Implement phantom variance markers

[jhpratt]

ACP accepted rust-lang/libs-team#488

Tracking issue #135806

[rustbot]

r? @scottmcm

rustbot has assigned @scottmcm.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[Sky9x]

Correct me if I'm wrong, but I think these should be the other way around

[jhpratt]

Swapped the descriptions so as to be correct, reworded them a bit for clarity, and added an example (contrived as it may be).

[scottmcm]

I never remember which variance is which, so hopefully a random roll will find someone who does 🤞

r? libs

[Amanieu]

In the ACP, we suggested making these types work like PhantomData by allowing them to be constructed without any fields. This doesn't have to happen in this PR, but it would eliminate the need for the trait and constructor method, and would bring these types in line with PhantomData and PhantomPinned. At the very least it should be listed as a blocker for stabilization in the tracking issue.

[jhpratt]

Why should that be a blocker? It's backwards compatible to eliminate the need for ::new() by turning it into a lang item. The trait only exists so people don't have to explicit state the variance when constructing a value, regardless of whether ::new() is needed.

[Amanieu]

While it is backwards-compatible, we are generally reluctant to stabilize functions that we known are going to be deprecated eventually.

In any case, this PR is fine for an initial implementation.

[jhpratt]

Gotcha. I'll add that as an unresolved question in the tracking issue.

[Amanieu]

@bors r+

[bors]

📌 Commit b3a5d0a has been approved by Amanieu

It is now in the queue for this repository.