RE : [lwip-devel] 1.3.0

[Frédéric BERNON]

I really would like to got a 1.3.0 release (my feeling is that lwIP release 
time was too long between 1.2.0 and "1.3.0", of course, there was lot of 
changes, but in a general way, I would be in favor of shorter release time). 
But before release it, some records should be closed. Here the full list (hope 
it will be readable), with MY point of view (it's not like this in the 
bugtracker).

TASKS
 #7421   Implement SO_RCVBUF                                                    
       1.3.0
 #7410   Don't include internal headers in api header files                     
       1.3.0
 #7244   Add sys_arch_ticks/jiffies in sys_arch                                 
       post-1.3.0
 #7235   Create timeout framework for NO_SYS=1                                  
       1.3.0
 #7212   Add Mutex concept in sys_arch                                          
       post-1.3.0
 #7040   Work on tcp_enqueue                                                    
       post-1.3.0
 #7017   Implement DNS client                                                   
       1.3.0
 #7013   Create option to have all packets delivered to netif->output in one 
piece     post-1.3.0
 #6995   Implement SO_REUSEADDR                                                 
       post-1.3.0
 #6969   Review usage of conn->err in netconn layer                             
       1.3.0
 #6930   Implement SO_LINGER                                                    
       post-1.3.0
 #6881   Create option to leave header fields in host byte order                
       post-1.3.0
 #6849   Test how checksum on copy could be integrated into the stack           
       post-1.3.0
 #6735   Provide new pbuf type: PBUF_RAM_NOCOPY                                 
       post-1.3.0
 #6683   Document lwIPs thread safety requirements                              
       could be closed (doc is an infinite task)
 #1551   Make ARP more independent of interface hardware type                   
       post-1.3.0
 #1549   lwIP function documentation                                            
       could be closed (doc is an infinite task)

BUGS
 #7040   Work on tcp_enqueue
 #21535  Retransmit Timer in SYN_SENT state                                     
       post-1.3.0
 #21494  TCP pcb->mss should be 536 if no MSS option received                   
       1.3.0
 #21492  TCP MSS can be larger than MTU                                         
       post-1.3.0
 #21491  TCP MSS sent is based on the received MSS option                       
       1.3.0
 #21433  Calling mem_free/pbuf_free from interrupt context isn't safe           
       post-1.3.0
 #21099  Request getting lost/unanswered in lwip stack [bug #20640]             
       post-1.3.0 (is it invalid or not?)
 #21049  Allow reuse of pbufs after return from APIs                            
       post-1.3.0
 #20900  Potential crash error problem with netconn_peer & netconn_addr         
       1.3.0
 #20779  Keep-Alive and SYNs                                                    
       post-1.3.0
 #20515  TCP delayed ack does not work as expected                              
       post-1.3.0
 #20511  No persist timer                                                       
       post-1.3.0
 #20287  tcp_output_nagle sends too early                                       
       post-1.3.0
 #20199  TCP appears to be "Shrinking the Window"                               
       post-1.3.0
 #20155  select() with errorfd set                                              
       post-1.3.0
 #19927  DHCP NACK problem                                                      
       post-1.3.0
 #19699  Bug in SNMP ASN1 decode; submitted patch                               
       post-1.3.0
 #13315  PPP PAP authentication can result in erroneous callbacks               
       post-1.3.0
 #3031   Implement a new fully pool-based pbuf implementation.                  
       Isn't it done???

PATCHS
 #6253   Added csum to struct pbuf.                                             
       post-1.3.0
 #5628   #17726 PPP: memory wasting with every reconnection by pppClose/SIG     
       post-1.3.0




Most of tasks/bugs I think we should integrate in 1.3.0 are already in 
progress. Once again, it's my point of view.

