I remember a similar example from https://wg21.link/p0315r4, see point 6.

On Thu, Apr 27, 2023 at 6:14 PM Brian Bi via Std-Discussion <
std-discussion_at_[hidden]> wrote:

>
> On Thu, Apr 27, 2023 at 11:08 AM Daniel Krügler via Std-Discussion <
> std-discussion_at_[hidden]> wrote:
>
>> Am Do., 27. Apr. 2023 um 17:05 Uhr schrieb Thiago Macieira via
>> Std-Discussion <std-discussion_at_[hidden]>:
>> >
>> > Is this intended with C++20's class-type NTTPs?
>>
>> No,
>>
>> see
>>
>> https://cplusplus.github.io/CWG/issues/cwg_active.html#2542
>
>
> But this doesn't really answer the OP's question. Consider that we can
> simply do this:
>
> template <class T> struct Structural {};
>
> template <auto Call = Structural<decltype([]{})>{}> struct S {};
>
> Now the template parameter type is structural, but there is still a
> question of whether the lambda type is the same every time `S<>` is
> referenced.
>
>
>>
>> > template <auto Call = []{}> struct S {};
>> > static_assert(!std::is_same_v<S<>, S<>>);
>> >
>> > It passes the compilation in the three major compilers that have
>> implemented
>> > the feature in the first place:
>> > https://conformance.godbolt.org/z/edf8Pv71j
>> >
>> > I'm not asking to change this. I actually want it this way. But I'd
>> like a
>> > confirmation that this is the intention and won't suddenly break my
>> code later.
>>
>> It is not intended and it is likely that compilers will adjust to make
>> that ill-formed.
>>
>> - Daniel
>> --
>> Std-Discussion mailing list
>> Std-Discussion_at_[hidden]
>> https://lists.isocpp.org/mailman/listinfo.cgi/std-discussion
>>
>
>
> --
> *Brian Bi*
> --
> Std-Discussion mailing list
> Std-Discussion_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-discussion
>

-- 
Regards,
Oleksandr Koval.