Implement CoerceUnsized for {Cell, RefCell, UnsafeCell}

[apasel422]

These impls are analogous to the one for NonZero. It's occasionally useful to be able to coerce the cell types when they're being used inside another abstraction. See Manishearth/rust-gc#17 for an example.

r? @eddyb

[eddyb]

LGTM. cc @rust-lang/lang Do we want such impls? While the feature is unstable, they would be usable in coercions from stable code.

[alexcrichton]

Would it be possible to add a few tests to the repo as well ensuring that we don't regress this functionality?

[apasel422]

@alexcrichton Done.

[alexcrichton]

Thanks! I think though that libcore doesn't actually have tests run, so could they be moved to libcoretest?

[apasel422]

@alexcrichton They're not actually tests -- they'll cause compilation to fail if we regress the functionality. This is similar to other assertions that types are covariant or Send/Sync.

[alexcrichton]

Oh gah, right! Carry on!

[nrc]

Usually, I'd be cautious about impls that let us skirt stability, but the DST coercions have been around for ages and I don't think there are problems with them, but we could do with more testing, so maybe this would help that?

[bors]

☔ The latest upstream changes (presumably #35666) made this pull request unmergeable. Please resolve the merge conflicts.

[nikomatsakis]

I don't know that these impls make sense for Cell -- cell relies on being able to copy memory in and out, and coercing to an unsized value means we can't do that, right?

RefCell and UnsafeCell make more sense to me and I think I am in favor.

[nikomatsakis]

I guess you could copy things in and out if you permitted dynamically sized stack frames (as I sort of want to permit now, in order to generally loosen the rules on DST). But no reason to jump the gun there. :)

[apasel422]

@nikomatsakis See the link in my original post for an example of where this is useful for Cell. This impl means that if you have Cell<Shared<T>> where T: Send, you can coerce it to Cell<Shared<Send>>.

[eddyb]

@nikomatsakisCell<Unsized> does indeed not make sense - but that's not what's being discussed here - that works automatically as long as you have T: ?Sized, which RefCell and UnsafeCell already do (RefCell<Trait> has been working for many versions now).

What's being added here is the ability to coerce {Unsafe,Ref,}Cell<Ptr<T>> to {Unsafe,Ref,}Cell<Ptr<Unsized>> where T: Unsize<Unsized>.

More concretely, Cell<&T> would coerce to Cell<&Trait> and that is useful.

[nikomatsakis]

Ah, right, thanks for the correction. =)

In that case, I have no objection. It is sort of nifty that this works. Seems like it'd be a variance error, but since there's no aliasing...

[alexcrichton]

Discussed during @rust-lang/libs triage discussion was also that this is good to merge, thanks @apasel422!

@bors: r+

[bors]

📌 Commit 1fd791a has been approved by alexcrichton

[bors]

⌛ Testing commit 1fd791a with merge 43204ff...

[bors]

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-cargotest, auto-linux-64-cross-freebsd, auto-linux-64-debug-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-opt-no-mir, auto-linux-64-opt-rustbuild, auto-linux-64-x-android-t, auto-linux-cross-opt, auto-linux-musl-64-opt, auto-mac-32-opt, auto-mac-64-opt, auto-mac-64-opt-rustbuild, auto-mac-cross-ios-opt, auto-win-gnu-32-opt, auto-win-gnu-32-opt-rustbuild, auto-win-gnu-64-opt, auto-win-msvc-32-opt, auto-win-msvc-64-cargotest, auto-win-msvc-64-opt, auto-win-msvc-64-opt-no-mir, auto-win-msvc-64-opt-rustbuild
Approved by: alexcrichton
Pushing 43204ff to master...