https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124565
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> --- Though, I've tried to reproduce it again (last night I just wanted to get the bootstrap/regtest done, so considered the i386.cc change as the only possible breakage and reverted it and wiped the build tree) and it currently is in stage3 while previously it failed earlier. So maybe it was just some random glitch (though, something like that hasn't happened to me on this box for years). It is non-PGO/LTO bootstrap, so I think it should be reproduceable...
