|
From: Jeremy F. <je...@go...> - 2005-02-21 17:26:05
|
Eyal Lebedinsky wrote:
> Running off cvs head, using:
> valgrind --tool=memcheck \
> --leak-check=yes \
> --show-reachable=no \
> --num-callers=50 \
> --error-limit=no \
> --run-libc-freeres=no \
> --time-stamp=yes \
> --trace-syscalls=no \
> --quiet \
> my program...
>
> And also with '--trace-syscalls=yes'.
>
> After the recent changes in memory handling, I now have a situation
> where my program seems to run to completion, but then after it
> exit()s it aborts with a segmentation fault.
>
> I do not see any error logged by valgrind.
Well, --quiet will suppress the printing about SIGSEGVs, so you wouldn't
see any crash messages. If it crashes after exit(), it could still be
some atexit/static destructor calls which are causing the problem. Does
it generate a vgcore file?
> This is a new problem from the last few days. Beforehand things
> were mostly working, except for my server running out of
> memory (it was not really using that much of it and I could not
> explain what was going on yet).
Well, the changes to memcheck reduced its memory use by a large amount
in a lot of common cases, so it could be that you're just getting further?
> If I should run my program with logging/debug options and post
> the logs then I am ready to do so.
Yes, please: strace, -v, --trace-syscalls=yes and --trace-signals=yes .
And file a bug in bugzilla.
Thanks,
J
|