klimek added a comment. Ok, so just for my understanding: the nested name specifier in the changed tests canonicaltype is *not* the user-written type (thus, what the elaborated type would get us), but the full nested name specifier of the canonical type? http://reviews.llvm.org/D11797