implementation causes an assertion in ats_scheduling
when address is not yet known
- fix assertion in scheduling - formatting
doxygen
unwanted commit
doxygen
doxygen
doxygen
doxygen
clean up code and code for utilization
new property
adding
unused
<none>
<none>
payload and data
fixing crash
fix
implementing api call
adding initial statistics call
- clean protocol number - added lookup protocol numbers - implemented
lookup sending
added lookup test + fix in test
api communication done
extended plugin api to support lookup function added new index to both
plugins implemented lookup functionality in both plugins namestore uses lookup
function
replace iteration with lookup
replace zone iteration with record lookup
removing unused tests and clean up makefile
initial changes for generic solver performance test
next step for generic solver test
add confs to make dist
fixed
move definition
next steps to generic perf test
fix for quota loading
function to fill buffer with random values
fix peer creation
minor fixes and log level changes
improved message
fix assertion: address has to be removed from peermap before notifying
solver
clean up
basic benchmark functionality working
implemented solver information callbacks for benchmarking
clean up unused vars
improved benchmarking
gnuplot basics
writing data files
centralized function to start solution
adding additional INFO
update info for benchmark
benchmark with updates
doxygen
fix for lookup
- fix: all records were created as shadow record by default - new: --
printing record flags -- improved error message on deletion
new test to lookup shadow record
implementation of shadow record functionality: on decrypt iterate over
included records - if a shadow record is found, check if: -- a non expired,
non-shadow record for this record type exists: if yes: filter shadow record --
shadow record is expired if not: - remove shadow flag and include record
minor fixes + test for shadows record filtering and usage
block expiration
fix tests and do not assert since this will break make check
internal time logging
remove non existing tests
new HELLO for fulcrum
new hello for gnunet.org
new testing hostkeys
test for address to sock
new identities for blacklist checks
integration test keys
new config for solver
randomized network selection
extend info cb and documentation
adding notification to solver
randomization was off by one
fix and eval function
splitting up calculation and notification for proportional solver to
ensure measurement correctness
doxygen
don't forget the house keeping: fixing memory leak
functionality to print ids in hostkey file
fixing blacklist ids for dv
renaming PSEU to NICK gnunet-namestore can set nick name in namestore
adding pseudonym NICK automatically to records sets
header
remove private flag
use nick instead of PSEU lookup
one shorten call to analyze
fix for size
fix segfault if no name is given
renaming and modifications to the shorten test - renamed to nick - added
identity
remove
improved debug output
- fix message log level - fix merging - fix docu
delete remove namecache content
remove DHT lookup code - do not shorten shorten zone
doxygen
coverity fixes
fixing bugs found by coverity
allocation evaluation
remove, not used anymore
additional hosts are appearing -> disable hostlist
utilization
improved evalution
preference iteration
- reduce debug level in proportional - store pref client in DLL - remove
disconnected clients
notify only once
preference client management and test
get normalized pref per client
evaluation
renaming timeout constants
missing fi
- renaming namestore references to namecache - implementing nick record
based shortening in resolver
advanced tests later
remove debug output
timeout is to short on slower systems (and when running with valgrind)
minor fix in check ... test test_gns_nick_shorten now passes
rename namestore to namecache
wrong string used
additional check for filename
SET building is working now ... but SET itself seems to fail
remember direct route to be released on direct disconnect
fix crash
missign break
local revocation test
evalutation
test for revocation flooding
- neighbor_table_consensus was never created - do not include own or
remote peer in set - print set status
improved cleanup
- FIX: GNUNET_SET_STATUS_HALF_DONE is never called only
GNUNET_SET_STATUS_DONE - FIX: Crash since neighbor_table == NULL on removal
check
debug messages for DV plugin
clean up test
minor corrections to revocation service revocation testtest now passes
fixing test
minor output improvement
- Tests did not clean up: TEST_HOME with namestore db was not removed
after test Therefore test_namestore_api_remove_not_existing_record passed when
executed on a clean system, but failed if ran after a previous test adding a
record
check for cli tools
check for cli tools
add cli check
disabling for debug
fix div by zero
use test specific namestore db
separate from default and clean up
adding message
changes to config and test due to test directory
improved debug message
debugging
debug message
return error as exit code
- detect GNUnet installation - detect if GNUnet is running
path required
bug in zone to name: called with disconnect error instead of signalling
"no name found"
fixing multiple segfaults fixing functionality: misunderstanding of
monitor callback, stored pkey under zonekey of returned existing mapping
remove debug
fixing parallel build dependency
reduce log level to debug
- fixing crash during shutdown - adding network functionality
enabling transmission to plugin
addign network
check in proportional
adding libgnunetats.la to libadd
doxygen
modify test to test shortened delegations in shorten and master zone
include nick in cache for debugging
debug message
rename to match functionality
expiration
add cfg option to disable caching
add nick name for cached records
no cache cfg option
new test with PKEY shortening check and lookup in shorten zone
config file
correct order of quotas
doxygen + remove break
fix uninitialized memory in network send
fix for mantis #0003156
fixed public key and identity comparison in disconnect messages
set to NO_TASK when done
fix ats information on session start
fixing segfault since session->plugin was NULL
basic qr code reading
detect python interpreter error handling
added config + silent mode
silence: other way around
- test if identity service is running - default ego is used if none given
adding docu
fixed uri parsing and exit codes
gnunet-uri return handler exit code
calling gnunet-uri
not working
working again
revert to http
make message appear less often
renaming according to naming conventions
disable debugging break
fix error message when testing service and directory exists
document options provided
use a define as statistic string to be sure to set the same value
be more descriptive for debugging
shorter session timeout for UDP
remove debugging breaks
fail to start service if we have a malformed configuration
do not start with malformed configuration
fix for core key exchange
fixing documentation
this is not possible, since identity check is enforced by
transport_clients on try connect call
Destroy address and session if blacklist denies connection Address
suggestion requests are not need, if a request was scheduled before...
remove addresses we cannot connect to
comment
fix assertion in line 1388
reduce switch log level, increase ats timeout log level
accept all requests, reduce log level
fix doxygen errors in comments
clean up: remove unused and outdated files disabling tests to fix them
additional signal handlers for test
disabling services
if start fails, clean up and exit
restart test
disable more services
UNIX Path
IPv6 not available: second messsage is better
re-adding disconnect test
restructuring configs + enabling additional plugins
disable udp broadcasting
remove unused config files
clique test removed unusued confs, tests
clean up for configs
a nat peer is included in clique test
unused test
unused
fix
sleep to have hostlist started
do not reschedule timeout on sending
disconnect test
disconnect test
fixing 0003221: Crash in UDP broadcast:
plugin_transport_udp_broadcasting.c:253.
fix for fix
check for keep alive function while loading plugin
new api function for timeouts
new timeout function
new timeout function and keep alives with nonces
fixing name to install
clean up test after execution
rescheduling sessions for udp
fix for plugin
fix for bt plugin and set nonce to 0
fixing test config
session timeouts for tcp
session timeouts for unix
session time out for http client/server
new timeout tests for WLAN and bluetooth session timeout for WLAN
test mode for bluetooth
tests and session timeout for bt
disable debugging messages
idea: try to prefer 'newer' addresses to improve mobility issues adding
time of add and time of last activity information, to allow to choose 'newer'
address
delay start of client to allow the server to start on slow systems
added function to disable receiving broadcasts to prevent buildbots from
connecting to the network
delay between starting clients
trying to analyze bug 0003218
add session to message
related to 3216
preventing receiving in disconnect
major bug: not converting NBO to HBO
fix relative time calculation
success message
at least they connect on transport level
fix indent
fixing file indent
new monitoring API
implementing monitoring api
neighbour monitoring api implemented
missing configs
test + documentation
more docu
fixed return value
renaming states according to naming convenations
extending ATS api suggest callback with peer id adding suggest
continuation
consistent documentation
docu
naming conventions for peer states
implementing 0003268
implementing 0003268
implementing 0003268 to inbound information in HELLO addresses All
transport plugin functions are modified to use HELLO addresses instead of
peer,address,address_length All plugins are modified to use HELLO addresses
internally
set default options
fix for printing address
increase log level for useful messages
improved log messages
additional function to check option
improved checking for direction
missing option assignment
copy info
fix for creating session
fixing issues with udp due to internal address length
fixing memory leaks
issue: as long as we do not have a peer identity we cannot work with
HELLO_addresses, so we have to stick to sock_addr fixed sockaddr-to-session
lookup
fixing dv plugin
fix memory leak
simplification for tcp plugin
fix crash on shutdown
fixed wlan plugin
remove inbound outbound notion from wlan plugin
fix bt plugin
fix mantis bug 0003270: do not broadcast on loopback
header for transport validation monitoring API
transport validation monitoring API (not yet complete) + CLI + fix for
crash in transport/plugin_transport_udp.c
printing validation information with cli
fixes for reconnecting and message handling
new state
correct string
test
documentation for validation monitoring test case
documentation for ATS
fixing documentation for hello
documentation for transport service
more fixed documentation
documentation for callback
perform blacklist before address switch this commit adds the blacklist
check before address switch in the next step the old blacklist check has to be
removed
fix ftbfs on system where SIN_LEN exists
removing useless fixme comment
fix use after free and mantis 0003257
blacklist check on try connect
initial preliminary commit for ats testing lib and simulator tool
not yet finished: moving perf_ats to testing
fixes
fixes for ats measurement
improvements
fixes for makefile
fix for evaluation
changes to traffic generation
reverting pogen.sh
missing file
traffic generation
put traffic generation in separate file
experimentation basics
new logging functionality
fixing test case logging
implementing operations, episodes...
example experiment
enforcing operations
added check if service is running
fixes for doxygen
solving slow shutdown issue: statistic was not destroyed
added multiple iteration support for averaging
average calculation
logging average
separate update and full results
fix compile error
doxygen fixes
documentation for the performance benchmark
more documentation and minor cleanups
improved benchmark output
clean up configuration parameters and implementation
minor fix in averaging
fix: only include peers with requests and addresses available, otherwise
system has no solution due to min_peer restriction
renaming options + adding glpk verbose option
renaming options and status variables
modifying hostlist options to surpress warnings
enabling ril and mlp benchmarks
documentation
enable logging as a cmd line arg
a test experiment to send data at a rate of 10 kb/s to 3 slaves
fix: destionation was wrong
fix and improved logging
improved logging
writign plots
remove extension
extending bandwidth tracker api to support notifications
send receive delay rescheduling support
constant traffic generator
renaming types
adding linear and saw wave support for traffic generator
sinus support for traffic generator
documentation and random generator
additional check for sinus generator
experiment parsing + test.exp containing different examples
safety checks
various fixes
fixes
removing set_rate, using start_send instead
logging all nodes in a single file
updated experiment
adding preference generation
enforcing preferences
preference logging
docu
fixing tests ftbfs
fixing 3299: use after free caused by shutdown order neighbours & plugins
fix for 3300: remember to call client_receive_done for blacklist client
fixing compile warning: cast to types required in doxygen documentation
fix indentation
filter ats information in non-verbose mode
eval
solver based evaluation
unix domain sockets used hardcoded path instead of configuration setting
- removing UNIX domain sockets port support and UNIX domain socket NAT support
moved eval tool to ats
preference parsing
experiment loading complete
example experiment
preference generation implemented
property generation
cfg handle
fixing crash: excess task has to be stopped when neighbour is freed
abstract unix domain socket support
stop notifications on delete
enforcing correct generator type
logging task
fixing property generation
improved episode parsing added request support
adding and deleting address from solver
some additional information about quota
improved quota loading + requesting and stopping address suggestions
address management
commit before removing address ids
property enforcement
properties
+ setting preference + set preference example + adding client_id to
suport multiple applications
basic logging
Continue with address suggestion after try connect black list check
Related to moving blacklist check before address switching as described in
mantis #3262
fixing issues: - using uninitialized memory (mantis #3322) - notify only
about new session created
RIL is allowed to set bandwidth to zero
transport's idea of being notified about an explicitc disconnect using bw
0/0 and the modification of the ril solver to allocate 0 bandwidth do not work
together at the moment this is problem of semantic and not of implementation
reformating and adding additional log info about hello import
disabling hello import in configuration
improved messages
no set_state to DISCONNECT_FINISHED required before free_neighbour fixing
log output
fixed clean up and improved logging
Remove blacklist check after address switch: blacklist check is now
performed before switching
fast reconnect: new state is reconnect, not init this fixes mantis
#0003330: assertion at transport_api.c:618
updated tutorial with Omar's remarks and fixes
fixed configuration testing: NAT testing was not working
changes to connect behabviour and a lot iof improved debug messages
removing duplicate state transission
additional condition for logging
formatting
unsetting primary address on when free'ing address for monitoring
addtional check for failed CONNECT messages + some testing
remove testing breaks and add documentation for checks
additional log info
new session end function including address
session end function must include address to notify address
use address in session end callback
fixed check for destruction of inbound addresses
renaming generator type for ats addign comment for bandwidth
improved log message
doxygen fixes
when switching addresses in connect_sent: no blacklist check required,
switch directly and send new connect
-lgcrypt fails if gcrypt is not in lib search path: use variable to link
libcrypt
fix initialized value in udp on invalid address fix log message
fix compile warning
print message if service is not running
remove unsued INIT_BLACKLIST state: blacklist check is done before
switching
task was not marked as canceled when canceled fixing mantis #0003345
checking incoming connections major fixes: - do not disconnect neighbor
on CONNECT_SENT timeout, but: destroy address and expect new address from ATS -
fixes related to mantis #3263 when peers reconnect: when peers receive CONNECT
in PS_DISCONNECT, disconnect was completed but -- no address requested -- ack
timestamp not saved -- ack state was not set to expect a SESSION_ACK message ->
reconnect failed!
FIX: In ATS fast reconnect the timeout was not set correctly if reconnect
was successful and neighbour was disconnected in master task after 1 second
checking incoming connection behaviour: - removing blacklist checks for
incoming checks - adjusting blacklist check on connect to report to ATS -
incoming blacklist check states did not set timeouts - timeouts not set for
PS_CONNECT_RECV_BLACKLIST_INBOUND in handle_connect
Do blacklist checks on CONNECT before giving CONNECT to neighbours. If
peer is blacklisted we do not need to to anything, this simplifies the state
machine: If peer is blacklisted: CONNECT is not given to neighbours If address
is blacklisted: address is not given to ATS and will therefore not be suggested
So neighbour can use this information without additional blacklist checks
fix compile warning
removing blacklist state not required anymore
remove state
remove debug break
remove debugging breaks
removing GNUNET_TRANSPORT_PS_CONNECT_RECV_BLACKLIST state
remove CONNECT_SWITCHING_BLACKLIST state
we do not need this address in neighbours: was already added when CONNECT
was received in _transport.c before CONNECT was given to neighbours
error handling for failed CONNECT messages in CONNECT_SWITCHING
Address was not deleted if SEND_CONNECT in FAST RECONNECT failed
Therefore alternative inbound session was not suggested ... This was a hard
one...
some testing code for debugging
reformat
remove blacklist check in neighbours on CONNECT
blacklist checks have to be cancelled when session is terminated,
especially when blacklist clients are active
one break to break it all: disconnecting blacklist client was removed
from first blacklist check only
improved network lookup
We do not need the address and session here
added functionality to use the CLI to disconnect peers fixed DISCONNECT
functionality
API header with disconnect call
fixing crash: cancel dead task 1004493731513019834!
improved printing
fixing crash in unix
re-add address on connect
make plugin fail if bind fails ...
changes in ACK handling
cancel delayed disconnect task on free
changes to switch when connected
destroy session if connect fails
debuggging message for "send ACK but no session found" issue
no need to be an error
debug message for assertion to figure out malformed address length
fixed check for session_ack handling
fix duplicate connect notifications
fix for connection counter
minor changes for address conversion
doxygen docu
fixing configuration for failing test
ril may suggest 0, so this check is wrong
defragmentation context not freed when session is deleted
fixng crash on shutdown
fixing mantis #3362
remove break
obtain papi only once
operation timeout to short for pi
fix for coverity
remove unused value
fixing unchecked return values
fix for unchecked return value
removing unused variable und double free (with null check)
fixing resource leaks
fixing resource leak
unused call
fixing resource leak
fix bootstrap
fix
wrong checks
more information why nat tests fail
add address to list since we use local addresses OR it is not loobpack!
