https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89766
--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> --- (In reply to Rimvydas (RJ) from comment #10) > Using 9.0.1 20190319 as reference several ICE cases reduce down to the same > snippet (regression on trunk)? None of these ICE on the trunk, there is a warning on the first one: warning: inline function ‘constexpr a::a()’ used but never defined What do you think the compiler does incorrectly? E.g. for the first testcase I guess it would be rejected if it is actually instantiated, but when not instantiated, I'd think it falls into the invalid, no diagnostic required category.