M68k exceptions code.

[Mike Stump]

> Date: Wed, 22 Dec 1999 12:11:23 -0500
> From: "Alexander V. Komarov" <sasha@mera.ru>

> Did anyone succeed with compiling exceptions on m68k?

Yes, generally speaking.

> if not, could You tell me which exception handling way is less
> machine specific - PC relative (__throw) or setjump-longjump ?

setjmp/longjmp is more portable, should be easier to get working, and
requires less knowledge about the CPU...

> Both does not work in my case, but __throw is killed when it takes
> the pointer of a label (btw, how it works?), and i did not dig
> sj-lj...

g++.mike/eh6.C is the testcase to debug, if you want to track it down.

To use sjlj exception handling you will want to change the default in
the source code to default to it (you can't just give the flag on the
command line and have it work 100%) (or desupport dwarf2 EH support).

Also, you'll want to run the whole g++ testsuite and check it against
something same (x86 linux?).