[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [GNUnet-developers] Reproduceable failure

From: Igor Wronsky
Subject: Re: [GNUnet-developers] Reproduceable failure
Date: Tue, 1 Apr 2003 00:28:14 +0300 (EEST)

On Mon, 31 Mar 2003, Krista Bennett wrote:

> can't speak for him (for one thing, I can't get the German accent down :)
> I think I vaguely recall him not being able to reproduce the mem-hogging
> situation, although I could be mistaken.

True. But I can reliably reproduce it. It suffices to let gnunetd run,
and the more you do anything with it, the more it hogs memory. On the
other hand, I don't have the necessary skill or knowledge to track
the problem. I've tried the script made by Christian, that didn't
help, I tried Valgrind, that claimed that openssl 0.9.7 had leaked 50
megs hopelessly and some anonymous piece 110 megs but in a way
that the pointers were still available. That latter one had covered
its tracks well though,

==19453== 50916288 bytes in 12216 blocks are definitely lost in loss
record 12 of 13
==19453==    at 0x4015CDA8: malloc (vg_clientfuncs.c:100)
==19453==    by 0x4029AB4A: (within /usr/lib/i686/cmov/
==19453== 111646128 bytes in 20463 blocks are still reachable in loss
record 13 of 13
==19453==    at 0x4015CDA8: malloc (vg_clientfuncs.c:100)
==19453==    by 0x4261B22D: ???
==19453==    by 0x42619040: ???
==19453==    by 0x426180AB: ???

"Still reachable" means that someone has pointers left to that
memory, so in theory it could've been freed before exiting, but

Currently I can't get gnunetd even started, so my chances
on finding out this is doubly on ice (not that the current
start problem with the config file were unsurmountable, I'm
just today not in mood to track each bug caused by a hit-and-run
cvs commit without compile/run -test done by the committer).

Perhaps this is a take-gnunet-down-piece-by-piece-and-see
-when-the-memory-hogging-stops... :(

> Sorry you're feeling frustrated, Igor.

Maybe not entirely related to gnunet, but seems recurring. ;)


reply via email to

[Prev in Thread] Current Thread [Next in Thread]