[Bug-gnu-arch] [bug #5414] --skip-present may fail in some situations

[nobody]

=================== BUG #5414: LATEST MODIFICATIONS ==================
http://savannah.gnu.org/bugs/?func=detailbug&bug_id=5414&group_id=4899

Changes by: Robert Collins <address@hidden>
Date: Tue 11/25/2003 at 21:32 (GMT)

            What     | Removed                   | Added
---------------------------------------------------------------------------
          Resolution | None                      | Fixed
              Status | Open                      | Closed
             Release | tla--devo--1.1--patch-170

 | tla--devo--1.1--patch-170


------------------ Additional Follow-up Comments ----------------------------
merged into address@hidden/tla--devo--1.1--patch-204



=================== BUG #5414: FULL BUG SNAPSHOT ===================


Submitted by: gnubert                 Project: GNU arch -- a revision control 
system
Submitted on: Thu 09/18/2003 at 23:46
Category:  tla                        Severity:  5 - Major                  
Bug Group:  bug                       Resolution:  Fixed                    
Status:  Closed                       Release:  tla--devo--1.1--patch-170   
Fixed Release:                        Merge Request?:  yes -- diffs included
Your Archive Name:                    Your Archive Location:                
Assigned to:  None                    

Summary:  --skip-present may fail in some situations

Original Submission:  

Hi,



when there is a patch to be skipped, and there also are

missing patches with a lower number of digits in the patch

level, --skip-present will not work as expected.



>         patch-168    (simple changeset)

>           Wed Sep 17 15:47:53 PDT 2003      Tom Lord <address@hidden>

>           --skip-present support (rbcollins)



I just wanted to try it out, and found, that it didn't seem

to work for my special case: patches-6..12 missing, but

patch-11 containing logs already in the local tree. So

patch-11 should be skipped.



In whats-missing.c/filter_present_logs, I could see that

`unfiltered' and `to_skip' had the values:



       unfiltered      to_skip



        patch-6        patch-11

        patch-7

        patch-8

        patch-9

        patch-10

        patch-11

        patch-12



Rel_join then joins the lists, including only items from

`unfiltered' that are smaller than corresponding items in

`to_skip'. But rel_join is using str_cmp, so that "patch-6" is

already greater than "patch-11", so that `to_skip' gets

incremented, and the rest of `unfiltered' just printed.



If `unfiltered' and `to_skip' were sorted before joining,

the lists would look



       unfiltered      to_skip



        patch-10      patch-11

        patch-11

        patch-12

        patch-6

        patch-7

        patch-8

        patch-9



and the result of rel_join would be as expected:



------------------------------------------------------------>8----------

--- orig/libarch/whats-missing.c

+++ mod/libarch/whats-missing.c

@@ -123,7 +123,14 @@

        lim_free (0, revision);

       }

     rel_free_table (has);

+

+    rel_sort_table_by_field (0, to_skip, 0);

+    rel_sort_table_by_field (0, unfiltered, 0);

+

     answer = rel_join (1, rel_join_output (1,0, -1), 0, 0, unfiltered, 
to_skip);

+

+    arch_sort_table_by_patch_level_field (0, answer, 0);

+

     rel_free_table (unfiltered);

     rel_free_table (to_skip);

     return answer;

----------8<------------------------------------------------------------





-- 

Michael

Follow-up Comments
*******************

-------------------------------------------------------
Date: Tue 11/25/2003 at 21:32       By: robertc
merged into address@hidden/tla--devo--1.1--patch-204

-------------------------------------------------------
Date: Sat 11/08/2003 at 11:41       By: robertc
address@hidden/tla--integration--1.1--patch-4 has this patch in it. Thanks for 
diagnosing this, and I'll heckle Tom to get it included.

-------------------------------------------------------
Date: Thu 10/02/2003 at 00:07       By: gnubert
I updated to tla--devo--1.1--patch-188, but the problem still exists:



Patch 182 introduces sorting of "has" and "needs" lists, but this has an

effect only on changesets from the other branch (that this branch

already `has' or still `needs'). Changesets from _this_ branch that are

already present in the other branch and should be skipped, are not

handled at this step, but later, in filter_present_logs() -- that's

where my diff would apply.

-------------------------------------------------------
Date: Tue 09/30/2003 at 22:44       By: robertc
Tom thinks he fixed this already, as he hit it in a different use-case to 
skip-present... can you please test and confirm by closing the bug.


CC List
*******

CC Address                          | Comment
------------------------------------+-----------------------------
address@hidden             | 



No files currently attached


For detailed info, follow this link:
http://savannah.gnu.org/bugs/?func=detailbug&bug_id=5414&group_id=4899

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/