[Bug c++/109340] New: Inconsistent diagnostics for invalid member types in union

[Bug c++/109340] New: Inconsistent diagnostics for invalid member types in union

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109340

            Bug ID: 109340
           Summary: Inconsistent diagnostics for invalid member types in
                    union
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Keywords: diagnostic
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: redi at gcc dot gnu.org
  Target Milestone: ---

template<typename T>
union U
{
  U();

  T t;
};

U<int&> ref;
U<int[]> arr;
U<int()> func;

G++ correctly rejects this, but has three very different errors for basically
the same problem:

union.cc: In instantiation of ‘union U<int&>’:
union.cc:9:9:   required from here
union.cc:6:5: error: non-static data member ‘U<int&>::t’ in a union may not
have reference type ‘int&’
    6 |   T t;
      |     ^
union.cc: In instantiation of ‘union U<int []>’:
union.cc:10:10:   required from here
union.cc:6:5: error: flexible array member ‘U<T>::t’ in union
union.cc: In instantiation of ‘union U<int()>’:
union.cc:11:10:   required from here
union.cc:6:5: error: data member ‘U<int()>::t’ invalidly declared function type


Specifically:

error: non-static data member ‘U<int&>::t’ in a union may not have reference
type ‘int&’

error: flexible array member ‘U<T>::t’ in union

error: data member ‘U<int()>::t’ invalidly declared function type

Could we use the same phrasing for all three?

The first one seems the best, and the third one seems the worst.

For comparison, Clang says:

union.cc:6:5: error: union member 't' has reference type 'int &'
  T t;
    ^
union.cc:9:9: note: in instantiation of template class 'U<int &>' requested
here
U<int&> ref;
        ^
union.cc:6:5: error: data member instantiated with function type 'int ()'
  T t;
    ^
union.cc:11:10: note: in instantiation of template class 'U<int ()>' requested
here
U<int()> func;
         ^
2 errors generated.


Not very good either.

EDG says:

"union.cc", line 6: error: incomplete type is not allowed
    T t;
      ^
          detected during instantiation of union "U<T> [with T=int []]" at line
                    10

"union.cc", line 6: error: a function type is not allowed here
    T t;
      ^
          detected during instantiation of union "U<T> [with T=int ()]" at line
                    11

2 errors detected in the compilation of "union.cc".

[Bug c++/109340] Inconsistent diagnostics for invalid member types in union

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109340

--- Comment #1 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jonathan Wakely <redi@gcc.gnu.org>:

https://gcc.gnu.org/g:31a909712014b75fc6ae2ca5eaa425f218bb5f32

commit r13-6934-g31a909712014b75fc6ae2ca5eaa425f218bb5f32
Author: Jonathan Wakely <jwakely@redhat.com>
Date:   Wed Mar 29 22:16:55 2023 +0100

    libstdc++: Use std::remove_cv_t in std::optional::transform [PR109340]

    We need to strip cv-qualifiers from the result of the callable passed to
    std::optional::transform.

    For std::expected::transform and std::expected::transform_error I
    noticed we were stripping cv-qualifiers but were also incorrectly
    stripping references.

    libstdc++-v3/ChangeLog:

            PR libstdc++/109340
            * include/std/expected (expected::transform): Use
            std::remove_cv_t instead of std::remove_cvref_t.
            (expected::transform_error): Likewise.
            (expected<cv void, E>::transform): Likewise.
            (expected<cv void, E>::transform_error): Likewise.
            * include/std/optional (transform): Use std::remove_cv_t.
            * testsuite/20_util/optional/monadic/pr109340.cc: New test.

[Bug c++/109340] Inconsistent diagnostics for invalid member types in union

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109340

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Gah, wrong ID in that commit message.