log level
debugging break
upnp support just silently failed if upnp client is not available
handle nat_register fail
fixing error message
adding extended proxy support for http(s) client
extended proxy support for hostlist
change debug level
using proxy settings
advanced logging
improved logging
stop pg generation on address deletion
change log level
changes to logging
file logging
prop logging
logging
improved network definition
fixes to adding addresses
remove debugging message
fixing shutdown
fix normalization logging
optional timestamp
support for log data directory
fix output format
minor changes
adding RIL settings to default conf
simplify definition of address delete: only peer id and address id is
required
fixing file handle lookup one type to store them all
fix: setting network in address with eval tool
improved logging for eval tool fix for proportional solver: reset
calculated quota after assigning
- refine preference calculcations to reflect all clients - refining
bandwidth distribution in a network - adding a configurable factor to better
respect preferences
fix: test tests for the wrong directory
fix log level
fixes for the logging
clarification for proportional solver
fixing network generation issue with mlp evalution tool
comment
only call callback on change
- combining performance selection and stability in proportional address
selection - adding stbility tolerance configuration setting
fixing memory corruption
major changes to proportional solver including stability support for
address switch
clean up for addresses and modification to perf to get values in averaging
fixes for the solver and perf tool
adding config option to check for feasibility only
dbg option to autoscale problem
logging improvement for perf tool intopt presolver support + cfg option
logging update times
extended error messages
increase numerical stability: replace BIG_M with network quota
supporting additional log formats like MPS
option to disable objectives
perf_solver_conf with mlp options
implementation of mip gap tolerance interval
implemented mip gap
adding function to read float from cfg
read coefficients with cfg functions
use cfg function in solvers
updating cfg
fix logging for failed solutions + cfg
fix check
add step duration
header
adding bt to data file header
log if peer is requested
remove unused break
fixing division by zero in normalization
setting feedback with preference generator
reduce log level when solution was found within optimality gap
- increase log level when quotas cannot be loaded from configuration -
fix for setting default inbound bandwidth
remove log level and initialize values
added some fixme
using new api
new api
new API
new api
improved output for gnunet-transport
remove debugging output
remove checking for packet order in quota compliance
improved log message
remove unreliability test - update reliabilitility test: this test only
ensures that all messages are delivered and are not corrupted out of order
delivery is not checked any longer
rename cfg files in Makefile
new test case
newline
address switch test
new configs
more configs
missing config
add tcp test to makefile
improved test
performance API indicates disconnect
fix for address switching when both peers switch at about the same time
and an additional SESSION_ACK is received
disable hostlist server
remove breaks not required anymore
fix: increase address counter only when address could be converted
fix address resolution process
remove fix me: probe_ctx are stored in a list and clean up on shutdown
performed
fix crash
change order for error check
minor changes
bug hunting
error handling was moved to clients all clients (gnunet-peerinfo,
gnunet-peerinfo-gtk, gnunet-ats/transport) support error handling with error
code
these messages should not be logged as errors when addresses have changed
(due to plugins disabled, ip address changing) this behaviour is expected
fix semantic for address_to_string
this is not an issue: no break and just print info
fix for bug #0003416: do not stop parsing uri when plugin is not found -
changed semantics for address generator cb: GNUNET_SYSERR indicates stop, >= 0
indicates bytes added
reducing output
debug code to track down assertion in line 668
remove protocol violation - this issue can occur when 2 peers use the
same addresses, for example private ip addresses So if a remote peer uses and
publishes the same private address as the local peer, we might receive our own
id if we try to connect it
fixing compile error
minor fixes including string internationalization
added fixme
renaming to perf test
fixing unused code
fixing issues found by coverity
fix type warning
printing peer ids fixing first set of cfgs
extended test to check configuration files + fixed cfg files
describing different tests
fix compiler warning about uninitialized variable
remove unused header struct
updated peer ids in test cfg
test documentation
adding documentation
docu for address selection
fixing comment
forgotten commit making penalty quadratic as used and documented
target not required anymore since we use HELLO_ADDRESSES
detecting duplicate connects
if suggested address == primary address only update quotas
reduced debug level
this break was never seen ... so not required anymore
make delay computation readable perform validation only if plugin is
available, otherwise reschedule
fix uninitialized value
log message for delayed validation
info when refuse otherwise debug
renaming
use static variable
fix fixmes + minor debug
fixing memory leak
addresses must not change the active value
adding useful info messages
remove
removing duplicate curl handles
fixes for neighbor:
reflow paragraph
do not notify transport multiple time for a session
clarify cfg loading
fix for cfg loading
tracking down connection exceeded issue
- introducing state for put requests - clean up shutdown - renaming
GET/PUT connections to "requests"
additional log info
wrong type
fixing memory leak and session disconnect
- fixing use after free - renaming connections to requests
fix statistics for reuqests
missing mapping for CERT
fixing USE_SUSPEND error with mhd daemon and free error
improving documentation
gnunet coding style template for eclipse
rename "name" to label
fix docu, make code readable, remove unclear FIXME, add max frequency
per peer back-off for transport connect attempts
do not use NBO internally
bandwidth stability for ATS
dns forward and reverse lookup functionality for http names
minor changes
clean up
fixing assertion in gnunet-ats.c:250: hashmap did not copy peer ids ->
crash
remove debug message
duplicate timeout logic: numeric_reverse frees memory but does not unqueue
fixing doxygen
doxygen fixes
fixing doxygen documentation
fixing documentation
minor changes
doxygen fix use '+' as separator in hello uris and use ipv6 references
delimited with '[' and ']'
using tcp terminology for 3 way handshake: syn/syn_ack/ack renaming all
functions, states, messages ...
rename switching state
profiler
disable autostart
improve logging on cfg error
profiler and cfg
profiler done
missing state rename
remove PORT related information from unix add unix perf cfgs
improved profiler
scritp
config files for tests
fix: bluetooth helper with suid installer
get improved warning
compare with expected value
no one cares about this message
fix for neigbours created multiple times
fix for missing receive not done call
removing unused "RF_PENDING" flag
removing experimentation: replaced by sensors
fix for mantis bug #3606: fixed path suffix appended to GNUNET_PREFIX
fixing testcase: task not correctly unset, passes for me now
fixed and improved test for address switching
test failed to compile on my system
new testcase to test delayed ATS service startup
additional test
test for solvers to support delayed start
minor beautifications
improved test description
fix crash: check if problem is already created before updating
RIP fulcrum: machine does not exist anymore
implement linear back off for restart attempts
fix: compiler warning for uninitialized value
removing GNUNET_ATS_Information from ats-testing and logging uses
GNUNET_ATS_Properties instead
removing unused variable switching from ATS_Information to ATS_Properties
in gnunet-ats-sim
adding number of preferences to allow iterating over preferences
adding new ATS struct, updating config files to use lower case solver
names
fix order in initializer
fix preference parsing
adding disconnect check for performance cb
re-renable ats-tests directory: tests and tools compile, tests pass on my
system
removing ATS preference count
adding a first set of fixmes to stun code
adding a first set of fixmes to stun code
Maximilian Szengel (147):
Adding arm list/info feature.
Using GNUNET_snprintf.
Added initial version of regex lib
cleanup
fix
towards dfa
- dfa construction
formatting
api changes
doxygen fix
- DLL instead of SList - comments
DFA evaluation
NFA evaluation
- dfa minimization wip - static functions - optimizations wip
fix
Automatic regex generation for testing
better testing
removing unreachable states
fixes
dfa minimization wip
fix
doxygen fix
comments
Added '?' operator
bugfix
- SCC detection - Graphviz scc coloration - optimizations - api
api changes
added accepting state info to api
test update
dfa minimization fix
recursion for dfa construction
- State merging fix - Proof creation WIP
Fixed compilation warnings
Fixed warning
fixes
Towards new proof algorithm
assigning proofs
Better proofs
even better proofs
Test for computed regex.
removed unnecessary brackets
simplified regex/proof generation
more regex simplifications. fixes.
Updated testcase to use fewer random regexes.
regex simplification wip
regex optimizations
new test for regex
regex bugfixes and optimizations
doxygen fixes
new and improved tests
cleanup, fixes
regex bugfixes
removed unnecessary debug statements
fixed leaks
Summary: regex cleanup and bugfixes Author: szengel
regex: actually checking the proof, fixes
regex: fixed static analyzer warnings
regex: added verbose flag to graph api.
regex: iterating over the initial states
regex: fixed iterating over the initial states
regex: fixed iterating over the initial states.
bugfix
Fixed coverty issues
- Added coloring option to graph saving. - Fixed a bug causing NFAs not
to print properly.
using strings as labels
Added multi-striding capabilities to regex.
Fixed eval test case and stack smashing protection
- added check for automaton traversal - fixed a bug that caused nfa's
state_count to be incorrect for certain regexes - only compute scc's when
coloring option is set
Fixes
coverage
removing temp file in graph test
comments
coverity
optimizations
Added warning to check the firewall when tests fail (which happend to
me). The long diff is due to running pre-commit script prior to check-in.
fixed leak
DFA path compression
refactoring
regex: iteration improvements/fixes
fixes
ip/prefix to regex
iptoregex test
tests
fix
- ssh check for all hosts - linking directly after registration
regex testbed testcase
simple overlay connection working
fix
mesh service connect skeleton
renamed test_regex_big / fixes
fix
replacing test_regex_big with gnunet-regex-profiler
doxygen
doxygen, indentation
Moved regex profiler to mesh, because of dependencies.
fix
Exposing GNUNET_TESTBED_host_get_hostname
regex profiler
doxygen
mesh connections in regex profiler
mesh connections in regex profiler
coverity
- Added a new regex to test_mesh_regex - Mesh connect and regex announce
in regex-profiler
regex profiler fix
Aborting profiler after a timeout
Updated regex profiler config with mesh and dht default arm services
regex profiler: loading a set of search strings from a file
regex profiler: measuring the time it takes to match all strings
regex profiler fixes
Added statistics service connection to regex profiler
formatting
regex profiler saving statistics
added libgnunetstatistics to regex profiler dependencies
fix
- Added regex specific stats to mesh service - Regex profiler fixes
libmysqlclient linking for /usr/lib64 (fedora 17)
regex profiler fixes
- Added path compression parameter to DFA construction API - Moved NFA
construction to internal header - Added regex simulation profiler (for
profiling the NFA, that results by merging several DFAs in the DHT, in a
database)
added max path compression parameter to regex profiler
using prepared statements for regex simulation profiler
passing mesage header instead of message
batch regex announcing in profiler
sequential mesh service connects and regex announcing in regex profiler
mesh regex profiler
Using regex for exit/vpn
regex profiler fixes
regex profiler cleanup
regex profiler fixes
fix
- removed unused function / var - indentation
Adding new regexprofiler service that is used in conjunction with the
gnunet-regex-profiler to announce regexes from a policy file
Commiting changes to regex_profiler_test.conf before moving it to contrib
Moving regex profiler config to contrib
regex profiler fixes
regex profiler fixes
Added regex prefix to config for regex profiler.
- regexprofiler service to daemon - enabling nse in regex profiler
regexprofiler service to daemon
Better mesh/regex profiler performance.
doxygen
optimizations
union of regexes in profiler
doxygen
Removed waiting time before linking
Cleaup, indentation, comments etc.
Saner default values for the infiniband regex profiler conf
Fix 'way too many REGEX PUTs' issue.
Moon (13):
initial NAT lib commit (UPnP and NAT-PMP support)
add src/nat/Makefile to AC_CONFIG_FILES
add src/nat/libnatpmp/Makefile and src/nat/miniupnp/Makefile to
AC_CONFIG_FILES
add missing test_net.c and fix bug about sockaddr* cast
fix copying in6_addr
fix handling of NULL addr argument
bunch of fixes to bugs found by grothoff
rework UPnP code to use GNUnet scheduler and network API disable NAT-PMP
support for now
don't search for "upnp:rootdevice" (bug 1609)
fix missing curl linking
fix choosing UPnP device URL and device type
Fix include path and include order in src/nat (patch by LRN)
Fix W32 interface listing, and correctly copy addr in test-nat (patch by
LRN)
N. 'ng0' Gillmann (9):
a first working version of guix.scm
guix.scm: Add mysql and postgresql as inputs.
trivial fix in guix.scm
Moved guix.scm to guix-env.scm into a proper module and added a small
documentation how it can be used.
+
Added myself to AUTHORS file of gnunet for Gentoo maintainer. While its
outside of Gentoo portage, it is still a job I do.
#:use-modules -> #:use-modules in guix-env.scm
revert r38067 .. wasn't working, needs testing
gnunet/guix-env.scm: Disabled tests until they succeed, added debug
output + CFLAGS -ggdb -O0
Nathan S. Evans (1128):
heap merge from .8, fixed testcase
forgot heap check
heap bugfix
moderate udp support, not really tested (:
remove comments
pre-commit
fix memory leaks
remove debug message
small changes
making udp maybe, possibly, do something
whiny christian fix
missing tcp data type
make recvfrom_amount a bit more nice
fixing api related changes
declaration fix, forgot to commit
still massively broken, but this stuff now compiles... transport_api.c
needs work (who is going to do that?)
it compiles, but definitely doesn't work
transport ping/pong message types
<none>
other things
start of ping pong changes
start of ping pong changes
<none>
missing commit
test case performs better, ping/pong happens correctly now (for me).
Still requires testing and fixes, also udp still not working as it doesn't
supply any addresses to be used in a hello. That should be a simple fix, the
reason the test case is not working is likely a misunderstanding of what is
supposed to be happening
remove debug hard coding
why didn't I commit this yesterday?
remove debug options when not specified, otherwise they linger in the
configuration between test runs
resolver for udp plugin, don't propagate welcomes in tcp
udp closer to working, still wrestling with how to do data structures
properly in gnunet-service-transport.c (grr)
transport plugin change from single MessageHeader to message buffer and
size...
still not really working... sending and receiving of messages a bit hinky
udp test_transport_api test now working, tcp not... but a good omen
allow message chunking in tcp, still not working seemingly due to a
problem in server.c, though not sure why so it's probably actually a problem
with me
found the bug, test cases should now work
peer disconnect bug fixed
makefile update, removal of debugging DEFINES
proper latency calculation, though I think maybe the time difference
calculation should be a bit more flexible
<none>
<none>
maybe this will fix buildbot errors...
try again
with debug flags
works for me with debugging... weird
hopefully partially fixing non worky problems on various machines. but
since I haven't tested it on any other machines, we shall see if this is the
case. Previous issues likely related to the ack_received stuff which was still
present which caused tcp connections to have undesirable behavior
fixing (hopefully) transport hang issue by checking tc->reason
comment fix
<none>
change test cases to not write configs
change default config file name
coverity fixes
coverity fix
try connect fix, just because we have a neighbor in our list doesn't
necessarily mean we can already talk to it. Functional bug apparently
coverity bugfix
some minor changes that made things not work, but other things still are
keeping things from working now
nasty missing GNUNET_SERVER_receive_done bug fixed... At least one
morning wasted
test case changes
minor changes, with any luck test cases will pass (literally luck though)
remove debugging flags which keep creeping in somehow
received ack removal
remove old testcase lines
cheapest address calculation fix
dying transport_api fix/hack
missing frees
debugging statements
debug
wrong order
christian will never change
set last activity to now to give a bit of leeway for a connection to
happen...
changes to continuations required change of state machine maintenance (or
so I claim)
longer timeout to make test work on slower machines, but the problem is
still that the timeout for notifying transmit ready somehow gets propagated to
core retrying a transmission
remove debugging flag
remove debugging flags
remove super verbosity
testing changes, beginning of topology stuff
testing related changes
missing files for testing
api mismatch fix
missing function reference
changes related to my api changes
dce
undo christian hasty commit
undo christians hasty commit
bad ordering
testing changes, including real topology test (only for clique, and not
that great of a test). Need to verify it will run on other machines... also
need to commit it since it currently works for me
better ordering?
missing definitions
start process change to allow stdout to be sent from child process
api header changes
speeling
testing changes
add args for GNUNET_OS_start_process's
add stdin functionality for os start process function
missed testcase somehow
actually add test case, and added proper stdin support to
os_start_process function...
possible start for windows passing of stdin and stdout to child process
adding -lm
use gnunet read function for testcase (mostly to give me an example), add
dwFlags as instructed by Nils
test case cleanup
semi functional nat punchy dealy
nat stuff... works on my machine
message types for nat punching
add verbosity level to gnunet-nat-server
change help option for gnunet-nat-server
test case change, better test of core perhaps
mangle hellos so that external address gets sent out
add support for sending port along as icmp payload. needs tested on NAT
boxes to see if it really works
changes so that udp_nat sends running port along with icmp echo reply for
optimistic nat traversal
coverity, doxygen cleanup stuff
coverity, doxygen cleanup stuff
bessere warning :P
unused variable warning
udp nat test case doesn't fail if not installed as root
towards better topology testing, still a kink or two
add support for multi topology testing to makefile (all work for me, but
only clique is currently a testcase)
bug fixes, feature additions
error when not using verbose (yay for Werror)
coverity fixes, strange behavior(s) in transport_api.c hopefully resolved
merge conflicts, coverity and other fixes
connecting peers relies on connections completed thus far, instead of
estimating how many should be complete
remove verbosity
speling
speelling
execinfo definition
finer grained debugging
finer grained debugging
remove debug
ifdef instead of if
disconnect neighbor on single failed address, wrong behavior
re use validation to keepalive connections. May need more work/thought
shell for dv plugin and service, not yet working
dv config files and fixes
incremental improvements, testcase which (should) work once dv works
non-compiling, but getting there
dv message types
wont work because someone made my random heap function unpossible.
committing for posterity mostly
adding random_heap function definition in case I add it another way
fondly remembering ciphertext_send... all the pieces are there, but still
doesn't work. Needs debugging, as usual.
