https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117311
--- Comment #3 from H. Peter Anvin <hpa at zytor dot com> --- It does, in fact, work just fine under -O0, although it will redundantly manifest the frame pointer in a different register (which is not a problem.) Now, it would seem to me that if this *isn't* something guaranteed to work, then we have a much bigger problem. Do you have a recommendation for how to deal with this, preferably that doesn't impose an undue burden on gcc development?
