[Bug libstdc++/120625] New: [15/16 Regression] std::formatter<__disabled> specializations cause errors in user code

[redi at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120625

            Bug ID: 120625
           Summary: [15/16 Regression] std::formatter<__disabled>
                    specializations cause errors in user code
           Product: gcc
           Version: 16.0
            Status: UNCONFIRMED
          Keywords: rejects-valid
          Severity: normal
          Priority: P3
         Component: libstdc++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: redi at gcc dot gnu.org
  Target Milestone: ---

#include <format>

struct X { };

template<typename T>
concept is_X = std::is_empty_v<T> && std::is_same_v<X, T>;

template<is_X T, typename C> struct std::formatter<T, C> { /* ... */ };

// instantiate the formatter specialization
auto s = sizeof(std::formatter<X, char>);


This compiles with GCC 14 or with -std=c++20 but not -std=c++23

This is due to:

#if defined _GLIBCXX_USE_WCHAR_T && __glibcxx_format_ranges
  // _GLIBCXX_RESOLVE_LIB_DEFECTS
  // 3944. Formatters converting sequences of char to sequences of wchar_t

  namespace __format { struct __disabled; }

  // std::formatter<__disabled, C> uses the primary template, which is
disabled.
  template<>
    struct formatter<char*, wchar_t>
    : private formatter<__format::__disabled, wchar_t> { };


I think we need the __disabled tag type to be complete so that we don't break
constraints on program-defined specializations of std::formatter.