Added the receiver CPOs (set_value, set_error, set_stopped)

[dietmarkuehl]

• fixed the year of the copyright of some of the files
• added missing std:: qualification to a tag_invoke in operation_state

[miscco]

Looks good to me.

The one thing I am missing is converting calls to set_value, which should also be possible and tested

[miscco]

One obvious caveat that I forgot is that we should check that the arguments are recevier's. But that wont fly until we get the receiver concept done so...

[dietmarkuehl]

I’ll add tests checking for conversions later today.

With respect to the receiver concept: although currently it doesn’t (which I assume is a mistake) the receiver concept would depend on the completion signal CPOs (set_value, set_error, set_stopped) as well on other CPOs (eg. get_env). Making the CPOs check for receiver arguments would create a similar cyclic dependency as would be between operation_state and the start CPO. As over there, I think that would be misguided. The CPOs conceptually don’t have anything to do with receiver except that receiver uses them. I do know how that could be implemented but it still is wrong and I think it would be especially wrong in specification.

[miscco]

If we do not go with receiver should we then constraint with movable-value?

[dietmarkuehl]

Maybe - that would need to go into the specification, though. The current specification just says "for some subexpression R" without adding any requirements on R. I believe the implementation allowing (and testing) the different lvalue overloads is as specified. An implementation not passing these tests does not match the specification.