https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106211
--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The question comes about variadic templates and deduction of auto.
>Further, all three compilers also accept this minor modification:
This one does need any deduction though so it is not even related.
So let's take the slightly reduced testcase:
template<typename, typename>
concept any = true;
auto t = []<typename... Ts>(Ts...) {
return [](any<Ts> auto... b, auto c) {return c;};
}(1);
auto t1 = t(2, 3);
We have now:
Ts being <int>
so we get:
return [](any<int> auto, auto) {};
Which is invalid as any<int> is invalid.
Maybe I misunderstand how this works though.
If we remove the any<Ts> then we get the same failure for clang as GCC:
<source>:7:11: error: no matching function for call to object of type '(lambda
at <source>:5:12)'
auto t1 = t(2, 3);
^
<source>:5:12: note: candidate function [with b:auto = <>, c:auto = int] not
viable: requires single argument 'c', but 2 arguments were provided
return [](auto... b, auto c) {return c;};
^
Which is to say the problem is the way variadic templates with auto is being
handled. I suspect the concept is being handled incorrectly for clang.
