[OpenSIPS-Devel] SF.net SVN: opensips:[5847] trunk/modules/nat_traversal/nat_traversal.c

[Thomas Gelf]

Dan Pascu wrote:
> I'm a bit confused. You say segfaults started with 5783, but you have  
> a backtrace from 5780?

I showed Bogdan my (few, seldom) segfaults that I noticed with 5780, and
that resulted in 5783 (causing MANY more segfaults) and finally 5784
(avoiding this special segfault at the cost of making keepalives no
longer working correctly).

> Can you confirm if using system malloc made the problem disappear?

No, I can't. I'm pretty sure that with 5783 I'd immediately achieve
once again such a segfault - but that's as far as I understood to
be expected and not related to the earlier problem (even if segfaults
are looking very similar).

Yesterday I tried latest SVN with both system- and pgk_malloc. And I
was able to crash both of them. But that was for other reasons (see
thread "Some ways to crash OpenSIPS with current SVN", NOT related
to nat_helper.

Now I'll better wait for a response to my thread "Unable to disable
PKG_MALLOC", as I don't feel quite well with an 1.4.4 patch for current
trunk ;-) Yesterday I removed -DPKG_MALLOC, reverted some statistic-
related changes (r5653-5655) and finally also removed -STATISTICS.

That way I was able to compile it (not sure if done correct, but
looked good to me). Please note once again, that I never ever wrote
C code - it's just thanks to OpenSIPS that I had to dig a little bit
into it and teach myself some basics. Therefore I'm mot sure whether
my changes where fine ;-)

As I absolutely need to keep this system up and running at least
during business hours (seldom crashes and restarts by monit are
something I can live with - just r5783 was a little bit too heavvy).
If you have an idea how to easily reproduce such a crash that would
be great - but if really related to memory corruption I doubt that
this would probably be not an easy task :-/

Cheers,
Thomas