Re: Trunk gcj ICEs on instance variable increment within try block when compiling to native object. (And a try/finally related bug.)

[Alexandre Petit-Bianco <apbianco at cygnus dot com>]

Mark J Roberts writes:

> $ gcj -c Test.java
> Test.java: In class `Test':
> Test.java: In method `Test.foo()':
> Test.java:5: Internal compiler error in fixup_abnormal_edges, at reload1.c:9516

This appeared one or two weeks ago (I posted something about it to
this list I think,) then went away and re-appeared a couple of days
ago. I filed other/3970 yesterday with a more complicated example, I'm
to adding yours to the case.

> $ gcj -o test --main=Test Test.java
> $ ./test
> Aborted (core dumped)

I'm going to have a look. Chances are that gcc's EH is having problems
and it's not gcj's fault.

> $ jikes Test.java
> $ gcj -o test --main=Test Test.class
> $ ./test

I would emit the same diagnostic for this one, but I'll have a look
too.

> Finally, running a jikes-built class file with gij:
> 
> $ gij Test.class
> Aborted (core dumped)

I have no idea where this can be coming from. Could you send us the
class file?

Thank you for reporting these problems.

./A