https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78959
--- Comment #3 from Martin Sebor <msebor at gcc dot gnu.org> --- Okay, thanks. I can add that but shouldn't the pass still correctly handle the pre-C99 HP-UX behavior (i.e., when HAVE_C99_RUNTIME is not defined)? (In case it's not apparent from the test, it fails because GCC optimizes calls like "sprintf(0, 0, "%hhd", 4105)" to 1 when on pre-C99 HP-UX the function returns 4.
