Re: SIGIO, xemacs random crashes on SMP system

[Anthony Mallet <anthony.mallet%laas.fr@localhost>]

On Tuesday, at 20:45, Matthias Scheler wrote:
| Is that only true for "xemacs" or can you come up with a simple
| programs that demonstrates this problem?

At the moment I couldn't reproduce any problem in a simple program (I'm still
trying though), and "xemacs" is the only application that doesn't run as
expected on my system. Of course this should be a big evidence of a problem in
xemacs itself and not in the system, but a few facts make me still doubt:

a/ it never ever happened until around January, and since then it's been pretty
   prominent and painful.
b/ even if xemacs trashes it's stack in some circumstances, I can't hardly see
   how an empty signal handler being present or absent could change something.
c/ if I stop all the cpus but one, the problem disappears (although, of course,
   xemacs is not even linked with libpthread) ...

I saw some "XXXSMP locking" comments in kern_sig.c, so I was wondering if this
could not be related. But I definitely lack of knowledge on signal handling in
the kernel to be able to understand the meaning of those XXXSMP.

Still looking for any suggestions on how to debug that...

Regards
-- 
Anthony