Re: Patches Applied for 21.1.8 [includes essential pathc]
Gunnar Evermann <ge204(a)eng.cam.ac.uk> writes:
> You could of course consider it rude of us that we don't set
flag in
> configure (I would prefer a "Sorry compiler too broken! Abort").
indeed. why don't we do that then?
Incidentally is there any reason why we do not try to print some kind
of C stacktrace from our signal handler?
Usual reason: Nobody wrote the patch. Boring but true.
The Sun Workshop version did that and sometimes it was really
helpful.
That means the code is already somewhere for solaris...?
somewhat platform specific but I think at least for Solaris and
linux/glibc it should be easy (didn't Ulrich post something about that
some time ago?).
Yes. And W.Perry tried it on their in-house software. Maybe they can
be asked for help. (Note that this is for glibc 2.1 and up).
At least we don't suffer from the "GDB development process too slow
for Linux" aka "Linux changing its signal handler trampoline too
often" aka "short backtrace" problem.
Jan