This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: [gfortran, committed] Fix regressions introduced by patch for PR 25045
On Wed, Feb 15, 2006 at 04:02:45PM -0500, Jakub Jelinek wrote:
> On Wed, Feb 15, 2006 at 12:58:23PM -0800, H. J. Lu wrote:
> > Do you have this
> >
> > http://gcc.gnu.org/ml/gcc-patches/2006-02/msg01249.html
> >
> > in your tree? My results are for revision 111111, which doesn't have
> > your change.
>
> ATM yes, but since I committed the gomp merge to the trunk till I wrote
> msg01249.html patch, I bootstrapped/regtested the trunk without
> msg01249.html patch twice on x86_64-linux and once on ppc64-linux,
> several times on gomp branch and many times on redhat/gcc-4_1-branch
> (which also includes the gomp backport), on 7 different arches.
It looks like the switch statement in resolve_omp_clauses is
miscompiled:
.loc 1 819 0
leal -6(%r12), %r13d
.LVL74:
.loc 1 825 0
mov %r13d, %eax
movq .L133(,%rax,8), %rax
movq %rax, (%rsp)
%r12 is 0, which I believe is list. We have %rax == -6 and
movq .L133(,%rax,8), %rax
segfaulted since .L133(,%rax,8) is too big.
H.J.