Generalize `impl<T> Clone for Box<T>` to unsized types #146381

a1phyr · 2025-09-09T18:19:59Z

I had to add a non-default method to CloneToUninit for this to work, but I think this is fine. (I did not do the renaming to CloneUnsized as suggested in #126799 (comment), as this is quite orthogonal)

This also adds some infrastructure that will be helpful for in-place initialization.

Commits can be reviewed independently.

rustbot · 2025-09-09T18:20:05Z

r? @workingjubilee

rustbot has assigned @workingjubilee.
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

workingjubilee · 2025-09-09T19:42:11Z

r? library

rustbot · 2025-09-09T19:42:14Z

Failed to set assignee to library: invalid assignee

Note: Only org members with at least the repository "read" role, users with write permissions, or people who have commented on the PR may be assigned.

workingjubilee · 2025-09-09T19:42:28Z

r? libs

library/alloc/src/boxed.rs

zachs18 · 2025-09-10T01:56:10Z

library/core/src/clone.rs

+    ///
+    /// # Safety
+    ///
+    /// `dest` must be a valid pointer to a valid value of the same concrete type than `self`.


Given that trait vtable pointers can be deduplicated, I'm not sure if "the same concrete type" is what we want to require here. cc #146381 (comment) . I.e. it's not possible in general to verify for certain that two dyn Trait pointers with the same metadata actually point to the same erased/concrete type.

This seems similar to the issue with Rc::get_mut_unchecked, where either:

if any other pointers exist to this value, we must only write a value that is valid for all those pointers' types/metadatas.

if there are no other pointers to this value, we can write anything that is valid for this pointer's type/metadata.

I think this is the right thing. It be not be possible in general, but it is in some cases (eg with TypeId). Another formulation would be that it is safe to create a mutable reference from dest.with_metadata_of(self).

library/alloc/src/boxed/uninit.rs

jhpratt · 2025-09-12T23:02:12Z

As CI is failing, @rustbot author

rustbot · 2025-09-12T23:02:16Z

Reminder, once the PR becomes ready for a review, use @rustbot ready.

rustbot · 2025-09-25T20:12:12Z

This PR was rebased onto a different master commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

This new module contains a type that is like an uninitialized `Box`, but works with any type, not just `Sized` ones. Additionally, this will be useful for future work on in-place initialization (see rust-lang#142518).

This is necessary to properly support `Clone::clone_from`.

a1phyr · 2025-09-26T20:41:55Z

@rustbot ready

jhpratt · 2025-12-17T04:15:55Z

I forgot I was assigned to this and apparently missed that it was ready for review. I'm busy at the moment, so passing off.

r? libs

Mark-Simulacrum · 2025-12-20T17:05:23Z

library/alloc/src/boxed/uninit.rs

+    #[cfg(not(no_global_oom_handling))]
+    pub(crate) unsafe fn new_for_metadata_in(meta: T::Metadata, alloc: A) -> Self {
+        let ptr = ptr::from_raw_parts_mut::<T>(ptr::null_mut::<()>(), meta);
+        let layout = unsafe { Layout::for_value_raw(ptr) };


What is the safety justification here? AFAICT, this can hit the last case ("otherwise, it is conservatively not allowed to call this function." in https://doc.rust-lang.org/nightly/std/alloc/struct.Layout.html#method.for_value_raw).

We should have safety comments on the unsafe blocks and new_for_metadata_in itself should have a doc comment with requirements to be called.

Mark-Simulacrum · 2025-12-20T17:06:02Z

library/alloc/src/boxed/uninit.rs

+
+        let ptr = if layout.size() == 0 {
+            let dangling = unsafe { core::num::NonZero::new_unchecked(layout.align()) };
+            NonNull::without_provenance(dangling)


nit: use https://doc.rust-lang.org/beta/std/alloc/struct.Layout.html#method.dangling ?

Mark-Simulacrum · 2025-12-20T17:13:19Z

library/alloc/src/ffi/c_str.rs

 }

-#[stable(feature = "more_box_slice_clone", since = "1.29.0")]
-impl Clone for Box<CStr> {


Do we have a unit test or compile test verifying this still exists? Want to avoid accidentally losing it if the unstable feature is removed without noticing that we've shrunk the API surface area.

Mark-Simulacrum · 2025-12-20T17:21:36Z

library/alloc/src/boxed.rs

+            // if the concrete types are different, so we never allow them to
+            // go through `clone_from`.
+            #[inline]
+            default fn meta_eq(self, _: Self) -> bool {


Hm, I'm not very happy with using specialization for this. At minimum I'd like to get T-types to weigh in before we merge. It seems like if we replaced this with an unconditional 'false', we'd only regress performance for callers of clone_from with slices?

Maybe the right thing is to add MaybeComparable or some such to a bound of Pointee::Metadata and force impls for all things that might be metadata? AFAICT, that would avoid specialization, right?

