https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125838
--- Comment #16 from Alfie Richards <alfierichards at gcc dot gnu.org> --- :) I imagine preserve_none's use case is pretty much orthogonal to variadic args use case. So I don't think it's worth trying to get Clang to change their behavior even if I marginally prefer ours/it may have slight performance benefits. I think the best option is to change our behavior to match theirs and fall back to AAPCS when using variadic args.
