https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125653
--- Comment #12 from Richard Sandiford <rsandifo at gcc dot gnu.org> --- Created attachment 64667 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64667&action=edit Candidate patch The early return bypassed the later: /* If STMT is a phi, remove any information recorded for its arguments. */ if (is_a <gphi *> (stmt)) reprocess_inputs (stmt); I'm testing the attached patch against the GCC testsuite, but it'd be great if you could check that it fixes glibc.
