---- Reported by robertj@gmx.net 2005-06-27 05:34:46 MST ----

Some older SUSE distros (7.x, 8.x SLES 7 and 8) have been
shipped with a buggy LIBC which consumes SIGRTMIN itself
leading to crash when libmono.so is loaded, for example
when --with-static_mono=no is applied or when libmono.so is
consumed by third party programs.

About the LIBC bug:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=53394

The patch presets mono_thread_get_abort_signal to return
SIGRTMIN+3. This signal is supposed to be available on
all systems that support SIGRTMIN, because POSIX requires
at least 8 signals.



---- Additional Comments From robertj@gmx.net 2005-06-27 05:36:09 MST ----

Created an attachment (id=168184)
sigrtmin.diff




---- Additional Comments From miguel@ximian.com 2005-06-29 19:45:45 MST ----

I would rather have some way of testing for this than mkaing the
change on every distribution.  

This patch only turns the problem into a potentially different problem.

What we probably need to have is a way of "probing" if these signals
are being used, something like this:

handler = -1;
for (int i = SIGRTMIN; i < SIGRTMAX; i++){
    sigaction (i, NULL, &old);
    if (old.sa_handler == NOT_SET){
       handler = i; break; 
    }
}

Now, the above only works if signals have been preset before Mono has
been invoked.  Am not sure if there is a reliable way of coping with
this problem 



---- Additional Comments From lupus@ximian.com 2005-07-07 14:17:04 MST ----

Fixed in svn: we avoid SIGRTMIN and select a free signal if available.

Imported an attachment (id=168184)