[Bug c++/114067] New: GCC gives wrong diagnostic in the definition of a static data member of same class type

[Bug c++/114067] New: GCC gives wrong diagnostic in the definition of a static data member of same class type

[jlame646 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114067

            Bug ID: 114067
           Summary: GCC gives wrong diagnostic in the definition of a
                    static data member of same class type
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jlame646 at gmail dot com
  Target Milestone: ---

The following program produces wrong diagnostic with gcc and clang:

```
struct A{
    int x;
    static const A a{1};
};
```
GCC says `in-class initialization of static data member 'const A A::a' of
incomplete type`.

While the actual problem is not that `A` is incomplete but that it is not an
integral or enumeration type. 

Only msvc gives correct diagnostic saying:

```
 error C2864: 'A::a': a static data member with an in-class initializer must
have non-volatile const integral type or be specified as 'inline'
```

So I think gcc should say something similar to msvc instead of saying that `A`
being incomplete is the problem.

[Bug c++/114067] GCC gives wrong diagnostic in the definition of a static data member of same class type

[jlame646 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114067

--- Comment #1 from Jason Liam <jlame646 at gmail dot com> ---
That is, even if the type of `a` was complete the program would've been
ill-formed. So saying only that `A` is incomplete is the problem doesn't seem
right.

[Bug c++/114067] GCC gives wrong diagnostic in the definition of a static data member of same class type

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114067

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |diagnostic

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---

Actually A at that point is incomplete type.

Note MSVC gives really bad diagnostic for:
```
struct A{
    int x;
    static constexpr const A a{1};
};
```
says undefined type A rather than incomplete type.

Anyways GCC's check for incomplete type happens before the other checks.
Plus GCC allows it with -fpermissive ....

[Bug c++/114067] GCC gives wrong diagnostic in the definition of a static data member of same class type

[jlame646 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114067

--- Comment #3 from Jason Liam <jlame646 at gmail dot com> ---
(In reply to Andrew Pinski from comment #2)
> Actually A at that point is incomplete type.

Yes, it is incomplete at that point but that is not the reason for the program
to be ill-formed. The reason is that `A` is a class-type and not an
integral/enumeration type.

[Bug c++/114067] GCC gives wrong diagnostic in the definition of a static data member of same class type

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114067

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Jason Liam from comment #3)
> (In reply to Andrew Pinski from comment #2)
> > Actually A at that point is incomplete type.
> 
> Yes, it is incomplete at that point but that is not the reason for the
> program to be ill-formed. The reason is that `A` is a class-type and not an
> integral/enumeration type.

As I mentioned there is a extension which allows it be accepted (only with
-fpermissive) if it was complete. And the checking for it being complete
happens before that other check. Now I fully looked into how fix this so the
diagnostic for the completeness happens after allowing the extension because it
seems like the check for completeness happens way before in a different
function.