function to pointer reference conversion

Question

I am wondering why the following code is correct:

void foo(){}
void foo2(void(*)()){};
void foo3(void(*)()&){};

int main(){
  foo; // type void(&)() lvalue
  foo2(foo); // void(&)() -> void(*)() function to pointer conversion
  foo3(foo); // ?? conversion
}

According to the function to pointer conversion here

An lvalue of function type T can be converted to a prvalue of type “pointer to T”. The result is a pointer to the function.

The conversion is fine from the void(&)() to the void(*)(), but not void(*)()&.

What is the reason for this code snippet (especially line with ??) to be correct?

Answer 1

The parameter of foo3 which is void(*)()& is actually a function pointer with a ref-qualifier. This is not allowed by the standard.

The C++17 standard draft n4659 states :

[dcl.fct]/6

A function type with a cv-qualifier-seq or a ref-qualifier (including a type named by typedef-name) shall appear only as:

(6.1) — the function type for a non-static member function,
(6.2) — the function type to which a pointer to member refers,
(6.3) — the top-level function type of a function typedef declaration or alias-declaration,
(6.4) — the type-id in the default argument of a type-parameter, or
(6.5) — the type-id of a template-argument for a type-parameter.

The parameter of foo3 does not meet any of the criteria above. The closest to what you are trying to do is (6.2).

So you can have this instead:

void foo(){}
class C {
public:
    void foo1() & {}
};
void foo2(void(*)()){};
void foo3(void(C::*)()&){};

int main(){
  (void) foo; 
  foo2(foo); 
  foo3(&C::foo1); 
}

This compiles on both GCC and Clang. Demo here.