Christian message size assertion
dht api, shell dht service, base of future test case.... not yet working
dht api
beginning of changes
dht related types and changes
initialize size to zero
remove initialize size to zero
getting dht closer to being this crazy meta dht thing
dht related changes to message types and whatnot
basic api to service communication has been hammered out, as well as
service confirmations going back. still needs work, many things not finished.
added better continuation behavior to get start, put, and route start.
test case now properly handles message confirmation receipts from service.
find_peer api call still needs implemented, but we are generally much closer to
a working point. i'm sure there are coverity issues as well as doxygen crap to
be addressed
api definition changes
doxygen cleanup
doxygen additions
doxygen
test case verbosity
use gnunet-arm instead of gnunet-service-arm
gnunet-arm option to delete configuration file and servicehome directory,
tailored only for testcases as users probably won't want to delete their
configuration file whenever shutting down gnunet (unless they really hate it)
use loglevel defaults for gnunet-arm
remove debugging from testcases, testing now used gnunet-arm to run and
stop gnunet, removes configs and directories by default. remote still needs
tested
write control_host information to testing configs (for remote starting),
hasn't been sufficiently tested
remove todo
small whitespace change
small changes
missing frees
coverity fixes, test case tweak
missing free
change hostname in config files when created, if running remotely...
Otherwise the testing controller doesn't know whom to connect to. Since these
daemons are ostensibly running on said remote hosts, changing the hostname
shouldn't pose a problem (unless it does)
more tweaks to allow testing to actually work
remove files
small fixes for dv
add quiet flag (fix for test case)
dv changes, closer to a working state
datacache instead of datastore
work in progress.
dv type, hello api call
hello api call, may be unnecessary
lots of fixes, on its way to actually working...
add find peer to api, went a bit continuation crazy but i think it's
better this way...
service does simple put and get into datacache, test case verifies it
works
doxygen fixes
forgotten commit
doxygen fix
add gnunet-dht-get and gnunet-dht-put binaries, changes to service
doxygen is annoying
put and get testcase
doxygen
more doxygen things, i think
basic find peer functionality, added to test_dht_api testcase
dht find peer binary
add dht to default build system
add dht to build system
bash vs sh
debugging fixes
see if timing is cause
hashing fix
type check
cleanup debugging
dont know how that got out of sync
make test case a bit more robust using timeouts and retries
inexorably closer to perfection
many changes, won't work without blacklisting and more work
dht api fixes, it works again (for me)
core api related change
memory leak fixes
initial transport level blacklist stuff
testing changes, which hopefully make testing code more robust, though
problems are inevitable
testing
testing
remove reference counts
dht and testing lib stuff
get rid of transport asking arm to stop services
only committing in case of natural disater, not to be used
changes so things compile
increase default timeout for startup (for SLOW machines)
reduce number of peers
less stressful testcases for slow machines (especially a certain sparc
buildot)
testing bug fixes (hopefully)
increase set key delay, to help figure out connectivity issues on slow
systems
set a running flag, so server can be disconnected (allows larger
topologies)
testing bugfix, add scale free topology
minor testing changes
print topology dot file with appropriate name
commit lib change
dht was removed at some point, re-adding
longer timeout for file locking
missing distance metrics in transport stuffs
in case of fire, or volcano. still not to be used by anyone for anything
coverity warning fixes
remove files on exit
dv changes. add propagation of disconnect messages, dv updates on peer
direct connections. Result of those two changes should be a much better
learned topology for all peers. This was periodic gossip is really only to
stop timeouts from removing peers. Also changed dv_api to wait for message
queuing by dv service before calling transport transmit continuation, so that
addresses that are no longer available can be caught by transport service.
Perhaps some other stuff I have forgotten about. Should be kinda sorta
working, need to make better testcases and do more debugging as errors
inevitably pop up
message type changes, and dv message changes
debug changes for transport
address printing for dv plugin, doxygen comment
doxygen
get rid of duplicate statistics handle in each plugin
copy config to /tmp/ so arm can remove
main change for gnunet-arm is that gnunet-arm now gets proper
confirmation when called to start or stop services. This means that calling
gnunet-arm to stop/start a process will return 0 only if the process is
stopped/started. For gnunet-service-arm, this is implemented by leaking the
client connection socket on a shutdown request, so that it is only closed when
the arm service actually shuts down (indicating everything is really down).
changes required for arm fix
arm related changes to util things, enabling leaky sockets and shutdown
messages from services
verbose debugging
fix for infinite loop on dht shutdown failure
verbose errors to help debug arm problems...
try to get peerinfo to start sooner for dht testcase
temporary hack for gnunet-arm. given enough time processes seem to go
away, so give them time
fix for failure of dht testcase due to long wait for hostkey generation.
testcase now uses gnunet-peerinfo to generate the hostkey first, then runs
gnunet. A much better fix would be to have services that need a hostkey not
report to arm that they are happily running when they may hang indefinitely
trying to generate or retrieve a hostkey. Or if most things really need a
hostkey have gnunet-service-arm generate the damned thing BEFORE starting
services.
output PASS
missing free
comment for fall through
doxygen param
library changes for testing
changes to testing... create hostkey using peerinfo before starting
peers, better topology creation and more options therein, changes to order of
starting peers, connecting peers... new order is hostkey create, topology
create, blacklist create, start peers, connect peers
change to testing api
change to testing api
change to testing api
sorta improvement on creating at least X conns per peer from allowed
topology. using the hashmap iterator to add exactly one peer per peer from a
clique topology created a star because the first peer in each peers map was
always the first peer. Now a more expensive hash map traversal is used, but
should be fine for testing
testcase configs for using the minimum and random connection options
enable more verbose logging to debug on buildbots
better failure handling
depth first peer adding, non-null check for peer lists
change testing library to use timeout for peer and peergroup startup so
testcase/user can decide how long to wait as opposed to arbitrary wait counts
doxygen
doxygen and testing api change
testing api change
testing api change
testing api change
test case fails properly now
some testing stuff, but i think there is a problem in
gnunet-service-transport
more testing changes
testing api change
missed change
another one
comment fix
blacklist testcase, testing_group support for blacklisting
config for blacklist testcase
oops, missing commit
sigh, another missed commit
warn on daemon termination failure
reduce debugging messages
ridiculous timeouts, sparcbot can take up to five minutes to create 12
connections (not sure where all that cpu is going)
remove datastore testcases
doxygen
removing useless signing until fixed for real
reduce max set key delay, exponential backoff anyhow but let there be 3
in a minute
churn fixes and testcase
add now instead of add delayed
moving closer towards a working dv implementation
add UDP inside ICMP for NAT punch
remove unneeded udp crapulence
remove optional port information, always use since data indicates NAT
boxes don't care (or would not work for other reasons before checking payloads)
somewhat simplified, somewhat working udp_nat implementation. let the
buildbots thrash on it and see what happens
doxygen
coverity fixes, revert to 11639 if broken
change from udp_nat to only udp, altered testcases and configs to suit
fixes
malformed address handling
proper test case configuration
minor cleanups
first implementation of NAT punching for TCP transport plugin... doubt
that it works, but testcases that do NOT use NAT still work. not to be viewed
by anyone.
minor but necessary commit
working TCP PWNAT implementation (at least on my machine), also base
testcase and reliability testcase
add support for using both NAT and non-NAT addresses, fix bug when
multiple ICMP probes received
nates public key for testbed
missing configuration files, may help stupid build errors
bugs from codesonar
extra check for udp based punch
explicitly add user provided external address to known addresses when
used with NAT
always add user provided external addresses (duh)
fix for bad transport test case configs
test case config files
change to configs, use readable names for topologies and connect options
instead of cryptic numbers
changes to use names instead of numbers for topologies and topology
options
changes to use names instead of numbers for topologies and topology
options
dv changes, still working out an issue
working distance vector transport plugin implementation. after spending
days on it I think the problem (not all peers connect all the time) either lies
in core, or is related to latency
some testing changes, including an api change that likely breaks things
for others
api change
testing api change
resolve merge conflict
api change
testing api change
alignment
clean up files
dv changes, better logging and so on
dv changes, remove logging, smaller test case
connect in line, not ring.
so buildbots can reveal errors in DV code
doxygen
UDP message types
codesonar changes/fixes
codesonar fixes
codesonar fixes, some other changes
enable debugging to help figure out why buildbots hate dv
remove ugly ifdef for message tracking
try really long timeout for making dv testcase work
missing null check
make test cases build, but not run
define nat port
don't stop churned peers, clean up files
tokenizer api change, a bit of source cleanup
don't send hellos to transport unless DV really adds the peer (useless
resending hellos)
someone changed how services start and didn't tell the dht testcase
extra check for blacklist, spelling
really stupid string bug fix
misc bugfixes, especially core dumps for dv test case
remove config file for churn, not removing directory specified in config
file. May be a problem, unless the assumption is that test cases create/remove
their own directories in which testing runs
fix execinfo debugging so it doesn't deref NULL pointer, still not
entirely useful for the common person
bad config file
test case error handling
nicer debug format
logging
hello size added to debug
better debug message
more verbose
plugin not receiving messages
plugin not receiving messages
plugin not receiving messages
plugin not receiving messages
plugin not receiving messages
move client listen call into transmit start callback, so we know service
exists and is connected... otherwise we may stop listening before the service
starts
remove extraneous debug messages
unused code
defaultservices started all the trouble...
test unknown sender hypothesis
fix recursive message sending (last first causes issues)
delay message sending to test hypothesis
trying message queues on receiver side
don't log blacklist stuff as warnings
remove variables
fixing christians overwritten changes
fixing christians overwritten changes
put number of additional messages in config file
put number of additional messages in config file
coverity and codesonar fixes
coverity and codesonar fixes
debug fix
proper check address behavior
doxygen fix
doxygen fix
more portability changes; so NAT punching also should work
check address change, remove unused debugging case
callback for peergroup shutdown in testing, requisite testcase changes
callback for peergroup shutdown in testing, api change
shutdown callback
shutdown callback
shutdown context
hashing random memory to get a peer identity just does not work
check address function
add test case for NO topology restrictions, connect in ring. good for
maximum number of peers testing
change to limit number of peers actually starting concurrently so as not
to exceed max file descriptor limit
oopsie
fix for null hostkey callback on daemons_start call
test to see if removing statistics makes test case stop hanging
twopeer dht test, mostly for class
two peer start, connect, ensure core connect message made it to the dht,
disconnect, stop peers
give dht service a chance to receive core connect message
comment wrong
reduce buffer size for all connections, and for mst processing
define smallest buffer for connections
valgrind tests
remove valgrinds
remove retransmit
remove retransmit
remove valgrind
remove max buffer size from everywhere in util
remove max buffer size from headers
fix core testcases which expected to always be given larger message size
than that requested
remove max buffer size from socket create calls
statistics api lying about message size
libgnunethello got removed
server mst change
added dv twice and dht never?
bad commit fix
bad commit fix
remove nate specific debugging junk
try again to fix makefile
codesonar fix; config change
failing datacache testcase
codesonar fixes, hopefully doesn't break anything
codesonar fixes
remove debugging
more debug removal
more debug removal
thought realloc was problem, but not properly freeing memory. believe
that transport segfault *could* be related to growing connection buffers on
demand... appreciate another look if someone has the time (Christian...)
remove redundant check
another DHT testcase
powerbot seemed to fail dv testcase when uid of 0 used with print message
(using llu instead of u for unsigned long)
change for powerbot
dht testcases and some minor other changes...
dht related protocol changes
ignore case
change testcase to use correct binary
correct binary
missing file
<none>
dhtlog plugins
missing file
add testcases to run on make check, fix dht a bit
missing file
check library location as done in configure.ac
http test not built if plugin not
minor blurb change
mhdless fix
plugin to write sql queries to file instead of connecting to server
ignore sigpipe
some dht changes
change from host as char * list to linked list
definition of LL
h2s_full, useful for testing
nicer names for C-dog
uninitialized value
further arm do daemonize hack, easier to debug arm startup (debatable)
allow control connections to transport service
don't try to connect to all peers, but only those advantageous for routing
testing function to get current topology from peer group
iterate topoogy header related stuff
core changes for topology iteration (get current connections)
spelling
warn for error case
type warnings
add actual date since we aren't inserting data NOW, call for inserting
topology information
bug
test printing connected topology
remove unnecessary services
add topology table creation, insert, update, get, etc. to the mysql
dhtlog plugin, also added to test case
topology stability testcase
mysql fix for certain distros
minor fixes
making things work a bit nicer
routing table changes, not stable
fix for overzealous peer disconnect bug
mosgly fixes for memory leaks, misunderstanding of some mysql library
stuff
useless line
add statistics iteration to testing_group, test case for same.
testing function definitions, minor stuff
change warnings to debug
dv needs testing
lots of little things, mainly report statistics at end of dht profiling
testcase
api change to allow client to set DHT malicious, still need to implement
it on the dht driver side
message types for malicious control messages
finish implementing malicious peer behavior
hopefull stop strict kademlia from being breadth first search of network
fix for kademlia routing
testcase fix
myriad changes
stagger shutdown
<none>
fixing unused testcase
add bindto for udp transport as well, in case it's used
some nat fixes
trying to debug
fixing tcp nat implementation, WINDOWS SUPPORT NOTWITHSTANDING
remove warning debug message
print correct size
test case configuration changes, to make nat tests more realistic
missing add to address list
initial try at windows version
initial try at windows version
initial try at windows version
windows crappage
windows crappage
windoze
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
jaunty hang fix... need to further test disconnect handling
windoze still sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze sucks
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
yes, windows still sucks
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
windoze suckssss
workingish?
workingish?
workingish?
workingish?
makefile correction
wrong socket
blame where blame is due
blame where blame is due
remove un-needed fake pong
dht changes i don't want to have only on my machine
minor dht alterations
api change to go along with service change
increase timeout for find peer request, since it relies on host key
generation, which may be slow on some machines
spelling
add new mysql dump plugin, in future maybe replace current with faster one
assorted dht changes, fixes, etc.
fixing stats iteration
wrong types
always stagger everything
attempt to equalize find peer activity for kademlia and gnunet routing
broken dht compilation fix
broken dht compilation fix
various fixes, improvements, fubars, for dht... still tuning
missing function definition
transport service hello reduction, klocwork fixes
klocwork fixes, dht routing changes (hopefully fixes)
klocwork fixes
klocwork fixes
missing statistic include
broken stuff
fix
check parent pid, die if gone
remove tcp_nat reliability testcase, because transport switches seem to
be happening and since a non-nat'd peer may have multiple addresses this isn't
really a problem
fix for not accepting pong with address null errors in NAT test case
putting the nat reliability test case back in, though as far as i can
tell from running the test on jaunty-arm there is no way for transport to
switch addresses, which leaves me puzzled
dht changes, mostly making the driver do find peers adaptively...
currently not seeing any crazy issues
missing definition
missing definition
forgotten commit, null check
making core request connect work
core request connect in addition to hello, minor driver stuff
allowing a single stats and/or peerinfo service per host
handle null continuation
niceties
strange off by one fdset count
change timeout, so multiple connect request messages can be sent
allow multiple request connect messages to be sent to transport (failed
to connect first time)
allow transport api to send multiple request connect messages
messages 'hone in' more the higher the number of hops, still needs some
tweaking
add math support
dhtlog updates
fix arguments
make insert trial easier to not screw up
add republish support in dht (default once an hour, just like kademlia),
improved next hop selection (or so I think), use 'TrialInfo' struct in dhtlog
testcase
support for running stats and peerinfo services per host (or multipler
per host) instead of per peer
committing so line numbers match
comparison bugfix
comparison bugfix
assert in dht to help track down bug
possible fix for buildbot failure on NAT test case; once a 'NAT' address
is used and a connection established, the new connection doesn't have a 'NAT'
address (port 0) any longer, because it was built off of an inbound connection
from the NAT'd peer. However, since we initiated the connection we expect that
whatever address transport gives us should match.
get rid of debugging messages
remove debugging
better values for dht bloomfilter, reduce collisions for peers even with
large routing tables
kademlia compromise, add some ifdefs, compile error fix on blackbox
verbose debugging for NAT tests
document
remove crazy debugging
some testing changes, features added
testing lib changes
test case change to match testing api change
allow null continuation for core_api_peer_request
remove port from transport section because it was getting in the way of
making sense of netstat
don't ignore sessions marked as inbound when checking whether addresses
match
documentation for new parameters
fix for api change in disk.c, thanks to patch from LRN
safe stdin/stdout inheritance under windows
add retry attempts to test_dht_twopeer, remove core from defaultservices
to ensure that the dht gets started before core... Otherwise the test can fail
because the data being searched for is the other peers public key, which is
only known once connected.
