[BUG] [GCC] UFCS with qualified function fails

[JohelEGP]

Title: [GCC] UFCS with qualified function fails.

Description:

Clang and MSVC accept: https://compiler-explorer.com/z/qb19TEGv1.
I haven't reported this before because I was sure that
UFCS is for member call syntax and a qualified function doesn't make sense there.
But Herb has used such calls before, and it might be actually useful: #741 (comment).

Minimal reproducer (https://cpp2.godbolt.org/z/5MTr6jcsh):

ns: namespace = {
  f: (_) = { }
}
main: () = {
  _ = 0.ns::f();
}

Commands:

cppfront main.cpp2
g++14 -std=c++23 -pedantic-errors -Wall -Wextra -Wconversion -Werror=unused-result -I . main.cpp

Expected result:

• A diagnostic by Cppfront, or
• GCC to work (maybe it's their bug), or
• if this is a feature, for a qualified function in a member function call syntax
  to lower without the UFCS macro as a free function call because a member call would never work anyways.

Actual result and error:

Cpp2 lowered to Cpp1:

//=== Cpp2 type declarations ====================================================


#include "cpp2util.h"

namespace ns {

}


//=== Cpp2 type definitions and function declarations ===========================

namespace ns {
  auto f([[maybe_unused]] auto const& param1) -> void;
}
auto main() -> int;
  

//=== Cpp2 function definitions =================================================

namespace ns {
  auto f([[maybe_unused]] auto const& param1) -> void{}
}
auto main() -> int{
  static_cast<void>(CPP2_UFCS_0(ns::f, 0));
}

Output:

build/main.cpp: In lambda function:
build/main.cpp:27:33: error: 'ns::f' is not a class member
   27 |   static_cast<void>(CPP2_UFCS_0(ns::f, 0));
      |                                 ^~
raw.githubusercontent.com/hsutter/cppfront/main/include/cpp2util.h:761:47: note: in definition of macro 'CPP2_UFCS_0'
  761 |     if constexpr (requires{ CPP2_FORWARD(obj).FUNCNAME(); }) { \
      |                                               ^~~~~~~~
build/main.cpp:27:33: error: 'ns::f' is not a class member
   27 |   static_cast<void>(CPP2_UFCS_0(ns::f, 0));
      |                                 ^~
raw.githubusercontent.com/hsutter/cppfront/main/include/cpp2util.h:762:34: note: in definition of macro 'CPP2_UFCS_0'
  762 |         return CPP2_FORWARD(obj).FUNCNAME(); \
      |                                  ^~~~~~~~

[JohelEGP]

and it might be actually useful: #741 (comment)

Considering Herb's comment (#741 (comment)),
it could actually be a feature because

it can help reduce the pressure to add special-purpose language features like |> and reduce pressure to use special-purpose library techniques like overloading |

Those accept |> ranges::to and | ranges::to,
unlike UFCS on GCC with .ranges::to.

[JohelEGP]

a qualified function doesn't make sense there

a member call would never work anyways

Turns out there's a case.
Calling the member of a base type.
https://compiler-explorer.com/z/K8zqsGzPP:

struct B {
  void f();
};
struct D : B { };
void g() {
  D d;
  d.B::f();
}

So 00ef370 is a wrong fix.

[JohelEGP]

I think this is IFNDR, and GCC is right.
From https://eel.is/c++draft/temp.res.general#6.1:

The program is ill-formed, no diagnostic required, if:
(6.1) no valid specialization, ignoring static_assert-declarations that fail ([dcl.pre]), can be generated for a templated entity or a substatement of a constexpr if statement ([stmt.if]) within a templated entity and the innermost enclosing template is not instantiated, or

[JohelEGP]

Calling the member of a base type.

But all reject the UFCS macro on this (https://compiler-explorer.com/z/fe3j4ex9Y, https://cpp2.godbolt.org/z/xKT6dzTjr):

namespace N {
struct B {
  void f();
};
}
struct D : N::B { };
void g() {
  D d;
  d.B::f(); // OK.
  static_cast<void>(
    [&](auto&& obj) {
      if constexpr (requires { obj.B::f(); })
        return obj.B::f();
      else
        return B::f(obj); // Error: No `B`.
    }
  );
}

[JohelEGP]

I think this is IFNDR, and GCC is right.

I was wrong.
I have opened https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111841
after reading https://eel.is/c++draft/temp.dep and https://eel.is/c++draft/basic.lookup.qual#def:lookup_context.

[hsutter]

Thanks for the research. So, my understanding is that this a GCC conformance bug, correct?

[JohelEGP]

That's right.

[hsutter]

Thanks. OK to close?

[JohelEGP]

Yes.