https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125838
--- Comment #4 from Alfie Richards <alfierichards at gcc dot gnu.org> --- Ah thats not ideal. Sorry about this. I don't think we're in too bad a state as the preserve_none ABI isn't stable or even properly defined anywhere. I think there's an argument that using all the registers for variadic args is the sensible thing to do, and in the spirit of preserve_none, so we could argue this is the sensible thing. However, we do want to have compatibility with clang, and as it was originally a clang thing they defacto specify its behaviour. So we probably ought to fix this one way or another. I will reach out to some LLVM/Clang people and see if they have a different point of view.
