CWG3137 [expr.type] Composite pointer type requiring both qualification conversion and pointer-to-member conversion

[t3nsor]

Full name of submitter: Brian Bi

Issue description: There is a gap in the specification of composite pointer type if both a pointer-to-member conversion and a qualification conversion are needed to achieve a common type. However, all implementations accept at least some such cases.

using P = int*;
using CP = const int* const;
struct B {
    P a;
};
struct D : B {};

constexpr P (B::*bnc) = &B::a;
constexpr CP (B::*bc) = &B::a;
constexpr P (D::*dnc) = &B::a;
constexpr CP (D::*dc) = dnc;  // work around GCC bug

static_assert(bnc == dc);
static_assert(bc == dnc);  // EDG rejects

Suggested resolution: Edit [expr.type]/3.7 and add sub-bullets as follows. Note that the cross-reference to [dcl.init.ref] is struck because it already occurs in an earlier bullet.

• if T1 is “pointer to member of C1 of type cv1 U1” and T2 is “pointer to member of C2 of type cv2 U2”, for some non-function type U, where

  • U1 and U2 are similar object types and
  • C1 is reference-related to C2 or C2 is reference-related to C1 ([dcl.init.ref]),

  the qualification-combined type of T2 and T1 or the qualification-combined type of T1 and T2, respectively;

[lprv]

The bullet immediately following this one is made non-operational by this change and can also be struck.

(I'd also drop the "object" from "similar object types", I don't think it's needed.)

[jensmaurer]

CWG3137

I agree that "object" type is not necessary (dropped). I've left the ", for..." formulation, because it attaches the "respectively" better to the "or" to which it applies.

I'm not seeing why the immediately following bullet (talking about similar types in general) is now unreachable.

[lprv]

Similar pointer types are handled by 3.5, this change moves the handling of similar member pointer types into 3.7, leaving nothing for 3.8.

(BTW: the proposed resolution on the issue page still mentions sub-bullets, despite there not being any in the diff.)