Implement alloc::sync::UniqueArc

[frank-king]

This implements the alloc::sync::UniqueArc part of #112566.

[rustbot]

r? @Amanieu

rustbot has assigned @Amanieu.
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

[steffahn]

Hah! I’m just noticing that this, as well as UniqueRc, absolutely must be invariant (w.r.t. the parameter T).

Otherwise, you can

fn extend_lifetime<'a, 'b>(x: &'a str) -> &'b str {
    let r = UniqueRc::new(""); // UniqueRc<&'static str>
    let w = UniqueRc::downgrade(&r); // Weak<&'static str>
    let mut r = r; // COVARIANT: ==>> UniqueRc<&'a str>
    *r = x; // assign the &'a str
    let _r = UniqueRc::into_rc(r); // Rc<&'a str>, but we only care to activate the weak
    let r = w.upgrade().unwrap(); // upgrade succeeds: Rc<&'static str>
    *r // &'static str, coerces to &'b str
}

(playground)

[bors]

☔ The latest upstream changes (presumably #135357) made this pull request unmergeable. Please resolve the merge conflicts.

[Sky9x]

Suggested change

	}
	#[unstable(feature = "unique_rc_arc", issue = "112566")]
	unsafe impl<#[may_dangle] T: ?Sized, A: Allocator> Drop for UniqueArc<T, A> {
	}
	#[unstable(feature = "deref_pure_trait", issue = "87121")]
	unsafe impl<T: ?Sized, A: Allocator> DerefPure for Arc<T, A> {}
	#[unstable(feature = "unique_rc_arc", issue = "112566")]
	unsafe impl<#[may_dangle] T: ?Sized, A: Allocator> Drop for UniqueArc<T, A> {

[frank-king]

Do you mean impl DerefPure for UniqueArc here? (is Arc here a typo?)

[Sky9x]

yes it should be UniqueArc that's my bad

[Sky9x]

Suggested change

	/// Inequality for two `UniqueArc`s.
	///
	/// Two `UniqueArc`s are not equal if their inner values are not equal.
	///
	/// # Examples
	///
	/// ```
	/// #![feature(unique_rc_arc)]
	/// use std::sync::UniqueArc;
	///
	/// let five = UniqueArc::new(5);
	///
	/// assert!(five != UniqueArc::new(6));
	/// ```
	#[inline]
	fn ne(&self, other: &Self) -> bool {
	PartialEq::ne(&**self, &**other)
	}

Implementations must ensure that eq and ne are consistent with each other:

• a != b if and only if !(a == b).

The default implementation of ne provides this consistency and is almost always sufficient. It should not be overridden without very good reason.

https://doc.rust-lang.org/std/cmp/trait.PartialEq.html

[frank-king]

But how about the user overrides <T as PartialEq>::ne? Then the behaviours of <T as PartialEq>::ne and <Arc<T> as PartialEq>::ne are inconsistent.

[Sky9x]

See #98655 for some context, but the only possible correct implementation of ne is !eq, so there is no reason to override it.

[frank-king]

I see, thank you. Should I remove the overidden implementation of <Rc as PartialEq>::ne too?

[bors]

☔ The latest upstream changes (presumably #138208) made this pull request unmergeable. Please resolve the merge conflicts.