https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71598
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> --- Note I can't find any bit in the C standard that would make the testcase well-defined and support reporters view. My clang version doesn't "miscompile" it though. With LTO we'd treat both enum declarations as compatible wrt TBAA but they wouldn't be compatible with unsigned int (so in GCC speak both enums get the same TYPE_CANONICAL but that isn't TYPE_CANONICAL of unsigned int).