testing fewer messages (timeout issue?) on arm and powerpc
feature creep
sql related changes
dhtlog inserts round info
many useless changes to the dht testing driver, mostly for churn (which
may not even be useful)
mostly options fixes, new convergence formula for testing
mostly untested testing code, throttling churning of peers in the same
manner as normal peer startup and shutdown is
debug messages in driver, correct time (even for mysql_dump), change
dhtlog testcase so it doesn't leave files after running
some fixes, debugging churn nonsense
testing api addition
remove gnunet_break on allowable condition
missing variable
missing return fix
fix compile warning
daemons_vary fix
insert new round on first round
fix for remote testing
try new method for sending find peer requests
missing return
commit round information earlier (trying to get more accurate timing
information)
additional useful hash function
additional topology creation option, for adding closest underlay capable
peers to topology
missing
increase max connections, add verbose flag for topology testing (buildbot
debugging)
testing max connections
force delay between find peer requests and puts/gets
move bit distance function into util
increase timeout, sparcbot testcase might pass
increase timeout to give sparc buildbot a chance
longer timeout, give chance for success on sparcbot
even longer timeout, still timing out on hostkey generation (sparcbot)
fix compile error on buildbots
insane timeout for testcase
backup previous changes before altering routing algorithm yet again
only get statistics once per running stats instance
fixme for missing def
new routing method, semi-tested
redefinition of enum value broke compilation
logging to figure out whether convergence options are properly set
logging at an annoying level
stupid mistake
removing debug message
comment out iv_seeds, see if problems vanish
fix, not removing clients on disconnect
fix get error (never canceled) in test_dht_api, remove valgrind from conf
type errors, test case errors, etc.
mis-named functions (this means that SOMEONE doesn't have exhaustive
testcases)
test case fixes...
test case fix, dht bugfix
dht service bug, reduce debug level
missing route2 call
reduce verbosity of testcases, minor fix
typo, continuations for dht operations use continuation instead of being
default scheduled (GNUNET_SCHEDULER_REASON_TIMEOUT indicates error)
unique id is quite useful for the dht, even for puts
find peer messages were never getting actually forwarded, factor out some
unused cls args
toelke: don't bug dht
add dht testing option
change connect topology in testing_group.c to call callback on all peers
connected (or possibly timed out connecting)
trying to make things clearer, probably failing
remove debugging stuff, fix for very small values of small world options
braindead
fix logging for transport testcase, do suid nat binary check in
reliability testcase (execution is pointless)
missing NO_TASK assignment
core_api possible fix for timeout case
redeclaration of converge_modifier inside function makes global useless
oops
binary converge option in dht, make find peer requests more likely to be
sent early in testing
replacing GNUNET_OS_process_get_pid
possible fix for dht testcase segfault
memory leak fix (may cause segfaults on exit)
big scheduler refactoring, expect some issues
scheduler removal mishap (fix)
codesonar fixes
codesonar fix
codesonar fixes
Adding replication parameter for initiating GET and PUT requests to the
DHT.
Count message splits
read put and get replication from config when testing
decrease error verbosity
Log topology information with less verbose logging as well.
missing pointer set
coverity fixes
bug fix
missing null check
testing if flushing files corrects a problem
not sure how this has been working, unless something got out of sync in
svn... also not sure it is fixed
Fix for testing where peer_start callback immediately calls peer_stop
(also white space changes)
add options to defaults.conf
Testing changes required to work with new core api. Revert
GNUNET_CORE_iterate_peers call to work properly.
Fix for blacklist testcase; more aggressive reconnect attempts.
remove debugging flags
change section for converge_modifier to make sense
bad free
annoying those with email notifications of commits
dangling pointer fix
bad cls value
set GNUNET_NO on can_drop for connect messages
better forward_count implementation, only timeout forwards, no limit for
local requests (only remove on cancel)
report failed connection attempts in driver, use a queue bigger than 1 in
core
dont try to send to more peers that exist
dont try to send to more peers that exist
Fix missing notify_cancel in dht service, dhtlog_dummy bad init return
memory leak
partial fix for missing connect notifications from remote peers, still a
problem somewhere though(!)
possible fix for timing issue (causing missed peer connect events) in
testing
memory leaks, forgetting to GNUNET_CONTAINER_bloomfilter_free
only free when necessary (double free bug)
possible fix for CORE_disconnect segfault, GNUNET_break instead of
segfault in other error case (will see if this causes more troubles)
checking if pending_head is NULL, may not be correct
more core_api checks
invariant pr->pending_head <> NULL still getting violated
invariant pr->pending_head <> NULL still getting violated
specify total connections desired after find peer requests for dht
testing, use float for better precision in replication
bad ordering for config values
allow bucket size, kademlia replication to be set by configuration,
increase base replication
are messages getting dropped, or is logging not finding all paths
are core_notify_transmit_ready calls failing?
count number of messages queued vs. number sent out
send multiple messages to peer if possible
restrict replies to one per request (use block library)
use remainder from forward_count
more efficient malicious behavior (discard messages earlier)
greater or equal cutoff
tracking down strange routing behavior
use hashmap instead of linked list of neighbors, fix for double free
add comment to explain
missing nat server check before trying to kill process
strange update neighbor performance function
initial dv changes for atsi information handling, test case BROKEN
fixes for DV with new transport distance and core changes over the past
few weeks
better wording for function prototype
off by one, should remove churn code at some point though
Always log local request starts, even if malicious
some machines report their hostnameinstead of localhost
allow dv to be built and installed
force request connect message in transport testcase
port 0 for tcp plugin is BAD
windoze friendly check for nat binaries
different indentation, somewhat cleaner
disable UPNP in defaults for now
i am a dumb dummy
really disable upnp
test case for testing large topologies
fix for bad fail condition
coverity fix
don't know how this could have worked before
remove impossible condition check
speeling and coverity fix
reduce config variables
core connect changes
pre-check
core connect change, forgotten dht driver change
possible null dereference
klocwork fix
klocwork fix
codesonar fix
doxygen comments, last commit of bugfixes (hopefully)
core related fix
Regression test fix.
change (possible) common error message from warning to debug
trust argument does not exist
is sparc problem a timeout issue?
better explain small_world_ring, use consistent variable names
asynchronous malicious peer setup, fix for timing issues
add test and dht to default block types. DHT should be there, test is
there so users who try to do puts and gets with the dht do not go crazy
Check for missing type, set to TEST block type if so (ANY should not be
used)
make testcase end more gracefully
hopefully send to self test will run on buildbots
implement sybil attack for testing
maybe a fix for send_to_send testcase
trial and error
comment out line in config
send to self, remove assertion failure
replace line
allow to build without malicious code
load configuration properly?
proper testcase functioning
tired of seeing buildbot error messages for this silly testcase
vpn requires mesh
add blacklist option to defaults
if F2F is set in testing, enforce via blacklist (instead of relying on
topology)
blacklist stats
make really loud unable to communicate with peerinfo messages go away
reduce verbosity
Allow caller to print peerinfo error message, if it so chooses
unknown issue
change how peers are filtered (peer id gets freed somehow?)
do not print pointer
compile warning
loop forever
printf debugging
all peers match bloom filter. inexplicable.
using blacklists causes this error message to happen A LOT, reduce
verbosity
starting point for reading topology from file
new topology option
enable reading of hostkeys from a file for testing, also loading of
topology from topology file
try larger testing values, shouldn't hurt when doing local testing
bump up even more
allow 0 settle time
log topology when extended logging is enabled
don't wait 90 seconds
malicious even when settle time = 0 and malicious_after_settle is true
malicious even when settle time = 0 and malicious_after_settle is true
larger bloom filter
fix testcase
adjust bloomfilter size based on number of malicious peers
allow connect attempts to not yet started services without dying angrily
bloom size fix
<none>
<none>
hack so that configuration can force non-unixpath connections
reduce topology overheads in testing, probably breaking some stuff
param
clean up error messages
Fix for unixpath/arm bug
reduce messages from dht service
option to always forward until a closest peer is found, exact formula as
described in paper
report connection per second information
it seems once a connection fails a few times, it will never work
remove debug message
bart fix for my crappy code
initial route path recording, not yet what dht api expects
path tracking commit, hopefully a fix for dht_api connect bug, needs
tested
testcase using hostkeys from file
3809 hostkeys in data file
print warning messages about hostkey file
api change
missing return
makefile was broken
topology for multipeer test
Alter config files to use hostkeys from file instead of generating during
testing
use hostkeys from file
doxygen
header for bart polot's gauger tool
gauger name
undo change made by hasty commit
undo change
unix domain socket transport, associated test cases
makefile warnings
remove attempt at throttling attempted connections
safey ssh key
style problems on mac
doxygen
don't die so tragically
better memory usage, fix leak
fix for block libraries so they don't need to understand dht struct
spaces
spellingGG
assertion check
debug instead of error
speeling
segfault in upnp code (tip of the proverbial iceberg?)
clean up debug messages
comparing wrong data
configurable connect options
defaults for newly required testing options
revert Makefile.am to normal
vastly reduce cpu overhead when connecting peers
test case changes required due to testing api change
test case changes required due to testing api change
test case changes required due to testing api change
remove code for cpu heavy connection scheduling
enable very minimal dht logging
give testcase some more time to connect
debug
disable dns testing for normal test case
extra assertion
assert fd's are in valid range
reduce logging
reduce logging under normal (non-debugging) circumstances
don't use totally empty unix path
fix
change to dht profiling driver to avoid possible find peer issue
change to allow number of outstanding ssh connections to be set separately
Testing api change and related test case fixes, remember retry task and
clean up on exit for unix domain socket transport.
ssh per host
fix for friend file copying using ssh, should still be set up to throttle
connections
attempt to avoid issuing connect request for already connected peers
fix
testing code
testing code
reduce verbosity
handle out of sync connect callback scenario
remove double connect notify, add is connected optimization for peer
iterate call
add specific peer check to core iteration api
change testing to use is peer connected
missing NO TASK
oops
missing decrement
fix warning
respect minimal logging
missing null check
attempt to detect and debug doom spiral
missing free
free in proper place
Allow testing connection process to be stopped when asked.
nicer logging
fix
bigger floating point
fix
fix
Testing and core related changes.
Offer hello api change(!)
gauger changes, thanks bart (:
offer hello api change
hack to make NAT topologies work
fixes to gauger calls
Testing api change.
configuration option
Consistent GAUGER naming.
possible fix for mantis bug 1663
clique topology optimization, progress meter for connecting
nat topology optimization
ugh
more efficient list deletion
Indent, fix removal bug
fixed divide by zero bug
gauger api changes (:
reduce verbosity
Remove annoying log messages
Meh.
code format, fix shutdown bottleneck
Remove useless hostkey dilly-dallying in testing.
Set hostkeycallback to null.
Nothing ever is right the first time.
grumble
My brain is melted.
Fixes for testing because of my addled brain.
create nested directory structure for greater than 32k peers
1000x speedup when reading large topology files
<none>
do not avoid mallocs
api change for malicious peer bug when running distributed tests, stop
arm from repeatedly trying to bind to blank unixpath
cleanup testing_group.c, fix bad shutdown in test_testing testcase
testcase fix for api change
allow peer connect call directly from peer started callback
force connection attempt to core (should core always be listed in
defaultservices instead?
memory leak in dht service (small)
remove core connect call, add core to defaultservices for test case
code not to be used by anyone, ever
better exception handling
better exception handling
debugging
remove slash
grumble grumble
schedule task instead of waitpid
reduce verbosity
remove logging line
set peer state properly
helper process that rather dumbly starts gnunet peers
license and description
core api change
core api change
fix for over-reporting number of necessary connections
formatting
fix for dht test case configurations
ifdef, obsolete crappy code removed
Option for inserting gauger data
change warning to debug, additional assertions for correctness and to
make codesonar be quiet
initial peergroup convenience implementation
convenience api for peergroup running
peergroup initial version with test case
peergroup api version 2
change verbose to use_progressbars, move option to config file
commit missing file, oopsy
api change for peergroup
add comment about extending this basic testcase
I finally figured out why this function was so unintuitive to me. Hence
indicates a time in the future, and this function operates on a time in the
past.
hostkeys timing
remove trialid from gauger data
compile warning on certain compilers.
fix warning
warning
warning
record peer start duration
log gauger data, differentiate between hostkey generation and hostkey
loading/copying
remove log messages if glpk not installed
bad value
cast issues
katie's public key
a commit from nate, amazing
arg
speeling
@brief
formatting
update instead of set stats values
update instead of set stats values
how did this ever work?
test stop condition, enough requests complete
fix memory leak
fix memory leak
fix for assertion failure on topology daemon shutdown
utf8 characters
remove disconnects which are clearly being used improperly. Not a
permanent commit, just to check if BB testcases will now pass
bad assertion (i think)
extra 0 after endif
additional stats on peers connected
connected peer lookup
list connections binary source
api call for connected peer lookup by peer id
initial nse commit
debug flag
expensive tests flag
<none>
tool to list transport level connections
remove extra PeerIdentity (unused, wasted space)
also print inbound addresses
also print inbound addresses
add set delay for TCP send and receive welcome
prefer inbound addresses
transport minor changes
testing HACK to see if it helps connection issues, NOT A FIX
nse changes
nse updates
header files
changes
forward messages
print topology
print topology to file (dot)
remove crap
<none>
nse changes
coverity fix
coverity fix
changes
changes
new function
changes
unfinished function
nse changes
ten minute timeout
larger history, connect limit
connection limits
default topology bug
stuff
clock skew, testing half
clockskew code
clock skew lib
document
better printing
print actual estimate
print actual estimate
churn a service
normal distribution
set to null
churn fixes
timeout too short... adaptive? backoff?
