make RefCell unstably const #137843

Daniel-Aaron-Bloom · 2025-03-01T06:57:47Z

Now that we can do interior mutability in const, most of the RefCell API can be const fn. The main exceptions are APIs which use FnOnce (RefCell::replace_with and Ref[Mut]::[filter_]map[_split]) and RefCell::take which calls Default::default.

Tracking issue: #137844

rustbot · 2025-03-01T06:57:51Z

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

Daniel-Aaron-Bloom · 2025-03-01T07:29:24Z

@rustbot label -T-libs +T-libs-api
r? libs-api

tgross35 · 2025-03-02T12:19:36Z

Cc @rust-lang/wg-const-eval

RalfJung · 2025-03-03T07:26:40Z

The main problem is dropping. Please make sure to also add a test actually using a RefCell; that will probably not work since the Ref/RefMut Drop impls are not const.

RalfJung · 2025-03-03T07:28:03Z

library/core/src/cell.rs

+        debug_assert!(is_writing(borrow));
+        self.borrow.replace(borrow + 1);
+        mem::forget(self)
+    }


What is even going on here? Seems like you added some hacks related to dropping, but there are no comments at all explaining them. Please add extensive comments any time you do anything unusual.

Apologies. I've reverted this now in favor of const Drop and added a comment to my clone helper (since there's no const Clone trait)

RalfJung · 2025-03-03T07:32:51Z

@compiler-errors what is the state of Drop in const? Can one impl const Drop and it will "just work"? Are const Destruct bounds ready for use in std (for APIs that are stable but const-unstable)?

traviscross · 2025-03-03T08:02:48Z

Since we probably all want to hear the answer to that question...

cc @rust-lang/lang

oli-obk · 2025-03-03T08:24:07Z

Are const Destruct bounds ready for use in std (for APIs that are stable but const-unstable)?

No. Until we know what const traits syntax to implement, adding more const trait impls is not something we want. The libstd churn is not worth it

oli-obk · 2025-03-03T08:24:37Z

Can one impl const Drop and it will "just work"?

Yes

RalfJung · 2025-03-03T09:49:41Z

@Daniel-Aaron-Bloom okay so please remove the fn const_drop hack and instead try impl const Drop for BorrowRef and BorrowRefMut. If that does not work, then I think it is better to wait until that works; nobody will be able to use RefCell in const otherwise anyway (not even on nightly).

Daniel-Aaron-Bloom · 2025-03-03T18:26:08Z

Should I also remove try to remove my added private function RefMut::as_mut and use const_deref instead?

RalfJung · 2025-03-03T18:29:12Z

Oh that's what those functions are for?

Sure, seems fine to me as long no incomplete nightly features have to be enabled for this.

RalfJung · 2025-03-03T18:49:41Z

tests/ui/consts/const-refcell-destruct.rs

Typically we prefer library tests over ui tests for things like this. Somewhere in library/coretests there should be a good place.

Thanks. Will do.

RalfJung · 2025-03-03T18:50:20Z

tests/ui/consts/const-refcell-destruct.rs

+
+fn main() {
+    let dummy = REF_CELL_TEST;
+    assert_eq!(DROP_COUNT.load(Ordering::Relaxed), 0);


What is the point of this?

Honestly it was just a dumb way to ensure the destructor is never called. But now I realize I could have just thrown a panic in the drop impl.

RalfJung · 2025-03-03T18:51:09Z

tests/ui/consts/const-refcell.rs

+    }
+}
+
+const REF_CELL_TEST: RefCell<Dummy> = {


Please also try to actually change the value in the RefCell, and then observe the change in a following borrow().

library/core/src/cell.rs

RalfJung · 2025-03-03T22:32:58Z

library/core/src/lib.rs

@@ -154,6 +154,7 @@
 #![feature(cfg_target_has_atomic)]
 #![feature(cfg_target_has_atomic_equal_alignment)]
 #![feature(cfg_ub_checks)]
+#![feature(const_destruct)]


@compiler-errors @lcnr @fee1-dead is this feature okay to use in libcore?

Not until the syntax is finalized by T-lang. I've marked this PR as blocked

Which syntax? There is no ~const in this PR.

All this feature gate does is

allow impl const Drop -- but we already have a few impl const in the standard library so that syntax can't be the concern (I'm not even sure if the feature gate is truly required for this)

adjust the const_check logic to allow drop in const if the type in question is const-destructible

impl const Trait for Type syntax is also not clear anymore that it will get accepted. So we're putting everything on hold and not adding more things we may have to revert

Should I go back to the private const_drop and as_mut helpers then?

