Java 3.3 regression details from sparc-sun-solaris2.7

Mon Jan 27 19:28:00 GMT 2003

On 2003年1月27日, Andrew Haley wrote:
> > > FAIL: TLtest execution - gij test
> >
> > Can you post a stack trace from gdb at the point of the SIGSEGV? (Make
> > sure you continue past the first SIGSEGV in GC_find_limit-- that one is
> > normal.)
>> SPARC runs with -fcheck-references, so any SEGV (except that first
> one) must be a code generation bug, or perhaps something wrong with
> the runtime. Linker problem, perhaps?

My guess is a problem in the runtime, since I don't see the same failures
on solaris 2.8:
http://gcc.gnu.org/ml/gcc-testresults/2003-01/msg01199.html
If Kaveh or someone can find out if the SEGV occurs at all, that'll be
another important clue.
(Come to think of it... why do we install a SIGSEGV handler at all on
-fcheck-references targets? It only seems to complicate debugging in
these situations.)
Jeff