changes
add math lib, remove noinst temporarily
duplicate
fixes for remote peer testing
load hosts from file, implementation in testing
remote host testing
remote host testing
remote host testing
remote host testing
remote host testing
remove send on connect
remove send on connect
always print number of connections
gnunet-arm may take longer to run, assume all is well
debug
gnunet-arm is too smart for its own good
start service
undo
undo
undo
undo
good
redo send on connect
timestamps
timestamps
dont remove files
timestamp fix
command line timeout option
use arm timeouts, bad GNUNET_TIME function that will probably get removed
fixing timeouts with arm, working well now (:
remove verbose logging message
don't break unless debugging
fix
unixpath causing trouble for distributed testing
oops
set clock offset
blergh
add libm to makefile
debugging dv/transport problems
-fix
- Fix script for buildbots that run as root (no sudo provided)
Nils Durner (175):
rm
cp
make ./bootstrap work
fix MinGW
ignore Eclipse project files
<none>
MinGW
poll PIDs for status information
MinGW
break infinite loop
events
up
up
more complete DISK API
typos
translate GNUNET_DISK_PERM_xxx
match older GNUNET_DISK_file_xxx functions
GNUNET_IO_handle => GNUNET_DISK_handle
GNUNET_DISK_OPEN_READ
-atoo
fix
fix MinGW
revive vector container
naming
rename NETWORK_socket API to NETWORK_connection (to make room for the new
actual NET API)
low level network API
return error if filename could not be expanded
MinGW fix
fix file open flag mapping on Windows
use new constant name
support relative paths for GNUNET_DISK_get_blocks_available() on MinGW
expect platform specific slashes
don't fail if conditional log output didn't actually happen
MinGW
don't fail in testcases
naming
fix
fix
naming
docs & GNUnet-style return codes
return codes
fix return code
avoid c++ keyword
Windows returns EWOULDBLOCK instead of EINPROGRESS
locking
fix
fix
fix
set explicit include path for buildbot on MinGW
use correct PRNG initializer
srandom II
use singly linked lists instead of vectors
constants collide with GDI
redo r8922
<none>
fixes
don't crash
leak
needed
mingw
always free
fix MinGW
incompatible with MinGW
fix crash
fix
hard memory limit for safe buildbot'ing on Windows
fix leak, improve API
[W32] don't busy wait if there are no sockets to wait on, activate error
check from other platforms
up
GNUNET_CONTAINER_slist_append
various fixes
increasing timeouts to accommodate higher select() latencies on Windows
set up logging
Windows lacks /dev/full
MinGW fixes
stpcpy()
fix MinGW
bye pthreads
fix MinGW
pass interface address to enumeration callback
improve logging
improved logging
just warn about parser errors (done in process_acl?(); especially
important on systems that don't support IPv6)
MinGW
- VERBOSE
better error msg
don't require gnunet-service-resolver in path
fix
fix
fix MinGW, more precise type/error code handling
MinGW fixes, thanks to LRN on #gnunet
<none>
define target in any case
MinGW
Linux
use search path
don't bypass GNUnet IO
MinGW pipes
move platform specific stuff to platform.h
ndurner
up
Windows impl of GNUNET_DISK_file_get_identifiers, untested
trying to fix python check for non-python systems
fix mem issues, add ".exe" extension to path (fixes #1570, thanks to LRN
on #gnunet)
report signal() errors, thanks to LRN (#1571)
lower case mkdir_p
HKDF (does not work yet)
fix
minor
The zeroed K(i)-field for K(1) was not included in the RFC (Appendix D,
point 5 of the *revised* (Crypto'2010) paper)
init result buffers
fix
fix
remove debug info
update
debug outpit
support chunked context parameter for HKDF
fixes
add cryptographic seed to pingpong
packed
reintroduce seeds, fix calculations
up
KDF code
use separate keys for message authentication
const
up
dynamic IVs
stronger randomness
stronger randomness
authentication of ciphertexts (+ seed)
use gcry_mpi instead of gmp
GMP removal
remove SHA-512 code, use libgcrypt implementation
authenticated ciphertext only, use seed for individual HMAC keys
style fixes, minor bugfixes
alignment, byte order, discard fake messages faster
style improvements
fix warning
fix warnings
fix warnings
define pid_t as HANDLE
up
up
fix warnings
fix PATH_SEPARATOR
HMAC check
minor improvements
fix
0001617: [PATCH] Fix one mistype and some errors in MINGW-only code
named pipes
EEXIST
fix
improve error reporting
fix MINGW
MinGW
fix MinGW munmap (#1628)
shutdown via control pipe (#0001616)
fix
docu
no AF_UNIX on Windows
Windows fix
strtok_r
disable gauger on Windows for now
cross-compiler changes
MinGW cross-build
MinGW
MinGW
fix
fail if plibc is not available
docu
docu
docu
setup docu
docu
docu: no login shell for gnunet user
Omar Tarabai (151):
SENSOR component inital commit
SENSOR service integration
SENSOR service integration
test sensor definitions
fixed loading sensor definitions
warning fix
warning fix
PEERSTORE initial commit
towards PEERSTORE
towards PEERSTORE file plugin
towards PEERSTORE sqlite plugin
PEERSTORE sqlite plugin store function
bug fix
getting single sensor information
towards sensor scheduling
fix in sensor scheduling
running gnunet-statistics sensors
fix
toward executing sensor processes
executing sensor processes OK
cleanup
towards PEERSTORE api
PEERSTORE api update
update to PEERSTORE api
PEERSTORE api update
PEERSTORE api fix
PEERSTORE cleanup and api test
PEERSTORE api fix
PEERSTORE store function
towards PEERSTORE iterate
warning fix
fix
PEERSTORE api overhaul
reconnect on error
peerstore helper file
added record expiry to sqlite
towards peerstore iterate
peerstore API now uses MQ
minor fixes
minor fixes
peestore: towards iterate functionality
peerstore: towards iterate functionality
peerstore: completed iterate functionality
peerstore: record expiry
warning fix
peerstore: iterate request timeout
few fixes
fix for memory leaks
sensor: disabling sensors
peerstore: towards watch functionality
peerstore: towards watch functionality
peerstore: watch functionality
header file changes
peerstore: fixes in watch functionality
peerstore: memory leak fixes
minor fixes
peerstore: added 'replace' option and other fixes
peerstore: db index + store test
peerstore: additional test cases
peerstore: stress test + minor fix
peerstore: changed sqlite to async operations
using PEERSTORE in SENSOR + fixes in PEERSTORE
peerstore: minor fixes
minor fix
peerstore: fix
peerstore: fix
sensor: fix
new sensors + fixes
additional sensor files
sensor: using data lifetime value
towards sensor analysis functionality
Sensor gaussian model skeleton
DLL instead of multihashmap
sensor: unified numeric datatypes
sensor analysis: gaussian model
initial sensor reporting component code
sensor dashboard (collection point) initial commit
sensor: towards reporting to collection point
sensor: completed reporting to collection point
moved common sensor functionality to a util lib
completed sensordashboard + fixes
sensor lifetime defaults to forever
new sensors + disabled reporting by default
documentation fix
logging more verbose
sensordashboard: received data stored in peerstore
peerstore doesn't shutdown when clients are connected
Added flag to API disconnect method to send pending store requests before
disconnecting. Added a test case for it.
config fix
fs now uses peerstore to store respect value
moving peerstore out of experimental
error handling
towards sensor update functionality
sensor: towards update functionality
peerstore: doxygen
sensor: minor fixes
sensor: towards update functionality
sensor: towards update functionality
sensor: completed update functionality
sensor: fix memory leaks
added a section for PEERSTORE in tutorial
fix
sensor, sensordashboard: doxygen fixes + indentation
peerstore: indentation
sensor: towards p2p anomaly reporting
sensor: towards anomaly reporting
sensor: towards anomaly reporting
minor fix
sensor: merged reporting module
sensor: separate monitoring functions
sensor: starting modules controllable from config
sensor: api changes
sensor: force fake anomaly
sensordashboard: storing received anomaly reports
sensor: test case + fixes
sensor: update to test case + fix
sensor: proof-of-work and signing library functions
sensor: test case for proof-of-work + fixes
sensor: fixes for proof-of-work, test passes now
sensor: proof-of-work ingerated with anomaly reports
peerstore: minor fixes
peerstore: fix
peerstore: fix
sensor: removed interval restriction
sensor: towards profiler
sensor: towards profiler
sensor: towards profiler
peerstore: towards profiler
sensor: towards profiler
sensor: profiler reads topology from file
sensor: towards profiler
fix
sensor: towards profiler
sensor: towards profiler
sensor profiler: disconnect peers functionality
sensor: towards profiler
sensor profiler fix
sensor profiler complete
peerstore: clearer variable name
sensor: update to gaussian model + optimized parameters
fix
sensor: pow benchmarking
sensor: minor fixes
sensor: profiler prompts for peer reconnection
sensor: retry reporting to collection point if fails
sensor: profiler updates
sensor: profiler update
sensor: profiler update
towards fixing #3581
#3581
flush peer respect value on disconnect
Philipp Tölke (351):
init: Makefile
a first simple test-program for the vpn-functionality
ignore files in git
parse ipv6-packets for the vpn-functionality
some improvements in the test-code for the vpn-functionality -print
continously -print the complete hexdump
vpn: pull the pretty-printer out of the packet-file
vpn: prepared to implement tcp
vpn: parse and pretty-print tcp-packets
vpn: general cleanup
vpn: tcp-natting in one direction
vpn: a better way to parse the packets using bitfields
vpn: cleanup of the code
vpn: Only read as much as I can handle
vpn: reported fixme;
vpn: Packet-Structure for udp and dns over ipv6
vpn: Revert "Only read as much as I can handle"
vpn: Missed a ; in the header
vpn: pretty-print udp
Begin implementing the gnunet-vpn-helper
Parse and pretty-print dns
Shuttle data back and forth
start the helper from the daemon, end it correctly
corrected the copyright-disclaimer
communicate with the daemon over a specified protocoll
read the output of the helper
Rename files to fit into the convention
renamed a file to fit into the convention
deleted unused functions
CG could answer my questions
Pretty-Print Packets from the helper
Changed the copyright-notice
bugfixes in gnunet-vpn-helper
only select for write, if it was not possible before
Handle EOF in the stream between daemon and helper
A little test-case
Another little test
remove unsused files
renamed files to fit into the convention
make it possible to restart the helper
restart the helper when something goes wrong
the manpage says that memcpy can never return an error
close the not-used pipe ends
Changed the communication between the vpn-helper and the daemon to use
GNUNET_MessageHeaders
clarifying comments
Use the mst to read packets
only read as much as fits inside of a GNUNET_Message
renamed the vpn-helper to fit into the scheme gnunet-FUNCTION-SUBSYSTEM
Make clang happy
really kill the helper if something went wrong
Pack the network-structs
Streamline the memory-allocation
If the tun-if can't be initialized correctly, close the vpn-helper
corrected the test-case, the test-packet needs to use the
GNUNET_MessageHeader
Give the tun-interface an IPv4-Address
Added IPv4-Packet-Structure
handle ipv4-packets
helper-program to use iptables
fixed the #includes
The beginning of the DNS-Service. Does not work for the moment, because
GNUNET_SERVICE_run does not give me my cls...
Hijacking the DNS should work now.
As requested by mwachs
Fixed a string-length
Get rid of the cls. Use a global.
start the code to send queries to the resolver
Worked through the easy klockwork-issues
removed a few debug-outputs and changed the remaining to GNUNET_log
send the dns-queries from daemon-vpn to service-dns
bug
get rid of alloca()
prepare the dns-pretty-printing for the service-dns
organize the imports
send dns-queries to the service-dns and pretty-print them
Missed two %m yesterday
check whether i really can open file-like things
implemented a way to add an element to the end of a slist
Queue packet sent to the service-dns
minor fixes
answered the question
Error-Checking
debug-output
Use the DLL as queue
close ressources
rename the service in the config-file
removed unnessecary debug output of the helper
hijack the DNS-Packets in another way
Send DNS-Queries on to the original destination
print all fields of a dns-rr
Receive DNS-Responses and print them
save a bit of state
send the answer from dns to the vpn
stub to send the dns-answer to the helper
-Stop the vpn-helper if an invalid message arrives -Note that restarting
the helper demands restarting the hijacker, too
I think I have not completely understood how this bitfield-stuff works.
Hijacking DNS-Traffic works as expected.
Handle the case, that DNS-Answers come faster than the helper can accept
them
Remove unnecessary DEBUG-Output
Remove unnecessary comments
add clarifying comment to a magic number
pull a function I will need for service-dns out of pretty-print
split the dns-datastructure
parse a complete dns-packet-struct to another struct so as to completly
have it in memory
Make the packet-pretty-printing more usable (concise)
for debugging, print the dns-packets running through the service
Strings are char* not unsigned char *
Detect and log queries for .gnunet.-Domains
Reconnect to the service-dns if the connection is lost (or at the start
of gnunet for whatever reason)
Disconnect from service-dns only when connected
fixes
added my ssh-key
valgrind was crying
if we have no connection to the service-dns, don't queue
template for the dns-block
prototype for a dns-block
hash and validate a dns-block
Fix compile-errors (with -Werror)
If I hashed the whole block I could not retrieve it without knowing it...
lookup .gnunet-addresses in the dht
The plugin was named incorrectly
pyblish a dns-record for testing
testing-config
Fix the hashing
Only reschedule after an hour.
integers are %d, not %s!
allow the service-dns to answer not only with an unchanged dns-packet,
but also with other types of answer
send an dns-response from the dht
Ratelimit the starting of the vpn-helper
less magic numbers
Send answers from the DHT to the daemon-vpn
renamed file as per policy
modified the DNS_Record-block
Restart the hijack, if the tun was down
Answer questions for .gnunet with an ipv6-address
remove unneeded field
get rid of the unneeded "closure"-struct
comment and indent gnunet-daemon-vpn
comment and indent gnunet-service-dns
Fixes and documentation
Only start the hijack if the vpn is running
Answer PTR-Requests (mock-version)
Mem-Leak
This bitfield-stuff does not work as I expect
Answer an icmp-request
Fix a uninitialized-warning
abort is called break
NULL-Pointer
Fix memory-leak (and superfluous copy)
Refactored the creation of fake-addresses
new Struct (it will be saved to the hashmap)
Store mappings into the hashmap
Answer Reverse-DNS with the real data
fold a bit
Turn the IP-Address around in the hash.
send UDP-Packets to the peer offering the service
Send the UDP traffic out over a socket and relay the answers back to the
requesting peer
rudimentary connection-tracking
Tell core that we want to have this packet delivered
Remove Debug-Output
Patch by Marenz: Allow parallel build (make -j 2)
test-case for cores send-to-self-feature
let the send-to-self-test fail if it fails...
A few static-analysis-changes
split up daemon-vpn in multiple files, remove exit-functionality
fix two "invalid read"s
first shot at gnunet-daemon-exit; not tested yet
debug the core-send-to-self
Mockup for the mesh-api
changes to the mesh-api
let -vpn and -exit use mesh instead of core
dependency
fix usage of mesh_api assert invariant in core_api fix mesh_api
fixed doxygen warning
cleaned up the suid-program
fix build on freebsd
Documented, cleaned and indented gnunet-helper-vpn
get the ip-addresses from the command-line
initialize variables
read ip-addresses from configuration
create the tun-interfaces with a name choosen by the user
refactor the interface to the helper a bit to make it easier to use from
-exit
brodski: remove warnings
Yes, I know.
actually, no, I don't
and it gets deeper and deeper
I should stop commiting now.
fix memory-leaks
debug DNS-DHT
Use the configures IP-addresses correctly
Let the -exit use an own tun and not socks
bug
Begone foul memory leaks!
get rid of unnecessary error
fix build
fix mem-leaks and other nasty stuff
fix jump to NULL
fix SIGSEGV if dns was slow to start up
Allow the use of all ports and not just 4 per connection
read the dns-config from the file
valgrind-clean
fixed FIXMEs
refactor to ease the handling of tcp
prepare for tcp
tcp from vpn to exit. way back still missing
Send TCP in both directions:
doxy-fixes
Implement connect_by_type.
fix
remove unused file
added TODOs
send hello to self
documentation-change
resolved FIXME to a comment
removed unneeded FIXME
FIXMEs
FIXMEs
remove old comment
put in FIXME as reminder to myself
add TODO
indent
save state in heap
make the mock mesh perhaps not working, but not crashing either
Fix #1676, take one byte more in data-segment...
Send queries to remote peers
cleanup
make the service-dns ready to receive answers to queries sent over from
remote peers
add TODO
add TODOs
fail more gracefully
remove IP-Mappings after some time
make TODO better grepable
better hashing for the connection-tracking in the -exit
Better hashing for the services in -exit
remove resolved TODO
handle the reception of dns-answers in daemon-vpn
fix compile-errors
begin writing user- and developer-documentation
send packets to remote hosts
fix TODO
send queries on to the internet and send the replies back
let -exit receive packets from the mesh
commented
specify application-types
send packets back from the exit to the -vpn
handle traffic from remotes hosts coming back to daemon-vpn
indentation
better portability
fix bug (invariant) ;
clear stack-memory
remove unused variable
clarifying comment
update the cost later
add debug-output
fix bug (invariant) ;
pack structs that get sent over the wire
use app-types, not msg-types for by-type
fix bugs found by compiling with -Wextra
Stop gcc from treating unused arguments as error when compiling with
-Wextra
Added Email-Address that should persist over the next years
When connecting to mesh, give your app-types!
copy only the dns-part
debug the mock-mesh
document an assertion
make the exit configurable;
read the service-conf from the global config-file
allow more than one mesh-hello-packet to persist
fix doc
build the mesh-hello correctly
receive the mesh-hello correctly
Correctly honor the PROVIDE_EXIT-option
Do not send too big mesh-hello-messages
Mesh calls me with a NULL peer to terminate the list of peers
print peer-names correctly;
make the log output shorter
Do not connect to self when you find that another peer can handle an
application-type
debug reception of dns-answers
debug the vpn
default-config for vpn
load the dns-block-plugin as default
debug fds
try to fix the helper-startup and so on
invariant
only start one helper
comments
do not use unused application-types
save the udp and tcp tunnels
save the dns-tunnel
when putting small shorts in nbo into chars, flip the nibbles!
debug the dns
debug vpn
reverse logic
how good to be working on intel
do not allow more than one notify_transmit_ready per tunnel
queue transmits to tunnels
fix a bug in -dns exposed by kernel 2.6.39.3
begin resolving A-records
reply ipv4-icmp
send ipv4 through gnunet
tcp is not udp!
let -vpn send back v4 packets
allow build once more
close the dns when the user presses ^C
memory-leaks
really reuse the dns-tunnel
Revert "really reuse the dns-tunnel"
Revert "save the dns-tunnel"
bug
Use the right list at the right time.
Debug the connection-establishment in DNS
Bug: fill in the peer-field of the tunnel
Bug. fill the complete IP-Address
saner defaults
cast
prevent double free
double-free
double free
send a hello-packet for the mesh regularily
documentation
do a dns-lookup to find the address of services other than localhost{4,6}
fix bug in -vpn
bug
compile before commiting
make the dns-resolver used for dns-exit configurable
cleanup doxygen-errors
no more hijacker-zombies
we do not want to crash if somebody leaves the net
do not notify more than one transmit at a time
stop the dht as fast as possible
bug
core gives a buf of NULL to indicate errors
send queries as both A and AAAA
allow "unparsing" of dns-packets
handle dns over ipv6 and 6-to-4 and 4-to-6
bug
bug
bug
handle !A and !AAAA better
boolean algebra is difficult
do not end the helper if no v6-address could be set
nbo strikes again
*really* do A/AAAA
i simply do not learn
bug
bugs
bugs
save at the tunnel whether this is an ipv4 or ipv6 connection
gnunet-daemon-exit uses the new mesh
gnunet-daemon-vpn uses the new mesh
gnunet-service-dns uses the new mesh
handle client connections
cancel any pending request to a client that just disconnects
documentation for developers/performance-metrics
clarification
Safey A.Halim (42):
* test_peer.c is a test case for peer.c * Changes in configuration.c and
gnunet_configuration_lib.h for dumping configuration diffs between the default
configuration object and an edited configuration.