I think we'll just hold off on merging. I'm optimistic we'll get a decision soon

library/coretests/tests/cell.rs

Co-authored-by: Ralf Jung <post@ralfj.de>

rustbot assigned Amanieu Mar 1, 2025

rustbot added S-waiting-on-review T-libs labels Mar 1, 2025

Daniel-Aaron-Bloom mentioned this pull request Mar 1, 2025

Tracking Issue for const_ref_cell #137844

Open

3 tasks

Daniel-Aaron-Bloom force-pushed the const_cell branch from ef5a67e to ddc8777 Compare March 1, 2025 07:19

Daniel-Aaron-Bloom changed the title ~~Draft: make Cell more unstably const~~ make RefCell more unstably const Mar 1, 2025

Daniel-Aaron-Bloom marked this pull request as ready for review March 1, 2025 07:29

rustbot added T-libs-api and removed T-libs labels Mar 1, 2025

rustbot assigned joshtriplett and unassigned Amanieu Mar 1, 2025

Daniel-Aaron-Bloom changed the title ~~make RefCell more unstably const~~ make RefCell unstably const Mar 1, 2025

Daniel-Aaron-Bloom changed the title ~~make RefCell unstably const~~ make RefCell unstably const Mar 1, 2025

Daniel-Aaron-Bloom changed the title ~~make RefCell unstably const~~ make RefCell unstably const Mar 1, 2025

RalfJung reviewed Mar 3, 2025

View reviewed changes

library/core/src/cell.rs Outdated Show resolved Hide resolved

RalfJung reviewed Mar 3, 2025

View reviewed changes

library/coretests/tests/cell.rs Show resolved Hide resolved

oli-obk added the S-blocked label Mar 4, 2025

RalfJung reviewed Mar 4, 2025

View reviewed changes

library/coretests/tests/cell.rs Outdated Show resolved Hide resolved

rustbot added has-merge-commits S-waiting-on-author labels Mar 27, 2025

This comment has been minimized.

Sign in to view

Daniel-Aaron-Bloom and others added 7 commits March 27, 2025 16:03

make RefCell unstably const

2ee7f02

remove const_drop and add tests

5282939

refactor tests and helpers as requested

4e26f96

Update library/core/src/cell.rs

c6fb9c7

Co-authored-by: Ralf Jung <post@ralfj.de>

add borrow tests

fa9ba3e

Update library/coretests/tests/cell.rs

0fdc84a

Co-authored-by: Ralf Jung <post@ralfj.de>

test const Ref[Mut]

484a39a

Daniel-Aaron-Bloom force-pushed the const_cell branch from 06262da to 484a39a Compare March 27, 2025 23:03

rustbot removed has-merge-commits S-waiting-on-author labels Mar 27, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

make RefCell unstably const #137843

make RefCell unstably const #137843

Daniel-Aaron-Bloom commented Mar 1, 2025 •

edited

Loading

rustbot commented Mar 1, 2025

Daniel-Aaron-Bloom commented Mar 1, 2025

tgross35 commented Mar 2, 2025

RalfJung commented Mar 3, 2025

RalfJung Mar 3, 2025 •

edited

Loading

Daniel-Aaron-Bloom Mar 3, 2025

RalfJung commented Mar 3, 2025

traviscross commented Mar 3, 2025

oli-obk commented Mar 3, 2025 •

edited

Loading

oli-obk commented Mar 3, 2025

RalfJung commented Mar 3, 2025 •

edited

Loading

Daniel-Aaron-Bloom commented Mar 3, 2025

RalfJung commented Mar 3, 2025

RalfJung Mar 3, 2025

Daniel-Aaron-Bloom Mar 3, 2025

RalfJung Mar 3, 2025

Daniel-Aaron-Bloom Mar 3, 2025

RalfJung Mar 3, 2025

Daniel-Aaron-Bloom Mar 3, 2025

RalfJung Mar 3, 2025

oli-obk Mar 4, 2025

RalfJung Mar 4, 2025

RalfJung Mar 4, 2025 •

edited

Loading

oli-obk Mar 4, 2025 •

edited

Loading

Daniel-Aaron-Bloom Mar 4, 2025

oli-obk Mar 4, 2025

This comment has been minimized.

make RefCell unstably const #137843

Are you sure you want to change the base?

make RefCell unstably const #137843

Conversation

Daniel-Aaron-Bloom commented Mar 1, 2025 • edited Loading

rustbot commented Mar 1, 2025

Daniel-Aaron-Bloom commented Mar 1, 2025

tgross35 commented Mar 2, 2025

RalfJung commented Mar 3, 2025

RalfJung Mar 3, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

RalfJung commented Mar 3, 2025

traviscross commented Mar 3, 2025

oli-obk commented Mar 3, 2025 • edited Loading

oli-obk commented Mar 3, 2025

RalfJung commented Mar 3, 2025 • edited Loading

Daniel-Aaron-Bloom commented Mar 3, 2025

RalfJung commented Mar 3, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

RalfJung Mar 4, 2025 • edited Loading

Choose a reason for hiding this comment

oli-obk Mar 4, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

This comment has been minimized.

Daniel-Aaron-Bloom commented Mar 1, 2025 •

edited

Loading

RalfJung Mar 3, 2025 •

edited

Loading

oli-obk commented Mar 3, 2025 •

edited

Loading

RalfJung commented Mar 3, 2025 •

edited

Loading

RalfJung Mar 4, 2025 •

edited

Loading

oli-obk Mar 4, 2025 •

edited

Loading