Re: PATCH: Fix Solaris 2/x86 signal handling in libgcj

[Andrew Haley <aph at redhat dot com>]

On 02/24/2010 02:28 PM, Rainer Orth wrote:
> While investigating the remaining Solaris 11/x86 testsuite failures, I
> found two causes for the remaining failures, where patches have already
> been posted:
> 
> 	PATCH: Properly mark 64-bit .eh_frame section on x86 as type unwind
> 	http://gcc.gnu.org/ml/gcc-patches/2010-02/msg00991.html
> 
>         PATCH: Fix Solaris 11/x86 MD_FALLBACK_FRAME_STATE_FOR
>         http://gcc.gnu.org/ml/gcc-patches/2010-02/msg00994.html
> 
> Even after those, a problem remained: configure.ac didn't define any
> SIGNAL_HANDLER for the platform, unlike Solaris 2/SPARC, which obviously
> isn't right.  In fact, the code currently in include/sparc-signal.h is
> in no way SPARC specific and works on x86 just as well.  It isn't even
> Solaris specific at all, and should work on any POSIX system.  Many of
> the other *-signal.h files are very similar, and could probably be
> merged into a single posix-signal.h to avoid much unnecessary
> duplication.

Hmmm, does this really work?  I would have expected any x86 target to
need the divide overflow fixups in i386-signal.h.  I guess you're
defining DIVIDESPEC=-fuse-divide-subroutine ?

Andrew.