<none>
<none>
Adding test cases for "Configuration Diffs writing"
-> Using GNUNET_DISK_mktemp for the creation of temporary configuration
file -> Test Configuration Diffs code refactoring.
Enabling OS X Panther
Arm Exponential backoff
Coverage for functions: peer.c/GNUNET_PEER_resolve() and
peer.c/GNUNET_PEER_decrement_rcs()
Test case for arm exponential backoff (Incomplete)
test arm exponential backoff
gnunet service manager creates listening sockets for services having port
fields in their configuration entries and runs those services whenever clients
try to connect to them
safey ssh public key
Adding "Monkey", GNUnet module for automatic debugging. Experimental code
so far.
Configuration file for monkey test case
Monkey sends e-mails reporting bugs detected using gdbmi C.jj grammar for
C
regenrate: shell script generating a Java Source Tree (using JTB and
javacc) from C grammar file in $MAINDIR/src/monkey. gtb: shell script that runs
regenerate script. GNUnetParser.java: should replace CParser.java resulted in
from running regenerate script. (The program is responsible for parsing GNUnet
source code: currently no expression database is generated out of it).
Monkey is able to either send emails for detected crashes or dump its
output in a text file. test_monkey_npe.sh: Verifies Monkey's detection of Null
Pointer Exceptions.
Checkin for Seaspider and Parser generator script.
jtb script checks for the existence of the syntax tree hierarchy before
creating it.
Deleting generated files
SeaSpider's parser, depth first visitor, and LineNumberInfo class
<none>
Handle to deal with Expression Database
Seaspider accepts any combination of source files, source directories.
One of the arguments should be the path of the Expressions Database file (a .db
file)
Improving Seaspider's recursive method.
Monkey parse command arguments using gnunet_getopt_lib. Adding a man page
for Monkey
A simple C program to test Seaspider.
Adding an inline function to Seaspider test.
Adding a macro to Seaspider test.
Monkey Expression Database API
Relocating SeaspiderTest.c
Expression cleanup.
Expression Database API, and Action API (without implementation).
Monkey Action API.
codesonar problems fixed
Codesonar fixes for gdbmi library
Action API fixed
Action API detects segmentation fault
Monkey supports bad memory access using Valgrind
Monkey supports Assertions, and inspects user-defined expressions
Memory leak problem fixed when running Monkey with Valgrind. "assert"
debugging supported. Watch a user defined expression supported (crypto_crc
performance analysis test is not working properly)
Fixing Valgrind report
Sree Harsha Totakura (1284):
added my ssh key
added API definitions for stream library
fixed missing doc comment
syntax
stream P2P protocol message specification
generic type for read data
test case for stream API
removed ambigious description
added shutdown call in testcase
added half-closed shutdown test
refined test cases
- protocol states
- moved stream message types to gnunet_protocols.h
-added configuration parameter
-stream api implementation skeleton(incomplete)
-fixed compilation errors
-added message sequencing
-protocol upto ESTABLISHED state
-added socket write handle
-added write operation
-fixed ntoh64 to GNUNET_ntohll
-added data message handling
added ack sending
-fixed compile bug
-refined semantics for STREAM_write
-read operation sketch
-documentation fixes
-receive buffer (re)allocation in handle_data
-added seperate io handles for read and write
-copy buffer and STREAM_read(incomplete)
-fixed relative boundaries in stream read
-added prepare_buffer_for_read
-added read io timeout
-removed copy buffer and prepare_data_for_read
-modified testcase according to modified interface
-bugfixes and testcases
-identify sockets deriving from lsocket
-debugging
Data message retransmissions
more assertions
peer interning
corrected HELLO_ACK message size
fixed byte conversion bugs
fixed retransmission task
using GNUNET_STREAM_ACK_BITMAP_BIT_LENGTH
fixed listen callback to happen after reaching ESTABLISHED state
fixed read timeout problem and added ack sending incase of ignored data
messages
fixed read packets removal after read processor, byte ordering bugs,
ack_bitmap handling in handle_ack ack_task cancelling in socket close
bugfix
fixed compile error from r20729
- de-duplication
-added write io cancellation
-added shutdown handle and cancel
-shutdown completion callback
-added STREAM_shutdown
-added STREAM_shutdown & test case for shutdown
-doxygen
-retransmission of close messages
-stream connection halfclose and test cases
-logging in mesh peer disconnect handler
-test cases renaming
-svn ignore
-makefile for new test_stream_local (commented)
-fixed mesh application types array and makefile
-fixed wrong assertion
-task cancellations in tunnel cleaner
-added test_stream_local.c to standard checks
-added error status GNUNET_STREAM_BROKEN
-added interface for lockmanager
-modified lockmanager and added protocol messages
-lockmanager build skeleton
-integrated lockmanager build
-extended lockmanager api
-naming convention
-test cases setup for lockmanager
-hashmap for storing lockingrequests in lockmanager
-hashmap for storing lockingrequests in lockmanager
-lock success callback
-fixed memory leak
-logging conventions
-doxygen
added list processing
lock acquire and release
processing upon lock release
lockmanager with new datastructure
NULL check on disconnect handler
logging and indentation
inlining of helper functions
-reverted to 21440
peer ids in logging and indentation
clean shutdown in lockmanager, test case for lock release and message
format checks for incoming msg in lockmanager API
change in API documentation and function for finding lockingRequest in
hashmap
warnings
removed local function
-removed local function
-doxygen
server crash test case
handling replies continuously from server
-added message queue
-doc header
-fixed freed memory access bug
-valgrind checks
-added lockmanager to experimental build
-fixed warnings from lockmanager service during tranport tests
testing system
-added lockmanager to main build
fixed segmentation fault due to missing GNUNET_STREAM_OPTION_END
testing port checking (incomplete)
testing port reservation
-freeing addrinfo
-fixed infinite loop
port reservation and test cases
port reservation - release
-logging convention
- AI_NUMERICSERV windows hack
- hostkeyfile reading
-fixed bug 2379
-configuration from template
-updated testing new lib
-reverted GNUNET_TESTING_configuration_create()
-modified hostkeys management
removed double rsa key free
-GNUNET_TESTING_peer_configure
-peer start, stop, destroy
-GNUNET_TESTING_service_run
refined GNUNET_TESTING_service_run
comments and fixed NULL check for tm
test case for peer startup in new testing library
-added testcase file
-fixed memory leaks and added windows support for temp dirs
-fixing windows _mktemp_s
-test case for service_startup()
-arm now starts childs with custom config
-network message struct
-testbed message types
-updated documentation
-cleaner code
-new test case for multiple packet transmission using stream
testbed service build system
-implemented some functions from testbed_api_hosts
fixed static to non static
-testbed controller startup
-handler for INIT messages in testbed server
-added host_run_
-Helper handle wrapper
handling add host
stream misc fixing
remvod verbose debugging in stream api; fixed warning in stream_big test
case
added test_stream_big to make check tests
-fixed unused variable
-stream testing hooks and options
-testcase for sequence wrap around effect and sequence handling fixes in
ack handling
-stream uses locks from lockmanager
-doxygen
testbed_api_hosts test cases
-testbed service reply queue and host success reply
-changes required for acquire retry
-restrictions on using GNUNET_LOCKMANAGER_cancel_request()
-removed a restriction for calling GNUNET_LOCKMANAGER_cancel_request
-lockmanager acquire retry
-fixed reply handling
-fixing #2442
-fixes for api change in testing library
-fixing compile warnings
-fixed test compile warning
-fixed assertion failure when GNUNET_TESTING_peer_shutdown is called in
GNUNET_TESTING_peer/service_run
lockmanager testcase new testing library
-typo
-stream testcases with new testing library
-stream listen timeout option
-fixing #2440 - cancel messages if RELEASE called when ACQUIRE still in
queue
-doxygen
-added stream listen success callback
-added stream read io cancel
- removed test_stream_api.c
-service sharing message handling
testbed host registration
-testbed host registration cancel
configuration serialization
-configuration deserialization
-control retransmission for HELLO and HELLO_ACK
-control retransmission handling
-configuration serialization size_t for size
-coverity
testbed api: controller link message
-host registration with controller
link controller handler
routing
-linking libz
added TESTING_IGNORE_KEYS
ignoring ACCEPT_FROM for transport plugins while testing
-LCFContext
-tracking lcf_proc_task
-clearing LCFContext queue on shutdown
-ADVERTISED_PORT updated with value of PORT if present and not present in
TESTING_IGNORE_KEYS
-controller startup, connect and duals
-TESTBED controller_link_2
-TESTBED controller_link_2 header
-doxygen
direct links to slave controllers
routing controller link messages
fixes
testbed api test case and fixes
function exports from testbed_api
testbed peer_create ()
-peer create message handler
testing doc
operation and peer_destroy
-added handler for peer_destroy
-client handle operation success
-removed static operations and operation queue
-peer create bug fixes and test case
-inlined peer destroy in testcase
-fixing #2493
more fixes
-debugging for stdin close
-reduced test timeout time
-fixing 2494
fixes for controller_start() API changes
implemented controller error callback
testbed helper
testbed helper testcase
testbed helper init message
relaxed time for shutdown
doc
init message handling
svn ignore
config localization and service startup
doc
extended HELPER api to notify when child crashes
separated helper init message
helper integration to GNUNET_TESTBED_host_run_() todo: add controller
start callback with the configuration with which the controller is started
helper exception callback lesser parameters
helper reply with modified config
fix compile
fixed test_gnunet_testbed_helper
mesh compile
libz linking
implemented controller_start success callback
fix
fix; modified testcase for test_testbed_api; valigrind suppression for
zlib
API internal function call changes
testbed api peer start
peer create handling at service
peer_stop and service handle for peerstop message: TODO: handler for peer
stop success
peer create callback
decide btw ipv4 or ipv6
peer start and stop
fix mem leak
fix NULL memory access
-doxygen
removed redundant info request
-get peer config & info
-handling peer get configuration at client
message checks at client receive side
hack for unix domain sockets
testcases for peer get information
fixed mem leaks with operations
-fixes mem leaks
-suppressions for logging and others
-helper service shutdown fixes; overlay connect message
helper shutdown on stdin close
removed slave context
fixed incorrect realloc size
fixed LCF forwarding
extensive message checking
fixes
testbed operations
testbed operations
fixes
removed redundant cls(data)
-removed operation type
peer create with new operations handling
doc
peer destroy with new operations handling
peer start/stop with new operations handling
peer info with new operations handling
state information for OperationContext
improved overlay connect handling
refined overlay connect and extended test case
hacking overlay connect
working version of peer connect
fixed memory leak with HELLO
cleaner peer destory
overlay connect operation failure reporting
controller hostname in init
towards peer create forwarding
Local and remote peers
api forward operation message
forward operation msg cancel
implemented peer create operation forwarding
fixes
fixes and removed slave2 from controller link test
fixes
fixes
controller link as operation
added controller link testcase to default tests
modified link controller forwarding to use forwarded operations
removed controller link testcase from default tests
added checks for passwordless ssh login during tests
added ssh batch mode param
added testbed testing ssh key
peer create fowarding with tests
peer destroy forwarding with tests
peer start/stop forwarding with tests
operation queue renaming
towards implementing service_connect
doxygen fixes
service connect checkpoint save
peer service connect and its test case
more doc
checkpoint save for testbed_run
GNUNET_TESTBED_run to return a handle which can be used for clean shutdown
GNUNET_TESTBED_run and test cases
added the forgotten test case
fixes
peer_list handling fixes
peer_create starts with 0
api handler for operation fail event
Removed RunHandle return from GNUNET_TESTBED_run
stop peers before destroying them
added testbed_run test to default checks
adhering to indentation standard
renamed gnunet-testbed-helper to gnunet-helper-testbed
indentation standard
check if peer has already been stopped before attempting to stop it again
restructuring and checks
implemented GNUNET_TESTBED_test_run
tests for testbed_test_run and some fixes
added new testcase to default checks
repect event mask before calling event callback
removed testbed testing key
input checks
doc
controller callback added in GNUNET_TESTBED_test_run
fix
sane logging and code
new stream testcase using testbed
sane timeout
enable offline checks
new stream/testbed testcase into defaults
save
new testbed interface
changed testbed API to include convenience callbacks
service connect callback instead of operation completion callback
fix
suppressions for mesh
migrate stream 2peers test to testbed
test_stream_2peers_halfclose to testbed
-doxygen
implemented continuations to testbed peer start/stop API
renamed GNUNET_MESSAGE_TYPE_TESTBED_OPERATIONEVENT to
GNUNET_MESSAGE_TYPE_TESTBED_OPERATIONFAILEVENT
error reporting in service_connect
cleanup when operations fail
stream option to set packet size
rename MAX_PACKET_SIZE option to MAX_PAYLOAD_SIZE
reduced payload size
stream speedup fixes
stub for stream performance tests
scaffolding
more scaffolding
stream performance - implemented 1 hop throughput calculation
reasonable buffer for performance measurements
fix
fixes
removed dead code
tests to show overlay_connect on already connected peers doesn't fail
file end comment
added work ssh key
disambiguating local overlay connect
added new message for inter-host overlay connects
removed OCC state
more checks on while working on overlay connect request
add to routing if delegated host is subordinate
stop peers before destroying
messages for inter-controller overlay connect
unstable checkpoint save
1/2 part - inter-host overlay connect
handler for overlay request connect
coverity fixes
extended rocc
rename
new test case (not working)
fixes and working testcase for 2peers_2controllers
doxygen fixes
checks for working local SSH
test_testbed_api_2peers_2controllers to default tests
fixes
appropriate logging string
sequence difference calculation fixes while processing DATA_ACKs
case to show lowered throughtput if payload size if 64000
fixing valgrind unconditional jump error
1 hop performance testing
fixing stream data retransmissions
measuring uplink and downlink
doc
overlay connect forwarding when peer1 is not local
added peer2_host field in overlay connect message
overlay connect using lateral connections
fixes
doc
towards generic topology creation
implemented generic topology overlay connections
topology test case
svn ignore
fixes #2568
towards starting 3 controllers on a single machine
removed getaddrinfo
rename
clean exit while aborting
fixes for #2570
removing duplicate specifiers
removed duplicate specifiers
more fixes for #2570
disabling VPN to remove verbose warnings for VPN
fixing #2571
doc
get slave config
checkpoint save
fixes #2573
remove duplicate specifier
slave get config client part
rename to extract_config_()
clean handling of slave
slave get config service implementation
allow lateral connections from other controllers
fixing #2574
test case for lateral connections between controllers
moved read_io task and its corresponding timeout task to read handle
removed session id
doc
test case for overlay connect via lateral links
removed fixme
added operation closure to GNUNET_TESTBED_controller_link()
generate keys to use 2048 bit rsa
testing now includes valid hostname rewriting
working test on inifiniband cluster
exponential delay in offering HELLOs
fail nicely
forwarding overlay request connect
added get peer config forwarding
towards on-demand configuration retrival for remote controllers during
overlay connect operations
fixing compile
towards handling suboperations during overlay connect
operation id generation from controller host id and internal counter
checkpoint save
controller linking with host ids
overlay connect suboperations
working overlay connect with implicit suboperations
disabling hostkey checking for localhost
testbed profiler
using header-defined hostkeyfilesize
host loading from file
fixes for invalid reads
fixes
host file support for testbed profiler
fixing scanf failures
tolerate stopped peers while shutting down testbed_run
fix w32 build
no auto allocation during scanf
send self config during overlay connect
added infiniband hosts
host registrations in profiler
correct ips
fixes
slave startup
fixes
peer create
fix
removed peer destroy. May leak memory, but anyway we need that till the
end
fixes
towards peer create
doxygen
operation closure in peer start
peer start
corrected profiling time measurement for peer create
Random graph topology
peer linking
output status for connects
timeouts during forwarded overlay connects
logging and fixes
reducing peers to avoid timeouts
peer to use slave handle
host registration queues in slave handles
checkpoint save
a decent way to auto link controllers during overlay connects
fixed memory leak
fixs for crash while adding slaves with id greater than 9
new operation queue for limiting overlay connects
stop peer before destroying it
more cluster nodes
peer destroy state
-topology test case relaxed to accommodated overlay timeouts -bounded
overlay connects -topology API to retry on timeouts -test case configuration to
include bounds on topology -cleanup of occs during shutdown
use controller's host when slave host is NULL
- relaxed to accommodate overlay linking timeouts - increased bound on
parallel overlay connects
print statistics
parameterized how many failures to tolerate
- script to clean up testbed run when it crashes - lowered parallel
overlay connect operations - added more infiniband cluster nodes
testbed results
remove remote peers upon destroy
removed redundancy
cleanup rocc
counting references in overlay connect context
reverted to version 24399
fix
doxygen
removed duplicate definition
tracking forwarded operations as part of link controllers operation
tracking forwarded operations
simplify
option for disabling auto retry
disable auto retry
remove username part
less verbose
fix
fix
allow single host deployments
clique topology
profiler topology option
fixes
fix
fix
doc
fix
trying to fix win32 build
fix for crashing service while shutting down when peer create fails
exit during errors too
fixes
removed exponential backoff while offering HELLO
multiple operation queues for an operation
removed GNUNET_TESTBED_operation_cancel
starting testing processes with a low priority to not overwhelm testing
hosts
for starting up non gnunet helpers (SSH for example)
locate helpers in libexec
fixes
fix crashes when aborting testing drivers
dead code
fixes failing tests
check all available addresses
fix for port-conflicts while starting multiple controllers on same host
fixing testing failures on pi
fix
topology option support for GNUNET_TESTBED_run() via configuration
fix
fix crashing checks
fix
fix test on W32
fix crashing test
fix memleak
reverting back to fix failing testbed tests due to late startup of peers
ignore gtags files
fix to keep testing running
fix
maintain argv for helpers
added testing configuration files
do not start gns-helper-service even on windows
merging configs
merging configs
reverted test configurations to include ATS
coverity fixes
coverity fix
coverity fixes
changing to fixed number of packets instead of fixed size test data
dead code
including sample_hosts.txt to EXTRA_DIST
nss uninstall script to extra dist
- testing write access
check for existing directories in reverse order
- typo
print help for timeout argument
reduce number of packets
- remove stale file
- reverting unintended modifications to po files
