@jensmaurer Could you please take a look at this issue?

I'm not sure there is a real bug here. What exactly goes wrong if struct C; (which declares a class C) creates a class scope which happens to contain no declarations? Note that the "if any" is a nod toward the understanding that struct C; is a class declaration (but not a definition), so this seems intentional.

If it were, I think the argument enforces this #4775 issue. Consider this example

template < class D >
struct B {
  D::type x;
};
struct A { using type = int; };
struct C;  //#1 if it introduced a class scope of C
struct C : A, B< C >{
};

That means from the point "P" at B<C> in the base-specifier, there is no yet a member-declaration in this scope, hence, we could say S(type, C) is empty, its base classes should continue to be searched as per [class.member.lookup#4], and the note in [class.member.lookup] also says that

[Note 2: If T is incomplete, only base classes whose base-specifier appears before P are considered. If T is an instantiated class, its base classes are not dependent. — end note]

Is the base class A considered here since its base-specifier appears before P and the class C is incomplete at this point? Which obviously conflicts with the note in [basic.scope.class].

In addition, Is that "if any" maybe just a nod toward that the class-specifier with an empty "member-specification"? I cannot figure out a normative rule that can interpret the above example except that we consider that a class scope is only introduced after the closing "{" of a class-specifier and a declaration(not a definition) of a class does not introduce the scope.