https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121436
--- Comment #2 from Sam James <sjames at gcc dot gnu.org> --- (In reply to Antoni from comment #1) > This test passes for me on x86-64. > The goal of this patch was to have the correct sign for char so that some > x86-64 SIMD intrinsics would work IIRC, so -fsigned-char is not necessary. > From what I understand, this test should only be enabled on targets where > char is signed by default. I don't follow this -- on such targets, wouldn't the sign of char already be correct before the change? It also does: - build_common_tree_nodes (false); + build_common_tree_nodes (flag_signed_char); so surely it's supposed to support toggling it?
