https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121231
--- Comment #3 from Gašper Ažman <gasper.azman at gmail dot com> --- Would that make it a /way/ reduced test case in that case? PR 94511 seems to be value-dependent. In this case, both clang and msvc accept and generate correct code. PR 94511 seems like you needed to tweak it to be accepted? Also, I hope not, that bug's been present for years (!)