Parameterized directory listing permission check
making readable check optional
- fix
-adding rever order checking of directories
- fix doc
adopting to new transport API
save tests from crashing
-exit instead of return
-reverted to allow cleanup
-fixes
- send HELLOs aggressively
- adopting to new transport API in rocc
-fix build
-fixed warnings from test run
- increse timeouts and fixes
implemented line topology
test case for ring topology
implementing small world ring topology
Options for 2D Torus
- SD calc
- try connect as task
-fix crashing test
-fix slave crashes
- remove redundant transport handles
-fixes
- try connect as task in rocc
kill all peers first and wait for them later
dynamically adjustable operation queues
-reducing testing peers and links
- towards testbed_create
- doc
adaptive parallel overlay connects
- test config fixes for adaptive overlay connects
2D Torus topology
- testcase for 2d torus and fixes
- de-duplication
small world topology
fix memleak
use statistics
- towards cleaner shutdowns
- towards cleaner shutdowns
cleanup write handle upon shutdown
- warn upon unclean socket close
- fixes
- shutdown other socket also
- merged r25055
scale free topology
extended overlay configure topology to return max connections
- topology name handling
topology loading from file
test case for scale free routing
seperate timeouts for MESH retries
- use GNUNET_TIME_UNIT_ZERO
exponential delay for data retransmissions
- simplify
- update bitmap not only on ACK's base number but also by looking into
its bitmap
configure option for using IBM LoadLeveler for scheduling testbed slaves
on SuperMUC
- testing build system - WITH_LL
hosts file support in GNUNET_TESTBED_run() - fixes 2667
making GNUNET_TESTBED_is_host_compatible() asynchronous
- fixes
display and update number of hosts checked
- fix
- fix
- fix
host compatibility check in GNUNET_TESTBED_run() and fixes
- fix
- fix
- check the host compatibility check status
- check the host compatibility check status
- remove duplicate functions
- mark test as skipped
- mark test as skipped
- fix
- use localhost without chcking when hosts file is NULL
- fix compiler warning
- Loadleveler code
- MPI test program
- fork within MPI
- remove MPI code
- reverting to include MPI code
- replacing fork with exec
- exec arbitrary commands
- cleanup
fix 2672
- warning when trying to read when a read handle is already present
allow calling GNUNET_STREAM_read() from DataProcessor callback
- fix 2672
- rename
- doxygen
GNUNET_TESTBED_test_run to return status
fix 2689
- fix compile
- fixes
- de-duplicate
- logging
- log peer id
- fixes
- osx fixes
- assert hard
- logging
- logging
- logging
- fix
- use pipe control only on WINDOWS
- log pipe name while receiving signal
- fix compile issue
- rename TOPOLOGY_FILE to OVERLAY_TOPOLOGY_FILE and remove
GNUNET_TESTBED_create() and its friends
- warn if connecting a peer to itself - peer indices in overlay topology
file start with index 0
- fix
- fix compile; NSE profile still incomplete
- rename controller_ip to trusted_ip
- doc
- no more explicit overlay linking
- fix
remove disconnecting client from waitling lists of other locks
- fix crash when controller dies
- use GNUNET_array_grow()
- fix crashing testcases on sparcbot
- make operations wait when max_active is decreased
adaptive parallelism for overlay connect operations based on SD
calculations
- doxygen
- fix
linear increase and decrease when SD is within 1 and 2 SD respectively
weigh all readings equally
fix #2720
- give more weight to failure events
-disable statistics
- logging to file
- extra logging
- tag op id in overlay connect debug messages
aggressively persuade transport to connect
- fixes and TIMEOUT increase
cache HELLO's of peers
- mark operation done in case of failure
- call controller callback before operation completion callback; only way
to ensure that GNUNET_TESTBED_operation_done() on
GNUNET_TESTBED_underlay_configure_topology() operations doesn't result in
segfault.
- use GNUNET_TESTBED_overlay_configure_topology(); simplifies and also
ensures that overlay connect operations continue when they are done 1 at a time.
allow customizing remote shell command and the program started by it
- export config from infiniband
- kill irrelavant stuff
- no disk io from peerinfo for normal tests too
- doc - moved write_cont callback towarding beginning of
GNUNET_STREAM_WriteHandle - rename stream_protocol.h to stream.h
- fix parallel build
- flush output
-don't open sockets from consensus dv and lockmanager
use all IP addresses of master controller as trusted ip
fix 2688: profiler now uses GNUNET_TESTBED_run ()
profiler now depends on config file to give topology option
- warn upon unused OVERLAY_RANDOM_LINKS option
- print time durations in debug output
- print summary of established links
operations now can explicitly register how many resources they require
test harder
- move data structures to shared header
separate hello cache as module
- separate overlay connect message handling as module
- rename
- rearrange
- towards operation queues for open file descriptors
- indent
bound overlay connect operation execution to limited number of open file
descriptors
read MAX_OPEN_FDS from configuration
fix 2746
- rename
bound remote overlay connect operation execution to limited number of
open file descriptors
- fix SEGFAULT while registering hosts
- fix
- rename "hello cache" to "cache"
- towards caching transport handles along with HELLO messages
- queue requests for accessing handles from cache
caching transport handles
- fix
- fix
cache transport handles with peer connect notifications
- fixes
- fixes
- fix
function to synchronously check if a peer is connected at CORE level
towards caching CORE handles
- unified notify callback for cached transport and core handles
- deduplicate
- get core handle in core startup callback
- fixes
- fixes and implements caching core handles
- call peer connect notifiy callback only after calling the cache callback
- rename and doc
GNUNET_TRANSPORT_get_hello() to call its callback asynchronously
- fixes
- doc
- fix
-fix
- doc & indent
- check whether an operation gets cleaned before calling secondary
callback
- expire cache entries (expiry set to 'forever' for now)
- cache expire after 15 sec
add completion callback for overlay topology configure functions
- rely on topology completion callback instead of counting completed
overlay connect operations
- fix
- fix crashing tests
- fix
stop helpers by closing stdin
fix 2780: Testbed doesn't clean up properly on error
- fail test without dumping core
separate SD calculations
move parallel overlay connect operation queue from controller handle to
host handle
fix 2793: Testbed does not allow to set TIMEOUT for operations
- fix crashing testcases
Hacks to reduce per-peer load on CPU by increasing transport
retransmission timeouts. This helps to keep the load on testbed node low.
- rename and install to bin
- send SIGTERM to peer and reply with success before waiting on it
-fix mem leak
Assosiate a configuration template while creating hosts. This paves a
way to manage the configurations of controllers running on a host automatically.
M testbed/testbed.h naming consistency include uncompressed config size
- do not use mempcpy()
host locking; warn upon API violation
send configurations of newly started slave controllers as part of
controller link operations
- fix compile
- remove unused
Use a sub operation while forwarding controller link operations. This
helps to propagate the change in a host's configuration along the controller
tree's branch back to root.
checkpoint save
Parsing LoadLeveler allocated hosts and creating testbed hosts from them
version to make updated testbed compile with pre-ecc age code.
reverting back to the previous working version
script to prepare tree for pre-ecc testbed compilation
GNUNET_TESTBED_hosts_load_from_loadleveler() function available only when
--with-ll is specified
get hosts from loadleveler instead of reading them from hostlist
define WITH_LL only if libllapi is found
not using GNUNET_PROGRAM_run() as it is eating away the trailing command
arguments
update configure.ac to HEAD
-fix
finish off with MPI before spawning
- fix #2809
cache peer identities
- fix
- do not create ephemeral keys while testing
- disable DHT
- fix
be verbose while sending error message instead of emitting a warning
- remove unused mpi binaries
Remove MPI code Starts the child process only in one instance (checks for
lowest host ip; and that host ensures all instances co-ordinate via explicit
lock file creation) Implemented basic ARM-like functionality for dealing with
child processes
SuperMUC testbed config
clean the lock file
Don't call gnunet scheduler from signal handler. Instead, write to a
pipe (ARM-style) to run a scheduler task when signal is received.
- run tests faster
- remove unused code
- fix
- supermuc's internal network address
testbed API function to shutdown (stop + destroy) all peers. Client side
functionality only.
- fix (case where peers <= 2)
- revert back; reference: the previous commit also implements
SHUTDOWN_PEERS server-side funtionality
- logging
- fixes
- cleanup peers after shutdown operation
- PO update
- fix
- fix
- fix crash monitoring
- fix crash after handling SHUTDOWN_PEERS due to forwarded operations
using `GST_peer_list'
- fix
- typo
Implement GNUNET_TESTBED_get_statistics() - resolves #2662 Test case for
testing GNUNET_TESTBED_get_statistics() Fix waiting for SHUTDOWN_PEERS reply
when controller crashes
- adding the testbed_api_statistics testcase source
- make ARM tests compile without errors
Call TestMaster callback upon timeout set through configuration fixes
#2833
- assertion check
- assertions for debugging
- fix
- fix
- fix
- fix bad assertion
expose Operation and associated closure in every event
-fix
- fix
- doxygen
use testbed helper functions and lose old code
- working version of NSE profiler
- code reordering - Call GNUNET_SCHEDULER_shutdown() upon errors
- mark statistics service connect operations done when statistics
iteration completes
- sane values
- fix
- fix crashes upon shutdown
- fix
- do not call statistics when some peers are NULL
- warn. Missing peers in the peer array is unusual
- have peers connected in line topology
- make hosts file optional when ENABLE_LL is not set; when it is not
given use localhost only
fix: 2839
- warn when operations are not freed where they could have been freed
-dead code
- style
- generate histograms with unique file names
- fix
- allow getting statistics of a particular subsystem and name in testbed
- request nse statistics only
- exit when an operation fails
- read unique peer-ids from testbed
infiniband nse evaluation config
- test 2d tourus overlay
- fix
- fix DLL removals while cleaning up operation contexts
- fix crashes in testbed_api_testbed upon interrupts
provide how many links succeeded and how many failed in test master
callback from GNUNET_TESTBED_run() and GNUNET_TESTBED_test_run() functions
- reduce number of parallel peer creates to save operation timeouts while
creating peers
current working infiniband hosts
- reduce parallel ops even further
- ARM based peer start stop
support for asynchronous peer start/stop and service start/stop using ARM
API
- doc & improvements
- allow disconnecting from ARM from callbacks
-dead code
- only have asynchronous peer stop
- fix 2851
- revert 26711 and fix 2851 with dirty hack
- release ports while destroying peer
- suppress still reachable blocks from CRYPTO
-fix compile warning
- doxygen
- manage services of peers (client API)
- manage services of peers (server)
- warn on missing cases
- test case for starting/stopping peer's services
- use topology service instead of statistics as statistics may not
terminate when asked for
- fix 2853
- reduce interval
- reduce open file descriptors in processes started through testing
- remove hosts with older kernel
- load monitoring in testbed
- dead code
- log load statistics to file
- disable load statistics on windows
- increase testbed setup timeout
- collect load statistics from testbed
- fix
- increase interval
- revert; do not enable traffic logging by default
- log number of connected peers at core and nse services
- stop NSE service instead of stopping peers
- doxygen
- restructure
- execute operations in the same order as they are queued when possible
- doc
- no starving
- maintain separate queues for operations that are in WAITING, READY, and
STARTED states
- support caching through inactive operations
- more test cases
- doc
- copyright sed -i -r "s/(.*\(C\))(.*)(Christian)/\1 2008--2013 \3/" *.h
*.c
- restructure
- use host's configuration while connecting to controller
- use host's configuration while starting controller on the host
- fix #2862
- towards on-demand controller linking
- doxygen
- sanify
- fix memleak complaints from valgrind
- remove dead code as controller linking operation does not require host
-configuration explicitly
- fix
- fix
- input source for lstlisting in c tutorial
- compile instructions
- fixes
- doc
- fixes
- fix
- revert to using memory mapped host keys file
- close file fd after mapping the file
- stop iterating through waiting operations earlier
- 32-bit key hashmap
- use hashmap for remembering active operation contexts
- use hashmap instead of DLL
- doc
- doxygen
- service sharing in testing
- logger service for aggregating stat data from testbed peers
- testbed logger client API
- doxygen
- simple test case for service sharing in testing deployments
fix #2666: Implement service sharing among peers
- test cases for testbed logger
- Support flushing of buffered data
- test harder by checking the data file size
- fix
- signal termination to peers when controller crashes
- mesh test case to show it doesn't respond to shutdown
- have shutdown task
fix #2877: stream doesn't distinguish between the application ports
- client should open connection after the stream server has started
listening; -variable arguments to GNUNET_STREAM_open() and
GNUNET_STREAM_listen() should -end with GNUNET_STREAM_OPTION_END
- minor fixes
- supermuc config
- showcase failing controllerlink case
- remove unused configuration parameter in
GNUNET_TESTBED_controller_link()
- simplify on-demand linking in overlay connects
- doc
- add GNUNET_TESTING_peer_stop_async_cancel()
fix #2665: peer reconfiguration now implemented
- fix duplicate symbols
- remove unused opaque handle
- generate histograms based on configure option --enable-nse-histogram
- update config
- have hostname part of the log file
- limit hostname as gethostname() has undefined behavior when hostname
exceeds - the given length
- fix parsing of shared service specs
- flush logger data before disconnecting
- have a timeout for cases where flushing takes too long
- complement HELPER API with functions to kill and wait on the helper
process
- avoid linear scan of hosts
-fix
- remove hostname/ip address from INIT message. They will be determined
by the controller automatically
- allow reading hostnames with dashes/hyphens
- fix compile-time warning
- resolve hostname in helper to IP address
- use hostip instead of hostname to avoid dependency on resolver
- use hostname instead of ip addresses
- switch to UDP transport plugin
- sort hosts returned from loadleveler
- supermuc specific island handling code
- merge; compare hostname to determine the node to spawn process
- compile warning fixes
- distribute peers equally among island nodes on SuperMUC
- fix compile
- document options for overlay topology in configuration
- do not wait for keystroke while profiling on supermuc
- expand $ variables in filenames
- fix filtering through event mask
- script for finding the unique and shared memory used by processes
- fallback to using ip addresses for hosts
- revert 27109,27108
- revert 27104
- fix
- style
- include logging of memory statistics on Linux
- fix scanf issues with plibc ignoring '.' before the last '-'
- doc
- log number of processes in the system
- do not create peers on the island controller
- fix memory usage calculation
- compile with CPPFLAGS and LDFLAGS; gcc doesn't take them automatically
- declare GNUNET_EXTRA_LOGGING to make gnunet_common.h usable by sources
not present in source tree
- remove island handling code (we are restricted to 1 island anyway)
- more hosts to infiniband :)
- coverity fixes
- remove dependency on DHT
- assert return values
- dead code; fix doc
- print warnings for failed function calls
- remove dead assignments
- avoid assertion failure if PRIVATE_KEY is not present in PEER section
- load default + given configuration
- fix parallel build
- fix compile
- debugging bind and SO_REUSEADDR
- try with TCP
- use SO_REUSEADDR for SOCK_STREAM type sockets only
- make no distinction between TCP and UDP port while reserving them
- rename PEER_GET_CONFIGURATION to PEER_GET_INFORMATION
- use unix transport plugin
-add option for xgettext to support non-ASCII characters in gettext
strings
- revert to #8472
- fix resource claiming when deferring operations
fix 2893: Move adaptive parallelisation mechanism to operation queues
- allow parsing hosts with optional username and port number
- test with `.' in hostname
- enable substitions
- no need for special parsing of load-leveler generated file since we use
MSH from now
- dead code
- testbed profiler config for experiments on LRZ linux cluster
- negative deviation factor
- fix transport-udp port in-use warning
- considering W32 portability, use tcp transport plugin
- average with failure operation time as overhead
- doxygen
- towards testbed barriers
- doxygen
- more barrier stuff
- more barrier stuff
- barrier stuff
- more barrier code
- propagate barrier cancel message
- fix compile
- barrier wait API
- build system changes for testing barriers API
- fixes
- fix windows build
- remove superfluous files
- ignores
- remove dead code and fixes
- warn and wait for all operations to be completed
- fix memleaks
- doxygen
- give out run handle through which master controller's handle can be
retrieved
- barriers test case; more fixes
- fix compile
- fix failing barriers test case
- barriers doc
- experiment driver propagates barrier crossed notification
- popularity corresponds to edge direction
- implement scale free topology correctly and introduce argument to cap
the number of connections to a peer in the generated topology
- capping node degree in scale free topology
- fix likely possible nonterminating loop for small `cap'
- doc
- testing for getcwd on windows
- use getcwd()
- fix crash when disconnecting from misbehaving client
- undo rev 29322
- fix crash upon testbed service crash or broken connection
- don't append libexec path to absolute paths
- generate test_testbed_api_barriers.conf
- generate test_testbed_api_barriers.conf in the testcase
- do not install mockup-service into libexecdir
- remove python wrapper for barriers test case
- ignores
- gdb python code for DLL iteration
- match according to type
- fixes
- fix hanging barriers test case on windows
- declare dependencies for supporting parallel build (make -j xx)
- profiling looks for testbed service
- fix build
-fix compile warnings and parallel build
- testbed daemon for transport blacklisting
- start the adaptive algorithm with 4 parallel operations
- decrease timeout for individual connection attempts and increase
timeout for testbed setup
- towards seperating connection pooling from cache
- rename to GNUNET_TESTBED_BARRIERSTATUS_*
- fix misc. memleaks
- fix peer reconfiguration when the peer is already stopped
- bail out if peer is not running
- cleanup dead code
- prefix env variable for specifying the tmp dir testbed/testing uses for
dumping generated peer config files
- fix and doc
- doc
- ignore GNUNET_TESTING_PREFIX in testbed helper
- fix testbed crash when controller linking operations fail at controllers
- change --enable-ll to --enable-supermuc - define a system wide
termination signal GNUNET_TERM_SIG. Set this to SIGUSR1 only when building for
SuperMUC
- fix invalid free()
- interactive option to disable waiting on keystroke but wait instead for
termination signal
- use get_value_filename when retrieving filenames and paths
-fix doxygen warning
- use pooled connections from connection pool module
- do not create GNUNET_TEST_HOME directory
- warn upon failed $-expansions
- cleanup; move to hashmap32
- tolerate zero cache size
- fixes
- doc link to hostfile format
- allow localhost testing
- use configuration template for duplicate configurations
- doxygen
-fix leaks with putenv()
fix #2944: modify the adaptive parallelisation algorithm
- fix
- raise a signal after receiving it from the control pipe by calling the
respective callback registered with GNUNET_SIGNAL_handler_install().
