Explicit this object parameter wonkiness

Question

There was some code floating around on Reddit that defined a member-function with an explicit this object parameter defined as type int. This made me wonder how this member-function could possibly be invoked.

After doing some tests, all of the compilers (Clang, GCC, and MSVC) seem to produce different results. So, this raises the question; is the following code legal and in accordance with the C++23 standard?

struct w { constexpr bool f(this int) { return true; } };
static_assert((*&w::f)(1)); // clang ok, gcc ok, msvc nope
static_assert((*&w::f)(0)); // clang ok, gcc nope, msvc nope

Demo

GCC's error message:

<source>:3:23: error: non-constant condition for static assertion
    3 | static_assert((*&w::f)(0));
      |               ~~~~~~~~^~~
<source>:3:24: error: dereferencing a null pointer
    3 | static_assert((*&w::f)(0));
      |

MSVC's error message:

<source>(2): error C2660: 'w::f': function does not take 1 arguments
<source>(1): note: see declaration of 'w::f'
<source>(2): note: while trying to match the argument list '(int)'
<source>(3): error C2660: 'w::f': function does not take 1 arguments
<source>(1): note: see declaration of 'w::f'
<source>(3): note: while trying to match the argument list '(int)'

I think Clang is correct, &w::f undoubtedly gives a function pointer of type bool (*)(int). — 康桓瑋
– 康桓瑋, Commented Jul 11, 2024 at 3:29
@RemyLebeau That's not true. The first parameter is this int, not int. — 康桓瑋
– 康桓瑋, Commented Jul 11, 2024 at 6:56
@user12002570 It does not seem to make this int invalid - that sentence just describes one thing the proposal allows you to do, it is just the only thing the proposal allows you to do. There is even an example that we added that demonstrates that this works. (source: my paper). — Barry
– Barry, Commented Jul 11, 2024 at 15:14
@RemyLebeau, actually &w::f here has the same type as the pointer to free function: gcc.godbolt.org/z/aG6bEPxsd — Fedor
– Fedor, Commented Jul 12, 2024 at 7:30
@RemyLebeau Since C++23, because of the addition of explicit object parameters, a lot of statements that originally were true for all non-static member functions are now only true for implicit object non-static member functions. Non-static member functions with explicit object parameters are somewhere between the old non-static member functions and static member functions in terms of behavior. — user17732522
– user17732522, Commented Jul 12, 2024 at 10:18

user17732522 · Accepted Answer · 2024-07-11 11:10:28Z

4

I also think Clang is the only compiler behaving correctly here.

this int in itself is fine. I do not see any restrictions on the type of the explicit object parameter.

&w::f is clearly defined to give a bool(*)(int) pointing to the member function.

(*&w::f)(1) and (*&w::f)(0) should then be normal calls that match the call arguments in order to the parameters of the function, so 1 or 0 will be the argument to the explicit object parameter. There won't be any problem with the call as 1 and 0 can be used as arguments for an int parameter.

So it looks to me like bugs in GCC and MSVC.

answered Jul 11, 2024 at 11:10

user17732522

78.1k3 gold badges82 silver badges147 bronze badges

Sign up to request clarification or add additional context in comments.

23 Comments

303 Over a year ago

So, w::f is a non-static member-function that, when taking its address (i.e. &w::f), gives back a function pointer of type bool(*)(int)? Which section of the C++23 standard describes this behavior?

303 Over a year ago

Also, could you please quote/reference the section of the C++23 standard that determines the invocation of (*&w::f)(0) as well-defined?

Richard Over a year ago

This answer seems wrong. The program is ill-formed for the reason explained in my answer. Also this isn't a language-lawyered answer.

user17732522 Over a year ago

@user12002570 Obviously obj.f(10) doesn't work. The whole point of explicit object parameters is that it shall be matched against the object expression (obj), not the first argument of the call via member access expression. You didn't leave any parameter to take the 10 argument.

user17732522 Over a year ago

@Alan I do not see any reason that a w object has to be used to call the function.

|

Richard · Accepted Answer · 2024-07-12 06:19:06Z

-3

This is CWG2688 that points out that currently [expr.call] does not properly handle explicit object member functions.

Note the issue is still open.

From CWG2688:

Section: 7.6.1.3 [expr.call]	Status: open	Submitter: Matthew House	Date: 2023-01-16

Subclause 7.6.1.3 [expr.call] paragraphs 1 and 2 do not properly handle explicit object member functions. Such a function can be called via a function pointer, via an lvalue obtained by dereferencing a function pointer, and via a class member access expression.

edited Jul 12, 2024 at 6:19

answered Jul 11, 2024 at 12:24

Richard

47k7 gold badges40 silver badges94 bronze badges

1 Comment

Machavity Over a year ago

Comments have been moved to chat; please do not continue the discussion here. Before posting a comment below this one, please review the purposes of comments. Comments that do not request clarification or suggest improvements usually belong as an answer, on Meta Stack Overflow, or in Stack Overflow Chat. Comments continuing discussion may be removed.

Collectives™ on Stack Overflow

Explicit this object parameter wonkiness

2 Answers 2

23 Comments

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

23 Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related