Tracking Issue for the Freeze trait

[oli-obk]

Feature gate: #![feature(freeze)]

This is a tracking issue for the core::marker::Freeze trait

This trait allows generic code to restrict generic parameters to types without interior mutability.

Public API

use std::marker::Freeze;

trait Trait<T: Freeze + 'static> {
    const VALUE: T;
    const VALUE_REF: &'static T = &Self::VALUE;
}

Steps / History

• Implementation: Expose the Freeze trait again (unstably) and forbid implementing it manually #121840
• Final comment period (FCP)¹
• Stabilization PR

Unresolved Questions

• naming the trait differently (see #121501 (comment))

Footnotes

1. https://std-dev-guide.rust-lang.org/feature-lifecycle/stabilization.html ↩

[RalfJung]

@rust-lang/lang -- quick vibe check, do you think this needs an RFC or can it be done with a writeup and FCP in this issue?

[nikomatsakis]

Clarification Ralf: what is "this"? Stabilizing Freeze? I'm inclined to say RFC. Adding a new auto trait feels like a Big Deal and I'd like a strong write up on the justification and expected usage (that said, I believe this is a good move and there are legit reasons for this)

…

On Thu, May 2, 2024, at 7:40 PM, Ralf Jung wrote: @rust-lang/lang <https://github.com/orgs/rust-lang/teams/lang> -- quick vibe check, do you think this needs an RFC or can it be done with a writeup and FCP in this issue? — Reply to this email directly, view it on GitHub <#121675 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABF4ZUTIXQGZWXW6K3WMFDZAJ3AHAVCNFSM6AAAAABD4CORHCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAOJRGE2DMNZVG4>. You are receiving this because you are on a team that was mentioned.Message ID: ***@***.***>

[RalfJung]

"This" = stabilizing "Freeze" bounds (but not impls)

[nikomatsakis]

OK. Then yes, I think it requires an RFC.

[RalfJung]

Okay, thanks.

So -- if someone wants to help push this towards stabilization, drafting an RFC would be a good next step. :)

[p-avital]

Here is a stabilization RFC for this feature: rust-lang/rfcs#3633 :)

[jdonszelmann]

I found an interesting case in which I needed Freeze. Currently, the diagnostics for this are very confusing (I might follow with a PR that improves that) but @WaffleLapkin said it might be nice to comment this here.

not-working example: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=9939beef497fc78bf2951342b379cf4b

fixed using freeze: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=1efad00dbd416d99abe2da4c9758e2a1

[RalfJung]

Yeah that's basically the same case as @p-avital's

[Kyykek]

What are the soundness implications of writing a value of generic type into a MaybeUninit buffer, promoting that (with sufficient alignment) and casting the reference back to the original type? This should be usable today as a workaround for types that have a written guarantee for their lack of interior mutability, though it might be nice if there were more guarantees like this in the standard library.