Re: Regression in git-subtree.sh, introduced in 2.20.1, after 315a84f9aa0e2e629b0680068646b0032518ebed

From: Duy Nguyen <pclouds@gmail.com>
Date: Dec 31, 2018, 11:20 UTC
Message-ID: <CACsJy8AQ1raB+2wjEBtDrAvJm0v-3exJ-FVB2ZyL6VpRFQkspQ@mail.gmail.com>
In-Reply-To: <11AC6C27-4C74-43B1-89F6-98ABBFE74E0E@msys.ch>

On Mon, Dec 31, 2018 at 6:13 PM Marc Balmer <marc@msys.ch> wrote:

Show 24 quoted lines

>
>
>
> > Am 31.12.2018 um 11:51 schrieb Duy Nguyen <pclouds@gmail.com>:
> >
> > On Mon, Dec 31, 2018 at 5:44 PM Marc Balmer <marc@msys.ch> wrote:
> >>
> >> Hi
> >>
> >> One of the last three commits in git-subtree.sh introduced a regression leading to a segfault.
> >>
> >> Here is the error message when I try to split out my i18n files:
> >>
> >> $ git subtree split --prefix=i18n
> >> cache for e39a2a0c6431773a5d831eb3cb7f1cd40d0da623 already exists!
> >>   (Lots of output omitted)
> >> 436/627 (1819) [1455]       <- Stays at 436/ while the numbers in () and [] increase, then segfaults:
> >> /usr/libexec/git-core/git-subtree: line 751: 54693 Done                    eval "$grl"
> >>    54694 Segmentation fault      (core dumped) | while read rev parents; do
> >
> > Do you still have this core dump? Could you run it and see if it's
> > "git" that crashed (and where) or "sh"?
>
> It is /usr/bin/bash that segfaults.  My guess is, that it runs out of memory (as described above, git-subtree enters an infinite loop untils it segafults).

Ah that's better (I was worried about "git" crashing). The problematic commit should be 19ad68d95d (subtree: performance improvement for finding unexpected parent commits - 2018-10-12) then, although I can't see why.

I don't think we have any release coming up soon, so maybe Roger can still have some time to fix it instead of a just a revert.

Show 24 quoted lines

>
> >
> >>   process_split_commit "$rev" "$parents" 0;
> >> done
> >>
> >> Please note that this regression can not easily be reproduced, normally a subtree split just works.
> >>
> >> Reverting the last three commits "fixes" the issue.  So I kindly ask the last three commits to be reverted.
> >
> > Please provide the SHA-1 of the "good" commit you tested.
>
> I reverted these three commits (actually the last three commits to contrib/subtree/git-subtree.sh):
>
> 19ad68d95d6f8104eca1e86f8d1dfae50c7fb268
> 68f8ff81513fb3599ef3dfc3dd11da36d868e91b
> 315a84f9aa0e2e629b0680068646b0032518ebed
>
> And then it worked.
>
> - Marc
>
> --
> > Duy
>

-- 
Duy