In a flat view, I think we can group all these records like this:

"TCP"
 #21535  Retransmit Timer in SYN_SENT state
 #21494  TCP pcb->mss should be 536 if no MSS option received
 #21492  TCP MSS can be larger than MTU
 #21491  TCP MSS sent is based on the received MSS option
 #21494  TCP pcb->mss should be 536 if no MSS option received
 #21099  Request getting lost/unanswered in lwip stack [bug #20640]
 #20779  Keep-Alive and SYNs
 #20515  TCP delayed ack does not work as expected
 #20511  No persist timer
 #20287  tcp_output_nagle sends too early
 #20199  TCP appears to be "Shrinking the Window"

"pbufs&driver"
 #21433  Calling mem_free/pbuf_free from interrupt context isn't safe
 #21049  Allow reuse of pbufs after return from APIs
 #6735   Provide new pbuf type: PBUF_RAM_NOCOPY
 #7013   Create option to have all packets delivered to netif->output in one 
piece 
 #6849   Test how checksum on copy could be integrated into the stack
 #6253   Added csum to struct pbuf
 #3031   Implement a new fully pool-based pbuf implementation
 
"sys_arch"
 #7244   Add sys_arch_ticks/jiffies in sys_arch
 #7212   Add Mutex concept in sys_arch

"API"
 #7421   Implement SO_RCVBUF
 #7235   Create timeout framework for NO_SYS=1
 #7017   Implement DNS client
 #6995   Implement SO_REUSEADDR
 #6969   Review usage of conn->err in netconn layer
 #6930   Implement SO_LINGER
 #20900  Potential crash error problem with netconn_peer & netconn_addr
 #20155  select() with errorfd set

"Doc"
 #6683   Document lwIPs thread safety requirements
 #1549   lwIP function documentation

"Task,code in long long stand by"
 #6881   Create option to leave header fields in host byte order
 #1551   Make ARP more independent of interface hardware type
 #19927  DHCP NACK problem
 #19699  Bug in SNMP ASN1 decode; submitted patch
 #13315  PPP PAP authentication can result in erroneous callbacks
 #5628   #17726 PPP: memory wasting with every reconnection by pppClose/SIG

Last, for the release number, I'm in favor of 'lwIP 1.3.0'

  
====================================
Frédéric BERNON 
HYMATOM SA 
Chef de projet informatique 
Microsoft Certified Professional 
Tél. : +33 (0)4-67-87-61-10 
Fax. : +33 (0)4-67-70-85-44 
Email : address@hidden 
Web Site : http://www.hymatom.fr 
====================================
P Avant d'imprimer, penser à l'environnement
 


-----Message d'origine-----
De : address@hidden [mailto:address@hidden De la part de David Empson
Envoyé : jeudi 8 novembre 2007 04:35
À : lwip-devel
Objet : Re: [lwip-devel] 1.3.0


Simon Goldschmidt <address@hidden> wrote:
> when did you plan to move on to 1.3.0? There are a couple of bugs  
> open
> that are planned for 1.3.0, most of them are issues in the TCP  part.
>
> As I don't see me having plenty of time to fix TCP bugs for the next
> months (since I have a 'real job', hehe - unfortunately, I'd love to  work 
> on those), I'd love to see 1.3.0 without the TCP fixes just to  get a new 
> release out with the numerous fixes 1.3.0 would include  compared to 
> 1.2.0!

If the TCP bugs are long standing, not easy to fix/test quickly, and fixing 
them will not involve changing any of the APIs, then I agree: release 1.3.0 
and defer fixing those bugs until (say) 1.3.1.

The primary goal should be to get everything done in one hit which will 
require everyone to port LWIP back into their own implementations. API and 
major structure changes are more difficult to deal with than LWIP internal 
bug fixes.

If the TCP bugs will require API changes, then they should be fixed first, 
or at least the required API changes should be done without fixing the bugs 
(and without introducing new ones!), so 1.3.0 can be released.

Jared Grubb <address@hidden> wote:
> We could issue a "1.2.5" release based on the current head... Just an 
> idea if there are still things we really want to finish for a 1.3.0 
> release.

The degree of changes in the API suggest to me that this should be called 
1.3.0 even if all the planned changes don't made it in. 1.2.5 implies that a 
portion of the changes intended for 1.3.0 have been applied to the 1.2.0 
code base and the differences have been kept to as few as possible. This is 
not the case, in my opinion. It would be better to call this one 1.3.0 and 
then do a 1.4.0 if there are further API changes. 



_______________________________________________
lwip-devel mailing list
address@hidden http://lists.nongnu.org/mailman/listinfo/lwip-devel

Frédéric BERNON.vcf
Description: Frédéric BERNON.vcf