https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119362
--- Comment #4 from John David Anglin <danglin at gcc dot gnu.org> --- Build fixed by https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;h=6c443e128802bd93158a3db7c4edf5fc1fc76c8d. What bothers me is the above change was to fix a performance regression caused by a more aggressive use of invariant equivs. This is a problem with skipping the gain reduction for invariant equivs that are invalid (not consumed). I worry the current code still may skip equivs that are invalid and shouldn't be used.
