On Mon, Jan 27, 2003 at 06:52:03PM +0100, Tom Ivar Helbekkmo wrote:
> Jeff Rizzo <riz@boogers.sf.ca.us> writes:
> 
> > 0x48081d93 in pthread__switch_away () from /usr/lib/libpthread.so.0
> 
> For what it's worth, I'm seeing core dumps with both pthreads-using
> applications I use (xmms and Mozilla) in this spot, on a uniprocessor
> system.  (Well, it's actually a dual-CPU box, but the kernel only
> supports one of them.)  This system was running well with a complete
> (kernel, userland and packages) environment CVS-ed Friday morning EU
> time, but fails after updating again today, Monday morning.
> 
> Installing the new libpthread.so introduced the behaviour, and
> completely rebuilding everything that used it (including all
> supporting libraries &c) did not help.


Ah HAH.  It turns out, there was a little more difference between
the two systems than I thought.  Installing libpthread.so.0.1 from the
working system into the non-working one caused it to begin
working (even with MP).  I haven't tested it yet, but it looks
like the lossage may be that no threads may be created, as it seems
to be dying at the first pthread_create().  I need to investigate more,
but I would be that whatever changed between Jan 21 and Jan 26 is
to blame here...

+j
-- 
Jeff Rizzo                                         http://boogers.sf.ca.us/~riz