https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92462
--- Comment #5 from Aleksei Voitylov <aleksei.voity...@bell-sw.com> --- (In reply to Richard Biener from comment #3) > Indeed -fno-strict-aliasing might be a workaround (apart from the atomicity > issue). Also using a character type for the access (uint8_t is _not_ a > character type) would make the alias issue go away. Actually, it is not. However, -fno-tree-pre can be used as such but the point is why a workaround is needed here in the first place.
