I wrote:
> In exception.c I changed:
> if (pc < start_pc || pc > end_pc) {
> continue;
> }
> to:
> if (pc < start_pc || pc >= end_pc) {
> continue;
> }
I've checked this change in. I still think a pass needs to be made
through the pc manipulation code (as Tim Stack pointed out in
subsequent mails) to remove various bogus -1s. However, the patch
fixes a problem that is annoying the nightly regression tests.
I also added a new test case 'CatchLimits.class.save'. This is
compiled from 'CatchLimits.j' which is a Jasmin bytecode assembler
file. It tests throws at the limits of exception ranges (just before
and just after the start/end). The test passes in all builds of Kaffe
now (and it failed in the interpreter builds), so I've got a bit more
confidence in this fix.
Hopefully the nightly regression tests will start being more useful
(and more prompt) now.
-Pat
----- ----- ---- --- --- -- - - - - -
Pat Tullmann [EMAIL PROTECTED]
"Forty-Two." -- Deep Thought
_______________________________________________
kaffe mailing list
[EMAIL PROTECTED]
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe
