Avis, Ed wrote:
Stephen J. Turnbull wrote:
> Ed> I'm having trouble persuading xemacs to make a core dump on my
> Ed> Linux system, despite 'ulimit -c unlimited' and getting core
> Ed> dumps from ordinary C programs that crash.
>
>How do you start xemacs?
>
>
>From a shell script looking roughly like
#!/bin/bash
no_exit_on_failed_exec=1
exec xemacs -geometry 120x78+0+0
> Ed> To diagnose this, is there a way I can force xemacs to crash?
>
>Untested, but kill -SEGV $xemacs_pid should do it.
>
>
Yes, and that does leave a coredump. It's strange that I don't get one
for the assertion failure. I'll investigate some more.
> Ed> M-x crash-and-dump-core?
>
>That's an interesting concept for a debug build, but we certainly
>wouldn't implement that function in a production build. ;-)
>
>
I think a better function for me would be M-x assertion-failure, to
specifically test what happens when assert() in the C code returns
false.
well, in 21.5 there's already `force-debugging-signal'; the intent is to
forcibly get back to the debugger by one way or another. Under Windows
this does DebugBreak() i.e. INT 3; elsewhere it just does ABORT(), which
is equivalent to assert(0).
maybe we should also add something that always ABORT()'s or assert(0)'s
or something to that extent.
ben