DR: 18.11.4.7: get_accessible_* and the name used for the access

[hubert-reinterpretcast]

The accessibility of f below from C depends upon the name used to name A. It is unclear what the application of get_accessible_member_functions upon the meta-object types produced in g() and h() h2() would generate.

struct A { static int f(); };
class B : A {};
class C : B {
  static auto *g() { return static_cast<reflexpr(::A) *>(0); }
  static auto *h() { return static_cast<reflexpr(A) *>(0); }
  static auto *h2() { return static_cast<
      get_class_t<get_element_t<0, get_base_classes_t<reflexpr(C)>>> *>(0); }
};

[hubert-reinterpretcast]

Ditto get_accessible_member_types, etc.

[hubert-reinterpretcast]

The injected-class-name for A is private to B from C, so the h() case is invalid.

[hubert-reinterpretcast]

More of a note-to-self: using-declarations can also change access.

[hubert-reinterpretcast]

I think this issue is not at all straightforward. Protected non-static members are also complicated.

[Axel-Naumann]

IMO that's a design issue out of scope for the wording review.