I tested a variation:

namespace mozilla {
struct Proxy;
}
struct Proxy {
};
using mozilla::Proxy;
Proxy mProxy;

and both gcc and edg refuse to accept mProxy complaining about a definition with incomplete type.

one "last" test both gcc and edg accept. It shows that they are merging the declaration and definition:

namespace mozilla {
struct Proxy;
}
using mozilla::Proxy;
struct Proxy {
int a;
};
mozilla::Proxy mProxy;
int f() {
return mProxy.a;
}

The relevant bit of code (in SemaDecl.cpp):

if (TagDecl *PrevTagDecl = dyn_cast<TagDecl>(PrevDecl)) {
  // If this is a use of a previous tag, or if the tag is already declared
  // in the same scope (so that the definition/declaration completes or
  // rementions the tag), reuse the decl.
  if (TUK == TUK_Reference || TUK == TUK_Friend ||
      isDeclInScope(PrevDecl, SearchDC, S, isExplicitSpecialization)) {

This code makes us not treat the second declaration of Proxy as a redeclaration even though lookup finds the previous declaration. Not sure what the right predicate is.

another testcase:

/* g++ accepts this but clang dignoses:
error3.cpp:18:5:
error: call to 'f' is ambiguous
*/

namespace A {
extern "C" int f(int i = 5) { return i + 2; }
}
namespace B {
extern "C" int f(int i = 5);
}
using A::f;
using B::f;
int main(void) {
/* extern "C" should conceptually nullify overloading

• so why does clang consider this function ambiguous?
  */
  f(3);
  }

@llvm/issue-subscribers-clang-frontend

Author: None (llvmbot)