https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127492
--- Comment #5 from GCC Commits <cvs-commit at gcc dot gnu.org> --- The master branch has been updated by Jakub Jelinek <[email protected]>: https://gcc.gnu.org/g:13fd34134abc930f473b558b97fdc6ec99d07a16 commit r17-4781-g13fd34134abc930f473b558b97fdc6ec99d07a16 Author: Jakub Jelinek <[email protected]> Date: Wed Sep 30 12:38:35 2026 +0200 testsuite: Fix up PR127492 test As I wrote earlier, I couldn't make On Tue, Sep 29, 2026 at 09:08:28AM +0200, Jakub Jelinek wrote: > --- a/gcc/testsuite/gcc.dg/pr127492.c 2026-09-22 14:58:11.232624179 +0200 > +++ b/gcc/testsuite/gcc.dg/pr127492.c 2026-09-29 08:51:56.240821492 +0200 > @@ -1,5 +1,5 @@ > /* PR middle-end/127492 */ > -/* { dg-do run } */ > +/* { dg-do run { target { { { int128 && sync_int_128_runtime } || { { ! int128 } && sync_long_long_runtime } } || libatomic_available } } } */ > /* { dg-options "-O2" } */ > /* { dg-additional-options "-latomic" { target libatomic_available } } */ to work, dunno if multiple occurrences of the same effective target once not negated and once negated don't cause it. So, here is a patch which splits the test into two, one tests just __int128 atomics, one tests long long atomics, on lp64 targets we can test both. And the effective targets are so that either the target has hw support of the atomics with no need of -latomic, or -latomic is available. 2026-09-30 Jakub Jelinek <[email protected]> PR middle-end/127492 * gcc.dg/pr127492.c: Require int128 effective target and either sync_int_128_runtime or libatomic_available. Drop the long long fallback, but define T only if it is not defined already. * gcc.dg/pr127492-2.c: New test. Reviewed-by: Richard Biener <[email protected]>
