https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72817
--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> --- (In reply to amker from comment #7) > Thanks for reporting. > This may indicate something is still wrong in NE_EXPR analysis. I am OoO > right now and will look into this once I get back. This has been an signed vs. unsigned char issue, see #c6. I believe it should be fixed now.