std::complex reflect and __cpp_lib_tuple_like 202311L aka P2819R2

[piotrrak]

Hello

Just FYI, since I looked at build issues with recent clang/libc++

The check if std::complex is tuple-like is something that popped up.

This test for libc++ is bit unfortunate since it implements tuple-like protocol for std::complex
P2819R2 but not all the other things tuple-like protocol things that is are dependency (__cpp_lib_tuple_like 202207L).
Authors of paper did choose update __cpp_lib_tuple_like and libc++ can't claim it since it doesn't have things required for 202207L.
libc++ chooses to implement it, implying that only version check of libc++ can fix it.

It'll also fail with g++, didn't test it but there's libstdc++: implement tuple protocol for std::complex (P2819R2)

I don't have opinion if it should be removed or '#ifdef'ed', but since it's something that will come to standard, maybe there is no point in testing it at all.

Cheers,
/P

[mattkretz]

Thanks for the heads-up. IIRC, my intent here was to make sure that simdize<std::complex<T>> doesn't work. Because it should work differently - with additional semantics.

Also, IIRC, this'll break GNU Radio's use of vir-simd. sigh