P1065R2 constexpr INVOKE about stl

[BillyONeal]

https://gcc.godbolt.org/z/ro-Wqf <-- all the examples in the paper indicating that we needed CWG1581 fixed look to work in all the compilers now, so I think we can go ahead with this now.

[kasperk81]

https://gcc.godbolt.org/z/ro-Wqf <-- all the examples in the paper indicating that we needed CWG1581 fixed look to work in all the compilers now, so I think we can go ahead with this now.

Not sure if conformance table is up to date, but was CWG 1581 also resolved in 16.8?

[BillyONeal]

Not sure if conformance table is up to date, but was CWG 1581 also resolved in 16.8?

I didn't try to figure out if it was fixed completely, I just made sure the examples called out as problematic there in the paper worked.

Notably, our std::invoke implementation doesn't use SFINAE to select the mechanism to use, only <type_traits>, so it's possible that we were always immune to the 1581 edge cases.

[StephanTLavavej]

When we talked to the compiler devs in Dec 2020, we learned that while CWG-1581 isn't completely implemented (so the conformance table is correct), it's sufficient for the STL's purposes in std::invoke as far as we can tell.