https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122172
--- Comment #17 from Hans-Peter Nilsson <hp at gcc dot gnu.org> --- FWIW, a patch constructed from comments #2, #3 and the fixup in comment #16 yields two regressions at r16-4259-g9242a89d088f44 for cris-elf compared to r16-3809-g46028a2f40329b (not counting the usual churn of tests, passing and failing), which is at a level I'd expect: g++.sum g++-dg-lto-devirt-2-01.exe g++.sum g++.dg/ipa/devirt-2.C
