Add an `__is_inplace_index` metafunction.

[mpark]

Summary

This is used to constrain variant's converting constructor correctly.

[EricWF]

include/utility

937

Missing __uncvref_t . Is this intentional?

[mpark]

include/utility

937

Oh, I can add that. As far as I understand, we're required to perform decay_t from variant's side anyway,
so uncvref would be a no-op for that case. But I guess that's fine?

[mpark]

Add __uncvref.

[EricWF]

include/utility

937

Should be fine. I guess it doesn't matter. but it makes me a bit uncomfortable to have two different metafunctions, one that does decay and another that doesn't.

[mclow.lists]

LGTM. Do we need any tests?

[mpark]

In D34111#778637, @mclow.lists wrote:

LGTM. Do we need any tests?

I could add some. Where do you think they should live?

[mclow.lists]

I could add some. Where do you think they should live?

libc++ specific tests (which this would definitely be) live in test/libcxx/.

How about test/libcxx/utilities/meta/meta.unary ?

[mpark]

Add tests.

[mpark]

Lined up the static_asserts.

[mpark]

Add more tests.

[mclow.lists]

LGTM!

include/utility

 constexpr in_place_index_t<_Idx> in_place_index{};
 
 template <class _Tp> struct __is_inplace_type_imp : false_type {};
 template <class _Tp> struct __is_inplace_type_imp<in_place_type_t<_Tp>> : true_type {};
 
 template <class _Tp>
 using __is_inplace_type = __is_inplace_type_imp<__uncvref_t<_Tp>>;
 
+template <class _Tp> struct __is_inplace_index_imp : false_type {};
+template <size_t _Idx> struct __is_inplace_index_imp<in_place_index_t<_Idx>> : true_type {};
+
+template <class _Tp>
+using __is_inplace_index = __is_inplace_index_imp<__uncvref_t<_Tp>>;
+
 #endif // _LIBCPP_STD_VER > 14
 
 template <class _Arg, class _Result>
 struct _LIBCPP_TEMPLATE_VIS unary_function
 {
     typedef _Arg    argument_type;
     typedef _Result result_type;
 };