https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71479
--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> --- BTW, for __builtin_add_overflow_p with _Bool as last argument (which isn't rejected, going to fix that momentarily) we don't emit the expected test whether the result is 0/1, but whether it fits 8 bits unsigned. Also, _Complex _Bool or _Complex enum ... is something way too weird for the middle-end.
