[Bug c++/67152] New: [concepts] bogus "partial specialization of âfoo<T>â after instantiation" error

["eric.niebler at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67152

            Bug ID: 67152
           Summary: [concepts] bogus "partial specialization of âfoo<T>â
                    after instantiation" error
           Product: gcc
           Version: c++-concepts
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: eric.niebler at gmail dot com
  Target Milestone: ---

The error happens on constrained partial specializations of a template that has
already been instantiated. But the template that got instantiated would fail
the constraint anyway, so the error is complaining about nothing AFAICT.
Consider:

template <class T>
concept bool HasType = requires { typename T::type; };

template<class T>
struct trait {
  using type = void;
};

struct has_type { using type = void; };

// Instantiation here
trait<has_type>::type foo() {}

// constrained version here. Type "has_type" would fail this
// constraint so this partial specialization would not have been
// selected.
template<class T>
  requires !HasType<T>
struct trait<T> {
  using type = void;
};


...yields the following:

test.cpp:17:8: error: partial specialization of âstruct trait<T>â after
instantiation of âstruct trait<has_type>â [-fpermissive]
 struct trait<T> {
        ^

If the partial specialization is the following instead:

template<class T>
struct trait<T*> {
  using type = void;
};

then there is no error. It seems to me that for partial specializations, a
constraints failure should be treated similarly to a pattern match failure WRT
such errors.