[Bug libstdc++/106376] New: The implementation of std::pointer_traits seems wrong

[Bug libstdc++/106376] New: The implementation of std::pointer_traits seems wrong

[hewillk at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106376

            Bug ID: 106376
           Summary: The implementation of std::pointer_traits seems wrong
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libstdc++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hewillk at gmail dot com
  Target Milestone: ---

[pointer.traits.types]: 
using element_type = see below;
Type: Ptr​::​element_­type if the qualified-id Ptr​::​element_­type is valid
and denotes a type ([temp.deduct]); otherwise, T if Ptr is a class template
instantiation of the form SomePointer<T, Args>, where Args is zero or more type
arguments; otherwise, the specialization is *ill-formed*.

So for the following, the instantiation of pointer_traits will be ill-formed,
and static_assert will cause a hard error, just like libc++ and MSVC-STL do.
libstdc++ specifies element_type as a meaningless type, so the static_assert
still   passes.

Although the behavior of libstdc++ is incorrect, but I am not feel good about
this static_assert hard error..

#include <iterator>

struct I {
  using iterator_category = std::contiguous_iterator_tag;
  using difference_type = std::ptrdiff_t;
  using value_type = int;
  value_type& operator*() const;
  I& operator++();
  I operator++(int);
  I& operator--();
  I operator--(int);
  I& operator+=(difference_type);
  I& operator-=(difference_type);
  value_type* operator->() const;
  value_type& operator[](difference_type) const;
  friend I operator+(I, difference_type);
  friend I operator+(difference_type, I);
  friend I operator-(I, difference_type);
  friend difference_type operator-(I, I);
  auto operator<=>(const I&) const = default;
};

static_assert(std::contiguous_iterator<I>);

https://godbolt.org/z/sx763oMn6

[Bug libstdc++/106376] The implementation of std::pointer_traits seems wrong

[hewillk at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106376

--- Comment #1 from 康桓瑋 <hewillk at gmail dot com> ---
oops, this is basically https://cplusplus.github.io/LWG/issue3545, I am waiting
for your paper.

[Bug libstdc++/106376] [LWG3545] The implementation of std::pointer_traits seems wrong

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106376

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |SUSPENDED
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2022-07-21
            Summary|The implementation of       |[LWG3545] The
                   |std::pointer_traits seems   |implementation of
                   |wrong                       |std::pointer_traits seems
                   |                            |wrong

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to 康桓瑋 from comment #0)
> libstdc++ specifies element_type as a meaningless type, so the static_assert
> still   passes.
> 
> Although the behavior of libstdc++ is incorrect, but I am not feel good
> about this static_assert hard error..

Right, it's completely unusable, see PR 96416.

Libstdc++ already implements my proposed solution, the standard just hasn't
caught up yet ;-)

Suspending until 3545 is resolved, but I expect to close this at some future
date, as there's nothing for libstdc++ to do here.

[Bug libstdc++/106376] [LWG3545] The implementation of std::pointer_traits seems wrong

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106376

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |WORKSFORME
             Status|SUSPENDED                   |RESOLVED

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Closing ,as libstdc++ matches the C++23 working paper.