| Issue |
182232
|
| Summary |
libc++: Nested variant compilation issue on use of `operator==` in C++23 mode
|
| Labels |
libc++
|
| Assignees |
|
| Reporter |
devgs
|
Here's a simplified test case.
```
#include <vector>
#include <variant>
struct Array;
struct A1 { auto operator<=>(const A1 &) const = default; };
struct A2 { auto operator<=>(const A2 &) const = default; };
using Var1 =
std::variant<
A1
, Array
>;
struct Array : public std::vector<Var1>
{
using vector::vector;
};
// This fails
using Var2 =
std::variant<
A2
, Array
>;
static_assert(std::equality_comparable<Var2>);
int main()
{
constexpr bool foo = Var2{} == Var2{};
return 0;
}
```
And here's a [demo on godbolt.com](https://godbolt.org/z/nWfenv9ej).
And here's the link to related [stackoverflow.com question](https://stackoverflow.com/questions/79891776/weird-stdvariantoperator-error-with-clang-libc-and-specific-number-o).
As can be seen, `std::equality_comparable<Var2>` produces true but the actual compilation fails. Problem exists only with **clang** + **libc++** combo. Thanks to **T.C.**'s observation it seems to stem from the fact that, in both variants, `Array` shares the same *alternative index*. This helped to simplify the original test case.
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
