Re: Reproducible, corrupt packfile after fresh git-svn checkout message

From: Ben Tebulin <tebulin@googlemail.com>
Date: Aug 9, 2013, 12:27 UTC
Message-ID: <5204E018.4000808@gmail.com>
In-Reply-To: <7vsiyk9j6a.fsf@alter.siamese.dyndns.org>

Hello everybody!

I have some _very interesting_ news regarding this issue! Here is the deal:

  1. I was able to *reproduce the error on a machine of a coworker!*

  2. I was able to rule out
      - HDD: It's reproducible from /dev/shm
      - Memory: Memory tests works fine

now the interesting part:

  3. Occurs on Linux kernels 3.7.10, 3.8.x, 3.9.11, 3.10.5
     _but not on:_ 3.6.11, 3.5.7 and 2.6.32

Both machines showing this problem are »Dell Latitude E6330« with an »i5-3340M @ 2.70GHz« CPU. Mine running stock kernels, coworker using plain Ubuntu 13.04.

Furthermore I need to stress, that we never had any issues with our devices during daily _at all_.

So what to do best now?

- Ben

Am 08.08.2013 19:38, schrieb Junio C Hamano:

Show 6 quoted lines

>>>> Can you try to reproduce with a version older than v1.8.3?
>>>> E.g. v1.8.2.3.
>>>
>>> It seems to run 'much better' 
>>>   v1.8.2.3 : 3/10 runs do fail
>>>   fb56570  : 9/10 runs do fail

[..]

Show 22 quoted lines

>>> They always fail on a big blob (39MB) as I wrote in my first e-mail:
>>>
>>> ben@n179 /tmp/project.git $ ~/projects/git.git/git-show 49cdd0b21a351f3366008615d2cf8d03ca943978 | wc -c
>>> error: sha1 mismatch 49cdd0b21a351f3366008615d2cf8d03ca943978
>>> fatal: bad object 49cdd0b21a351f3366008615d2cf8d03ca943978
>>> 0
>>> ben@n179 /tmp/project.git $ ~/projects/git.git/git-show 49cdd0b21a351f3366008615d2cf8d03ca943978 | wc -c
>>> 39517156
> 
> Hmm, from this, and a later one ...
> 
> ... I am reading that (1) the packfile and repository is basically
> OK, (2) reading that object sometimes fails, and (3) the symptom is
> not limited to fsck but anything that reads the object with
> parse_object().  And that symptom exists only on that single machine
> (I am assuming that the repository was bit-for-bit copied, not
> "cloned", for the purpose of testing it on the other machine).  That
> makes me suspect something outside the control of Git (e.g. faulty
> memory or disk controller cable).
> 
> Are there other big blobs in the repository, and would "show | wc" fail
> if you attempt to read it on that machine?