[Openswan dev] problem with ASSERTION FAILED

[David McCullough]

Jivin Paul Wouters lays it down ...
> On Thu, 4 Feb 2010, David McCullough wrote:
> 
> >> Something has really broken in openswan-2.6. ASSERT doesn't cause core
> >> dump any more. This makes it quite much harder to debug problems.
> >>
> >> Anybody know how to fix this?
> >
> > Looks to me like ASSERT is not used anywhere in openswan.
> >
> > There is passert() and some others in openswan/passert.h.  This
> > needs DEBUG to be defined.
> >
> > You should be able to call assert().
> >
> > Have a look around the code for assertions and follow their example,
> 
> What Tuomo is saying is that the assert/passert calls seem to cause
> openswan to crash now, instead of producing a core dump. I do think
> there is a problem now with this mechanism. This is also causing
> plutorun issues where plutorun is no longer starting a new pluto
> process when it detects pluto has crashed.

You should be getting at least a log entry/file/line for any abort that
happens.

It does all it's logging etc and then calls abort (depending on all kinds of
obfuscation of what actually gets called).

pluto doesn't appear to be installing a SIGABRT handler and AFAICT a SIGABRT
(kill -6 pid) will cause a core dump (man kill),  unless someone changed
that also :-(

I know on our systems we dump core on an abort/sig 6.  But then we can't
write the core out (RO filesystems ;-)

This is a ulimit thing by any chance ?

Cheers,
Davidm

-- 
David McCullough,      david_mccullough at mcafee.com,  Ph:+61 734352815
McAfee - SnapGear      http://www.mcafee.com         http://www.uCdot.org