- fix
- don't timeout SHUTDOWN_PEERS operation
- support PREFIX strings with arguments
- write PEERID to PATHS so that it can be used for dollar expansion
- move do_start_process to util/
- allow $-expressions anywhere in the given string
- do not use SIGUSR1 on SuperMUC
configure option --disable-autostart to disable auto-startup of services
by ARM
-doxygen
new underlay API
- doxygen
- fix #3187
- add underlay api implementation
- underlay link model message
- fix #3203
- fix testcase crash; debug code added
- fixed missing call to call server_done in server message handler
- decrease timeout
- debug - htonl source line number while calling - do not select a
suspended channel upon receiving the first audio message
- extend mesh audio messag to contain source line number. This is needed
to find the correct client channel to link unreliable audio channel
- reset counters to continue the resumed call
- handle logging clutter properly - make testcase tighter
- call GNUNET_SERVER_receive_done() when returning from message handlers
- call MESH_channel_done() after successfully handling messages
- fix ntohl called on uint16_t
- hint about /help
- do not send a PHONE_HANG_UP message while terminating a caller
connection
- fix free after call to continuation
- fix use after free cases
- fix bad log
- fix overlooked `;' after `if'
- fix #3217
-style
- fix for incorporating day light savings time
- testbed underlay daemon
- read whitelist from database
-fixes
- fix ftbfs
- underlay topology generation
-fixes
- populate underlay db
- use asynchronous to speedup sqlite db writes
- unique constraint
- restrict underlay based on whitelist for now
- fixes
- allow peers to connect in both directions
- testbed underlay testcase
- test properly
-fixes
- utility to list peer ids from a given hostkeys file
- fix
- fixes
- add underlay testcase to `make check`
- fix parallel build
- remove spurious parameter tuple
- latency logging daemon
- only log changed values
fix #3275 with solution from
https://gnunet.org/bugs/view.php?id=3275#c8029
Include test_testbed_api_statistis.conf in dist
Skip testcase when DNS is not available
Allow testcase to run just with loopback interface.
Fix #3294: Decrement call number when the call is cancelled or hung-up.
Revert #32132.
Create UNIX domain sockets as abstract sockets when running in LINUX and
the option USE_ABSTRACT_SOCKETS is present in configuration.
- Declare dependencies. Local dependencies should not be prefixed with
$(top_builddir) path.
- Allocate buffer large enough to contain UNIX_PATH_MAX size pathnames in
case of UNIX sockets.
- create the directory for the UNIX socket file before binding to it.
- Cleanup cache items after removing from the hashmap.
- Add suppressions file for util.
- Add patch for valgrinding testbed
- Allow integration tests to run on loopback.
- Do not set permissions for abstract sockets.
- Remove lockmanager and add abstract sockets support.
- Do not try to delete empty string directories.
Consider the resources from failed operations as overloaded and not use
them until the parallelism is refreshed.
Do not let parallelism fall below the default parallelism value for
adaptive queues.
Allow redirection of STDERR when starting processes.
-Always add measured average operation completion time into history.
- increase history size for the measured average operation execution
times.
- Define and use BUFFER_SIZE for the size of the outgoing buffer.
- Copy PATHS section as the shared service configuration may refer to
options from here.
- Warn upon deleting unsent message. - Send the message immediately if
its size equals to the buffer size.
- remove dead code
- Delay shutdown until all connections are closed.
- Add gitignore
- Expose speedup routines as an internal header.
- Add speedup.h
- Working testbed configuration on SuperMUC.
- Avoid dereferencing undefined pointer when the request status indicates
an error.
- Log NSE histogram to file in addition to testbed-logger when
HISTOGRAM_DIR variable is present in configuration.
- Remove trailing '/' for the directory path.
Add gnunet-nse program to show network size estimates.
Fix 3429: do not print degree of parallelism.
Enable (X)DHT service to be compiled with malicious code.
Add profiler stub for DHT.
Initialize `arg_str1'.
Fix libgcrypt linker flags.
Use Crockford Base32 encoding instead of RFC 4648.
Add DHT profiler.
Print O, L, U instead of 0, 1, V in Crockford encoding.
Print `I' instead of `L'.
Do not cleanup operation queues if they are not empty. Instead, mark
them as expired and clean them in destructor. --This line, and those below,
will be ignored--
Enable tests to run concurrently.
xdht: track topology successor and predecessor neighbours.
- untabify
- English.
-Fixmes
Include libpq-fe.h instead of postgres/libpq-fe.h.
Do not bail out when pg_config is not found.
Try fixing 3522.
More fixes for #3522
Add options for SCALE_FREE topology.
Start PUTs and GETs only from active peers
Fix counting started peers.
Mark operations as done when they are finished.
Reconnect to DHT services for doing GETs after PUTs have been made.
Change the mode in DHT disconnect handler.
- fixes
After completing DHT PUT operations teardown the service connections in a
seperate task.
Do not realloc as this will render the pointers to ActiveContexts invalid.
Make the delays to start PUTs and GETs configurable as program parameters.
Teardown DHT connection after clearing up the ActiveContext
Do not continue if we are going to shutdown.
Wait for Bandwidth statistics to be collected.
Fix #3562: continue test if required number of links have been
established, despite failures as they will be retried.
Reduce the number of peers to 2.
Fix regression caused by SVN 34522
move sample testbed configs to a dir
-add a check to bail out on false signatures
-fix assertion checking the encoded s-expressions
-fix doc
- fix docu
- fix memleak
validate the parsed RSA private key
-fix typo
Introduce function to duplicate RSA public keys
-use defined name
code cleanup
more assertions
fixes
reconnect upon broken server connection
stop asking for random peers if we are shutting down
disable socket inheritance from ARM for the testcases
- check for uint32 value overflow
- set the connection handle to NULL upon disconnecting
-fix #3722
-update docu links
-fix debug message
testbed: disable host key checking along with password based
authentication
-remove startup through valgrind
rps tests: Add fixmes
-deduplicate string constant
Fix broken testbed controllerlink testcase.
introduce more message parsing checks
testbed: do not start unnecessary services
Fix #4718: Do not echo BARRIER_STATUS messages back to the main
controller/test driverAlso add a few FIXMEs to be fixed and fix testcase logic
to accommodate new scheduler shutdown semantics
propagate barrier status messages to sub controllers
disable more unnecessary services in testcases
update documentation
Supriti Singh (141):
incomplete code for finger and friend tables
-Modified struct PeerTrailSetupMessage. -Modified struct
PeerTrailSetupResultMessage. -Added stubs for find_predecessor. -Added comments
to understand the flow.
Implemented logic to randomly choose a friend. Modified struct
TrailPeerList
Handling trail setup result message.
1. Adding an entry in routing table. 2. Peer arithmetic to get finger id.
Finding successor of a peer id.
Computing finger identity using libgcrypt functions.
1.Modified struct PeerTrailSetupMessage and PeerTrailSetupResultMessage.
2.Handling trail setup result message.
Adding function GNUNET_CRYPTO_compute_finger_identity
1. Refactored the code for trail setup and trail setup result. 2. Adding
an entry into finger table.
Fixed memory allocation for peer list.
- Adding self to trail_peer_list - Adding functions to handle concurrent
node joins. - Modified but incomplete find_successor logic. - Added new message
types for concurrent node joins.
- Modified find_successor - added a new field successor flag to identify
successor in finger table.
- verify_successor - 64 bit finger identifier - predecessor flag
- verify successor result - notify new successor
Added a dll to store finger trail in struct FingerInfo
Updated find successor
Merged update_successor and update_predecessor with finger table add
Framework for put/get/monitor
1.Removed GNUNET_CRYPTO_compute_finger_identity 2.Removed myself as the
first element of the trail. 3.Multiple entries in the routing table.
4.Put/Get/Monitor from old DHT added back in clients file.
Framework for trail rejection and threshold on links first phase changes
for get.
changes in get and put code.
Removing current_trail_index, current_destination_type where not
required. Cleaning up handle_get, handle_put, handle_trail code.
Adding correct include file "gnunet_dht_service.h" in neighbours.h
Removing routing table entries corresponding to disconnected peer
Removing compile warnings Adding a new message type, tear down.
Handling trail rejection message, not using a failed trial list.
Handling all the cases while adding a new entry in finger table.
Modifying struct FingerInfo to store two trails to reach to same finger. Adding
code to handle threshold on number of trails through a friend.
refactoring the code for finger_table_add(),
compare_and_update_predecessor()
- Changing send_verify_successor_message to a function call from a
periodic task - Fixing the return value in case I am destination in
find_successor() - Refactoring of finger_table_add()
- Adding a new field in struct PeerTrailTearDownMessage, new_first_friend.
Handling the correct return type from find_successor() in
get,put,get_result functions.
- Fixes in routing table functions
Bug fix
- Adding a new message type,GNUNET_MESSAGE_TYPE_DHT_P2P_ADD_TRAIL -
removing compare_and_update_predecessor() - refactoring trail rejection.
- Adding code to check for congestion and threshold in find_successor() -
Updating trail rejection to get congestion time from congested peer.
Refactoring
Using trail id Finger table add functionality complete Adding a new
message trail compression
Adapting verify successor code to use trail id Linear lookup in
find_successor() Typo fix in comment of trail compression
Removing warning messages
-Adapting trail teardown and add trail messages for trail id.
Fixing issue #0003414
Not passing finger_map_index in trail setup message. Not passing trail
length in any message.
Removing warnings Using array to store finger table Storing routing
entries at endpoints routing table Refactoring the code.
adding my key
Minor fixes
Xvine_DHT: Minor fixes
xvine dht: few fixes
xvine - minor fixes
- Initializing head, tail in datacache struct get_context - Static
analysis bug fixes - Other minor fixes
xvine: minor fixes
xvine:bug fix
-- New message type, GNUNET_MESSAGE_TYPE_XDHT_P2P_PUT and
GNUNET_MESSAGE_TYPE_XDHT_P2P_GET -- Changing name of all the messages of X-Vine
by appending X in front of DHT. -- Not sending trail teardown in case next_hop
or prev_hop is me (remove routing trail due to disconnected peer)
xvine:bug fix
Passing correct argument, is_predecessor to select_closest_peer()
don't send trail teardown in remove_matching_fingers()
xvine: fixes
Handling 'put', 'get' and 'get result' correctly
xvine: bug fix
xvine: bug fixes
xvine: fixes
Xvine: fixes
xvine: Fixes
Use correct key while doing GET. Wait for all PUT and GET requests to
complete.
Close client connection to DHT once GET is succeeded or failed.
Add DEBUG to GET requests. Add replication degree as an option.
xvine: fixes
code cleanup
Removing warnings
xvine:fixes
Calculating index from finger value and my identity using shift operation
XVine: Fixes
x-vine:fixes
x-vine:fixes
xvine:fixes
xvine:fixes
Exponential backoff for find_finger_trail_task
Check for duplicate peers in trail setup and verify successor
- added basic layout for b/w and successor stats collection in profiler -
fixes in neighbours
xvine: fixes
X-Vine:Fixes
- Collect bandwidth statistics in profiler - Get average put/get length
in profiler - Check for GNUNET_CONSTANTS_MAX_ENCRYPTED_MESSAGE_SIZE,instead of
GNUNET_SERVER_MAX_MESSAGE_SIZE
Using statistics to collect current successor of a peer in profiler
Checking if circle is formed.
Xvine: Fixes
xvine:fixes
xvine:fixes
Act Malicious service code
Check that you are not present in trail twice Check that sender peer is a
valid peer
xvine: fixes
Remove the messages of disconnected peer from pending queue.
xvine:fixes
Added correct routing table entry in trail setup
Removed trail compression message. Checking for friend to source in trail
setup.
- Return value in select_closest_peer() instead of pointer - Handle no
trail found in routing table
xvine fixes
xvine:fixes
Check for loop in put and get path
- PUT/GET Code refactor
Fixing warning
Xvine: Fixes
trail rejection fixes
xvine:fixes
xvine: fixes
Fixing trail setup rejection
Changing the logic to check for circle completion
Proper log messages
Code refactoring Handling the case where friend which is a finger may
have trail len > 0
Free successor_hash_map
Fixing Warning, Creating new delays for puts and gets
Changed find_finger and verify_successor timeouts
Commented out Exponential Backoff
Handling missing next hop in verify successor
debug prints
Adding a new message type,
GNUNET_MESSAGE_TYPE_XDHT_P2P_NOTIFY_SUCCESSOR_CONFIRMATION
Waiting for verify successor result and timing out if result not received
in some time
Check if task is dead
Fixme
Modified logic for notify successor
Change in put/get logic
Code refactoring
Fixing bug causing cancelling of unscheduled task in send_notify
Fix
Fix
Fix
Code refactoring
Randomized PUT/GET delay interval.
Changing delay_put and delay_get
- Adding r5n profiler. - Changes in makefile for r5n profiler - Correctly
adding the trail in add_trail - Passing correct trail id in PUT message routing
lookup
Checking for GNUNET_CONSTANTS_MAX_ENCRYPTED_MESSAGE_SIZE in put message
in X-Vine
If 0 max search, use r5n dht
Exponential backoff
Reset the successor send time in notify confirmation message
Exponential backoff for find finger trail task
Readding trail compression to check if it can reduce trail length
Removing r5n profiler, using the same profiler for both x-vine and r5n
Removing trail compression message
- Act malicious API complete - Using multiple trails in PUT/GET
X-Vine: Fixes
xvine: remove warnings
xvine: better log prints
hark (1):
experimental gstreamer audiobackend (not working)
xrs (3):
get rid of segfault in gnunet-social
test_multicast.c: Added some comments and a check
Add new testcase for multipeer scenario.
-----------------------------------------------------------------------
hooks/post-receive
--
Main GNUnet code base.
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [GNUnet-SVN] [gnunet] Main GNUnet code base. annotated tag initial-import-from-subversion-38251 created. initial-import-from-subversion-38251,
gitolite <=