[Bug c++/104094] New: Alias template shown in diagnostic with wrong template parameter name

[Bug c++/104094] New: Alias template shown in diagnostic with wrong template parameter name

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104094

            Bug ID: 104094
           Summary: Alias template shown in diagnostic with wrong template
                    parameter name
           Product: gcc
           Version: 12.0
            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> struct trait;
template<typename A> struct trait { using type = A; };
template<typename B> struct cond { static constexpr bool value = true; };

template<bool> struct enable_if { using type = bool; };
template<> struct enable_if<false> { };

template<typename C>
class pair
{
  template<typename D>
    using constraint = cond<trait<D>>;

  template<typename E, typename enable_if<constraint<E>::value>::type = true>
    pair(E&&) { }
};

pair<int> p(1);


GCC prints:

name.C:18:14: error: 'pair<C>::pair(E&&) [with E = int; typename
enable_if<cond<trait<D> >::value>::type <anonymous> = true; C = int]' is
private within this context
   18 | pair<int> p(1);
      |              ^
name.C:15:5: note: declared private here
   15 |     pair(E&&) { }
      |     ^~~~


Why does it show cond<trait<D>> in the error? There is no D in that
constructor's parameter list, and D is not explained in the [with ...] template
arg list.

I think it should be shown as constraint<E> or as cond<trait<E>>, but showing D
is just confusing.

[Bug c++/104094] Alias template shown in diagnostic with wrong template parameter name

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104094

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2022-01-19
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

[Bug c++/104094] Alias template shown in diagnostic with wrong template parameter name

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104094

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2022-01-19 00:00:00         |2024-1-17

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Another example with recent trunk:

In file included from
/home/jwakely/src/gcc/gcc/libstdc++-v3/testsuite/std/text_encoding/requirements.cc:20:
/home/jwakely/gcc/14/include/c++/14.0.1/ranges: In function 'constexpr _Cont
std::ranges::to(_Rg&&, _Args&& ...)':
/home/jwakely/gcc/14/include/c++/14.0.1/ranges:9332:21: warning: typedef 'using
_RefT = std::ranges::range_reference_t<_Pattern>' locally defined but not used
[-Wunused-local-typedefs]
 9332 |               using _RefT = range_reference_t<_Rg>;
      |                     ^~~~~

Where does _Pattern come from? There is no _Pattern in that function template.
The source line shows it should be range_reference_t<_Rg> instead.

_Pattern is used in lazy_split_view, split_view and join_view, none of which
are used at all by `ranges::to`. Is there some has collision going on here when
the pretty printer finds names? Does it need an additional disambiguation step
if more than one name is found?

[Bug c++/104094] Alias template shown in diagnostic with wrong template parameter name

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104094

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #2)
> _Pattern is used in lazy_split_view, split_view and join_view, none of which
> are used at all by `ranges::to`. Is there some has collision going on here
> when the pretty printer finds names? Does it need an additional
> disambiguation step if more than one name is found?

Maybe that is PR 66968 (or PR 99) which I see are already linked here?

[Bug c++/104094] Alias template shown in diagnostic with wrong template parameter name

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104094

--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Yeah, maybe this is just a dup of my older bug report. It's been frustrating me
for years.

And obviously I meant "hash collision" above not "has collision" :-)