[Bug c++/107429] New: misdiagnosed "constraint depends on itself" in overloaded functions

[Bug c++/107429] New: misdiagnosed "constraint depends on itself" in overloaded functions

[jwjagersma at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107429

            Bug ID: 107429
           Summary: misdiagnosed "constraint depends on itself" in
                    overloaded functions
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jwjagersma at gmail dot com
  Target Milestone: ---

Given the following code:

  struct tag_foo { } inline constexpr foo;
  struct tag_bar { } inline constexpr bar;

  template<typename... T>
  auto f(tag_foo, T... x)
  {
    return (x + ...);
  }

  template<typename... T>
  concept fooable = requires (T... x) { f(foo, x...); };

  template<typename... T> requires (fooable<T...>)
  auto f(tag_bar, T... x)
  {
    return f(foo, x...);
  }

  auto test()
  {
    return f(bar, 1, 2, 3);
  }

g++ claims that "satisfaction of atomic constraint [...] depends on itself",
but the overload of 'f()' that is checked by 'fooable' is not constrained in
any way.

I am not 100% sure if this is valid code, but clang and msvc do not make the
same complaint, so I'm inclined to believe g++ is in error here.

[Bug c++/107429] misdiagnosed "constraint depends on itself" in overloaded functions

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107429

Patrick Palka <ppalka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
                 CC|                            |ppalka at gcc dot gnu.org
         Resolution|---                         |DUPLICATE

--- Comment #1 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Thanks for the bug report.  This is a consequence of CWG2369, and the
constraint recursion diagnosis is strictly speaking correct correct.  I think
other compilers accept it because they don't fully implement
https://wg21.link/cwg2369 which moved the constraint satisfaction check during
template argument deduction to _before_ the checking non-dependent conversions.

A workaround (that should be valid pre- and post-CWG2369) is to encode the
non-dependent conversion in the constrained overload of f as an additional
constraint that's checked before the fooable constraint:

  template<std::convertible_to<tag_bar> U, typename... T> requires
(fooable<T...>)
  auto f(U, T... x)

see PR99599 and its dups for more info

*** This bug has been marked as a duplicate of bug 99599 ***

[Bug c++/107429] misdiagnosed "constraint depends on itself" in overloaded functions

[jwjagersma at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107429

--- Comment #2 from jwjagersma at gmail dot com ---
Thanks, I thought I searched but somehow missed that PR.
From a user perspective it seems very surprising that this workaround compiles,
but the obvious implementation doesn't.  I do hope it gets addressed someday.