https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89536
--- Comment #15 from Jakub Jelinek <jakub at gcc dot gnu.org> --- How is BIT_NOT_EXPR expanded for the prec > 1 BOOLEAN_TYPEs btw? If it is normal QImode or SImode etc. one's complement, then I'd say it is a bug if match.pd generates such BIT_NOT_EXPRs.