Re: Reproducible, corrupt packfile after fresh git-svn checkout message 3 of 20)

From: Thomas Rast <trast@inf.ethz.ch>
Date: Aug 8, 2013, 13:20 UTC
Message-ID: <87y58c1fpo.fsf@linux-k42r.v.cablecom.net>
In-Reply-To: <9aa1672d5c04994f416dccd84b5983c960c0fdf9@localhost>

gitml.jexpert@recursor.net writes:

> So - now the puzzling thing: With valgrind it seems to work! 
> If I run it plain, it doesn't:
>
> /tmp/project.git $ valgrind --track-origins=yes  ~/projects/git.git/git-fsck

[...]

Show 10 quoted lines

> ==3431== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 2 from 2)
> /tmp/project.git $ ~/projects/git.git/git-fsck
> Checking object directories: 100% (256/256), done.
> error: packed 49cdd0b21a351f3366008615d2cf8d03ca943978 from
> .git/objects/pack/pack-6a6f5355584a5d71215d5fc867ce09602ceab533.pack
> is corrupt
> *** glibc detected *** <unknown>: free(): invalid pointer: 0x00007f8576682010 ***
> ======= Backtrace: =========
> /lib/x86_64-linux-gnu/libc.so.6(+0x7eb96)[0x7f85ad765b96]
> [0x4e727c]

[...]

> Any further hints?

Hrm.  Can you try getting a backtrace from

  $ gdb ~/projects/git.git/git-fsck
  (gdb) run
  ... wait until it crashes ...
  (gdb) backtrace full

I would have been more interested in error output from valgrind, because memory corruption invariably happens long before glibc finally figures out that something is amiss. But as things stand, the backtrace is probably the only thing we have...

-- 
Thomas Rast
trast@{inf,student}.ethz.ch