https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108359
--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> --- (In reply to Andrew Macleod from comment #5) > NOt sure we want to do that for this release tho...? Why not? Regression fixes are fine even in stage4, we are still in stage3. I'd go with something like your patch if it fixes this regression and defer more detailed handling of equality shifts for stage1.
