[darcs-users] [patch262] Resolve issue1014: mark the test as pass... (and 4 more)

Eric Kow kowey at darcs.net
Sun Jun 6 19:44:27 UTC 2010


On Fri, Jun 04, 2010 at 15:22:17 +0200, Petr Rockai wrote:
> Well, if this still induced commutation failures, they would still crash
> darcs. So no, we are not silently ignoring the bug. I don't quite
> understand how this bug actually came into existence, since I don't
> think the problem is in any way inherent. It indeed looks like an
> implementation barf in get_extra. Which no longer exists...

So we've discussed this a bit on IRC with Ian this morning,

http://irclog.perlgeek.de/darcs/2010-06-06#i_2408192

I'm not sure I followed all of the discussion, but I think we've come around to
thinking that issue1014 needs a closer look (and more aggressive testing).  This
may be one of those succeeding-for-the-wrong-reason bugs which are way way worse
than crashing.  If I understand correctly, now we have a situation where

1. named patches A and B contain only identical primitive patches.
2. named patch C depends on A (or B)
3. merging AC+BC now leads to a weird repo where we have the named
   patch C appearing *twice* (not a set of patches anymore, but a
   bag :-/)

Please correct me if I'm mistaken.  Perhaps we could extend the issue1014
one by grepping the changes --xml output for the patchinfo and passing it
to wc -l.

> If it was however an inherent bug, the partitionFL would break, since
> the partitioning point would be ill-defined: there would be patches that
> need to go to the left while at the same time they depend on patches
> that need to go to the right. This however does not happen. It would
> still happen in cases 1 and 2 that you describe below.

For the interested, Petr later on realised that actually,

| It won't fail as get_extra used to in the remaining cases, it will err on the
| "common" side -- i.e. some patches that are actually common may fail to be
| recognized as such if there's a buggy patchinfo somewhere.

(Sounds like we should have an example of this cooked up, maybe not as a test
 case but just so we have something to point to as an example of what could go
 wrong).

Anyway, as for the patches in this bundle...

| Thu Jun  3 14:29:57 CEST 2010  Petr Rockai <me at mornfall.net>                                                                                               
|   * Resolve issue1014: mark the test as passing. Likely addressed by NewSet.                                                                               

Needs looking into some more.
                                                                                                                                                            
| Thu Jun  3 14:31:33 CEST 2010  Petr Rockai <me at mornfall.net>                                                                                               
|   * Resolve issue1337: mark the test as passing. Likely addressed by noslurps.                                                                             

I'm happy to see this is solved (applied!)
                                                                                                                                                            
| Thu Jun  3 14:34:00 CEST 2010  Petr Rockai <me at mornfall.net>                                                                                               
|   * Resolve issue1401: mark the test as passing. Likely fixed by NewSet.                                                                                   

Ian said this was the same as 1014, so it seems we should working to
extend this test too (or maybe getting rid of it if it really tells
us nothing new).

| Thu Jun  3 14:37:18 CEST 2010  Petr Rockai <me at mornfall.net>                                                                                               
|   * Resolve issue1610: mark the test as passing. Likely fixed by NewSet.                                                                                   

Hmm, I guess I should just assume this really is fixed.  Yay!

| Thu Jun  3 14:37:35 CEST 2010  Petr Rockai <me at mornfall.net>                                                                                               
|   * Skip the case folding test on case-sensitive systems.   

Already applied.

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
URL: <http://lists.osuosl.org/pipermail/darcs-users/attachments/20100606/011691bd/attachment-0001.pgp>


More information about the darcs-users mailing list