https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81847
--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Valentine from comment #2) > (In reply to Andrew Pinski from comment #1) > > Try -flto-partition=one which might allow this to be reduced further. > > Thanks, just tried -flto-partition=one, still the same ICE. Other > -flto-partition options didn't help either, the only thing that fixes this > ICE for me is the "none" option. Right, the whole point of my saying: >Try -flto-partition=one which might allow this to be reduced further. Was that the deterministic of the ICE becomes very deterministic now and that even adding another .cpp file, it will still crash :). See https://gcc.gnu.org/wiki/A_guide_to_testcase_reduction#Reducing_LTO_bugs for other tricks of reducing the testcase to something which you can provide.