Mark-Simulacrum · 2025-12-20T17:22:46Z

tests/ui/box/unit/unique-pinned-nocopy.stderr

   |
 LL | struct R {
-   | -------- doesn't satisfy `R: Clone`
+   | -------- doesn't satisfy `R: CloneToUninit`


This seems like an unfortunate regression to probably relatively common diagnostics... especially the lack of the derive suggestion seems not-great. Is there some do_not_recommend attribute or tweak to the compiler we can land that would avoid this regression? CloneToUninit doesn't seem like it should be surfaced to users, at least until it is stabilized, and even then only if you're dealing with something ?Sized I suspect.

rustbot assigned workingjubilee Sep 9, 2025

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-libs Relevant to the library team, which will review and decide on the PR/issue. labels Sep 9, 2025

This comment has been minimized.

Sign in to view

rustbot assigned jhpratt and unassigned workingjubilee Sep 9, 2025

This comment has been minimized.

Sign in to view

zachs18 reviewed Sep 10, 2025

View reviewed changes

library/alloc/src/boxed.rs Outdated Show resolved Hide resolved

zachs18 reviewed Sep 10, 2025

View reviewed changes

library/alloc/src/boxed.rs Outdated Show resolved Hide resolved

zachs18 reviewed Sep 10, 2025

View reviewed changes

theemathas reviewed Sep 10, 2025

View reviewed changes

library/alloc/src/boxed/uninit.rs Show resolved Hide resolved

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Sep 12, 2025

a1phyr force-pushed the box_clone_more branch from f32cd66 to bdc70ea Compare September 17, 2025 14:15

This comment has been minimized.

Sign in to view

a1phyr force-pushed the box_clone_more branch from bdc70ea to 32deea5 Compare September 17, 2025 14:32

This comment has been minimized.

Sign in to view

a1phyr force-pushed the box_clone_more branch from 32deea5 to 6f53c47 Compare September 25, 2025 20:12

This comment has been minimized.

Sign in to view

a1phyr force-pushed the box_clone_more branch from 6f53c47 to 9cc7f3f Compare September 25, 2025 21:14

This comment has been minimized.

Sign in to view

Add alloc::boxed::uninit

06a9fc6

This new module contains a type that is like an uninitialized `Box`, but works with any type, not just `Sized` ones. Additionally, this will be useful for future work on in-place initialization (see rust-lang#142518).

Add CloneToUninit::clone_to_init

3784678

This is necessary to properly support `Clone::clone_from`.

a1phyr force-pushed the box_clone_more branch from 9cc7f3f to 84c014a Compare September 25, 2025 23:30

a1phyr added 3 commits September 26, 2025 01:32

Generalize Box::clone

129ff6f

Update CloneToUninit example to use Box

6da1cdf

Update ui tests

9331815

a1phyr force-pushed the box_clone_more branch from 84c014a to 9331815 Compare September 25, 2025 23:32

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Sep 26, 2025

a1phyr mentioned this pull request Oct 19, 2025

[lib] In-place initialization infrastructure #142518

Open

rustbot assigned Mark-Simulacrum and unassigned jhpratt Dec 17, 2025

Mark-Simulacrum requested changes Dec 20, 2025

View reviewed changes

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Dec 20, 2025

Uh oh!

Generalize impl<T> Clone for Box<T> to unsized types #146381

Are you sure you want to change the base?

Generalize impl<T> Clone for Box<T> to unsized types #146381

Uh oh!

Conversation

a1phyr commented Sep 9, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Sep 9, 2025

Uh oh!

This comment has been minimized.

workingjubilee commented Sep 9, 2025

Uh oh!

rustbot commented Sep 9, 2025

Uh oh!

workingjubilee commented Sep 9, 2025

Uh oh!

This comment has been minimized.

Uh oh!

Uh oh!

zachs18 Sep 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zachs18 Sep 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

a1phyr Sep 26, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jhpratt commented Sep 12, 2025

Uh oh!

rustbot commented Sep 12, 2025

Uh oh!

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

rustbot commented Sep 25, 2025

Uh oh!

This comment has been minimized.

This comment has been minimized.

a1phyr commented Sep 26, 2025

Uh oh!

jhpratt commented Dec 17, 2025

Uh oh!

Mark-Simulacrum Dec 20, 2025

Choose a reason for hiding this comment

Uh oh!

Mark-Simulacrum Dec 20, 2025

Choose a reason for hiding this comment

Uh oh!

Mark-Simulacrum Dec 20, 2025

Choose a reason for hiding this comment

Uh oh!

Mark-Simulacrum Dec 20, 2025

Choose a reason for hiding this comment

Uh oh!

Mark-Simulacrum Dec 20, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

Generalize `impl<T> Clone for Box<T>` to unsized types #146381

Generalize `impl<T> Clone for Box<T>` to unsized types #146381

a1phyr commented Sep 9, 2025 •

edited

Loading

zachs18 Sep 10, 2025 •

edited

Loading

zachs18 Sep 10, 2025 •

edited

Loading