[PATCH] Temporary workaround for openmp.c miscompilation
Steve Kargl
sgk@troutmask.apl.washington.edu
Thu Feb 16 18:03:00 GMT 2006
On Thu, Feb 16, 2006 at 09:35:53AM -0500, Jakub Jelinek wrote:
> Here is a workaround (the compiler is probably not yet smart enough to
> figure out code->op will be never EXEC_NOP and by adding case 0:
> that is not optimized out the problem goes away).
> Do you think I should apply this temporarily and revert when
> PR middle-end/26316 is fixed?
>
> 2006-02-16 Jakub Jelinek <jakub@redhat.com>
>
> * openmp.c (resolve_omp_clauses): Add a dummy case label to workaround
> PR middle-end/26316.
>
> --- gcc/fortran/openmp.c.jj 2006-02-16 08:22:59.000000000 +0100
> +++ gcc/fortran/openmp.c 2006-02-16 15:30:25.000000000 +0100
> @@ -865,6 +865,10 @@ resolve_omp_clauses (gfc_code *code)
> : list == OMP_LIST_MULT ? "IOR" : "IEOR",
> n->sym->name, &code->loc);
> break;
> + /* Workaround for PR middle-end/26316, nothing really needs
> + to be done here for OMP_LIST_PRIVATE. */
> + case OMP_LIST_PRIVATE:
> + gcc_assert (code->op != EXEC_NOP);
> default:
> break;
> }
>
This fixes the failures on amd64-*-freebsd. I think it is
OK to commit; otherwise, we'll be closing numerous duplicate
bug reports until PR 26316 is fixed properly.
--
Steve
More information about the Gcc-patches
mailing list