LWG-4113 Disallow `has_unique_object_representations<Incomplete[]>`

[StephanTLavavej]

LWG-4113 Disallow has_unique_object_representations<Incomplete[]>

We need to report this to:

• C1XX
• Clang
• EDG

so they can reject this code:

C:\Temp>type meow.cpp

#include <print>
#include <type_traits>
using namespace std;

struct Incomplete;

int main() {
    println("{}", has_unique_object_representations_v<Incomplete[]>);
}

C:\Temp>cl /EHsc /nologo /W4 /std:c++latest /MTd /Od meow.cpp && meow
meow.cpp
false

C:\Temp>clang-cl /EHsc /nologo /W4 /std:c++latest /MTd /Od meow.cpp && meow
false

C:\Temp>cl /EHsc /nologo /W4 /std:c++latest /MTd /Od /c /BE meow.cpp
meow.cpp

C:\Temp>

[AlexGuteniev]

Do you not want library workaround until compilers have it?

[CaseyCarter]

Do you not want library workaround until compilers have it?

It's probably not worth our trouble to devise, implement, and review a workaround for such a weird corner case. I could be convinced otherwise if someone could exhibit a problem in an actual program caused by this issue, but for now I think we should focus on other priorities.

[frederick-vs-ja]

Clang 19 w/ MSVC STL already rejects has_unique_object_representations<Incomplete[]> due to LLVM-95432.

libc++ has been always rejecting this. It formerly used __has_unique_object_representations(remove_all_extents_t<_Tp>) as a workaround, and the workaround was no longer necessary and removed in LLVM-143854. Such workaround works for MSVC and EDG.

[frederick-vs-ja]

Reported DevCom-11008458 (MSVC) and DevCom-11008471 (EDG).