Allows using-declaration to refer to scoped enumerator

[Quuxplusone]

Bugzilla Link	PR9124
Status	NEW
Importance	P normal
Reported by	Johannes Schaub (schaub.johannes@googlemail.com)
Reported on	2011-02-02 09:55:46 -0800
Last modified on	2016-08-24 04:28:22 -0700
Version	trunk
Hardware	PC Linux
CC	dgregor@apple.com, gonzalo.gadeschi@gmail.com, llvm-bugs@lists.llvm.org, richard-llvm@metafoo.co.uk
Fixed by commit(s)
Attachments
Blocks
Blocked by
See also

Clang is supposed to reject the below, but doesn't

enum class A { B };
int main() {
  using A::B; // should be an error
}

Although the spec (N3225) doesn't say so currently (DR already filed), clang
should presumably also reject referring to an unscoped enumerator in a using
declaration:

enum A { B };
int main() {
  using A::B; // presumably should be error too.
}

I think best may be to forbid the using declaration of having a nested name
specifier referring to an enumeration.

[Quuxplusone]

[namespace.udecl]p7 is clear that we should diagnose this, but it's not obvious why. The functionality is useful and doesn't have any obvious semantic issues.

Johannes: Do you have a DR# for this? I couldn't find it in the issues list, and I'd like to suggest a different resolution :)

[Quuxplusone]

I sent this issue to WMM twice with a year of distance between each iteration,
but he did not respond -.- So there is no DR# yet.

Hello Mike,

7.3.3p7 says

   A /using-declaration/ shall not name a scoped enumerator.

Presumably, this is intended to forbid the following case too.

   enum A { B };
   void f() { using A::B; }

However, A::B is an unscoped enumerator according to 7.2p2. It should
probably forbid the nested-name-specifier of the using-declaration to
name an enumeration, instead.

Thanks,
Johannes.

[Quuxplusone]

Of couse we cannot straight out forbid referring to a unscoped enumerator because it needs still to be allowed as a member declaration if the NNS is a class.

[Quuxplusone]

FWIW with inline variables:

static constexpr auto B = A::B;

Is safe and does the right thing.

While I would like "using A::B" to work, I would rather warn on this until this is allowed by the standard.