https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121478
--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Andrew Pinski from comment #6) > This patch has a few regressions. I think due to convert_for_assignment > part. I will look at fixing that in a few. Simple fix, use omit_one_operand_loc instead of build_compound_expr in both convert_arguments and convert_for_assignment. That seems to have fixed the regressions in my quick testing. So full build/testing now.
