Hi,
Is it possible that apache's
SetOutputFilter DEFLATE directive could be causing this?
This is just an uneducated guess.
cu
christian unger
On Mar 22, 2007, at 11:16 PM, Florent Daignière (NextGen$) wrote:
> * Ph. Marek <philipp.marek@bmlv.gv.at> [2007-03-16 07:54:55]:
>
>> On Friday 16 March 2007 00:40, Florent Daignière wrote:
>> ...
>>> Today I have encountered an other problem with the same repository :
>> ...
>>> I wonder if I'm really (un)lucky or if the hardware is to blame.
>> As it seems to be (something like) reproduceable - may I suggest
>> using strace?
>> strace -f -tt -i -o /tmp/subversion-bug -p <pid of httpd> -p
>> <other pid> ...
>> would write a series of logfiles into /tmp/subversion-bug.PID -
>> you could keep
>> only the last 100 or so, remove the other ones to keep storage
>> usage low.
>>
>> That would tell us whether it's software - and with a copy of
>> /proc/<offending pid>/maps
>> we might see (from the instruction pointer) which library is to
>> blame,
>> although that's more a hope than anything else.
>>
>> Regards,
>>
>> Phil
>
> Today the repository got corrupted twice (no software update has been
> done in the meantime). I realized it too late to get the first strace
> log ... but I got the second one. Unfortunately I was unable to get
> the memory map from /proc (I'm using -grsec on my kernel and it hasn't
> helped).
>
> Each time it has been a "Found malformed header in revision file"
> error
> (like the first two ones I reported).
>
> Here are the two revision files:
> http://emu.freenetproject.org/~nextgens/12276
> http://emu.freenetproject.org/~nextgens/12281
>
> http://emu.freenetproject.org/~nextgens/subversion-httpd-strace.gz
> - be carefull : 20M - I don't have the exact commit timestamp so I
> kept
> data around... the post-commit hook has been triggered at 20:51 I
> think.
>
> I am reluctant to publish apache's logs here but may do it on demand.
>
> Thanks for your time,
> NextGen$
Received on Thu Mar 22 23:34:04 2007