Fix incorrect impl JoinSemiLattice for State by tmiasko · Pull Request #164081 · rust-lang/rust

tmiasko · 2026-10-10T11:57:32Z

The implementation had an incorrect short-circruting behavior, i.e., it failed to join borrow when qualif changed. This lead implementation to accept code that should have been rejected.

This fixes a part of #164194.

Fix the issue and fallout in the standard library and tests.

rustbot · 2026-10-10T11:57:35Z

Some changes occurred to constck

cc @fee1-dead

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

rustbot · 2026-10-10T11:57:37Z

r? @petrochenkov

rustbot has assigned @petrochenkov.
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

Why was this reviewer chosen?

The reviewer was selected based on:

Owners of files modified in this PR: compiler
compiler expanded to 77 candidates
Random selection from 18 candidates

RalfJung · 2026-10-10T12:06:55Z

Thanks!

This is a breaking change, right? We should crater it to see whether we can just land it as-is or whether we should try to make the analysis smarter to reduce the fallout.
@bors try

Fix incorrect impl JoinSemiLattice for State

RalfJung · 2026-10-10T12:14:49Z

+        let mut v = ManuallyDrop::new(Vec::with_capacity(6));
        let mut x = 1;
        while x < 42 {
            v.push(x);
            x *= 2;
        }
        assert!(v.len() == 6);
-        v.const_make_global()
+        ManuallyDrop::into_inner(v).const_make_global()


Uh, that's not very nice. At the very least this needs FIXME(const-hack). But I am not sure we should land this at all. Does it help if we enable const_precise_live_drops?

View changes since the review

Yes - assuming it is fine if we enable if for the whole test crate.

IMO that's fine, yeah.

rustbot · 2026-10-10T12:51:41Z

⚠️ #[rustc_allow_const_fn_unstable] needs careful audit to avoid accidentally exposing unstable
implementation details on stable.

cc @rust-lang/wg-const-eval

rust-bors · 2026-10-10T13:12:07Z

☀️ Try build successful (CI)
Build commit: c79f6bc (c79f6bc4fc2389852aeff667626cb3e58adfe71c)
Base parent: 34f4a80 (34f4a807e2d84184fff4e42d4539092569b6fcd6)

RalfJung · 2026-10-10T13:13:08Z

@craterbot check

RalfJung · 2026-10-10T13:38:09Z

@craterbot check
is the bot alive?

craterbot · 2026-10-10T13:38:12Z

🚨 Error: missing start toolchain

🆘 If you have any trouble with Crater please ask in t-infra on Zulip
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

RalfJung · 2026-10-10T17:14:07Z

@craterbot run mode=check-only start=34f4a807e2d84184fff4e42d4539092569b6fcd6 end=c79f6bc4fc2389852aeff667626cb3e58adfe71c

craterbot · 2026-10-10T17:14:20Z

👌 Experiment pr-164081 created and queued.
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

theemathas · 2026-10-11T12:17:06Z

Crater rollup in #164148

@craterbot abort

theemathas · 2026-10-11T12:18:32Z

@craterbot abort

craterbot · 2026-10-11T12:18:36Z

🗑️ Experiment pr-164081 deleted!

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

petrochenkov · 2026-10-12T12:39:57Z

r? @RalfJung

rustbot · 2026-10-12T12:40:13Z

RalfJung is not on the review rotation at the moment.
They may take a while to respond.

The implementation had an incorrect short-circruting behavior, i.e., it failed to join `borrow` when `qualif` changed. This lead implementation to accept code that should have been rejected. Fix the issue and fallout in the standard library and tests.

theemathas · 2026-10-18T12:47:18Z

@craterbot check crates=https://crater-reports.s3.amazonaws.com/pr-164148-crater-rollup/retry-regressed-list.txt p=1

craterbot · 2026-10-18T12:47:21Z

🚨 Error: missing start toolchain

🆘 If you have any trouble with Crater please ask in t-infra on Zulip
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

theemathas · 2026-10-18T12:47:52Z

@craterbot check crates=https://crater-reports.s3.amazonaws.com/pr-164148-crater-rollup/retry-regressed-list.txt p=1 start=34f4a807e2d84184fff4e42d4539092569b6fcd6 end=c79f6bc4fc2389852aeff667626cb3e58adfe71c

craterbot · 2026-10-18T12:47:56Z

👌 Experiment pr-164081 created and queued.
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

theemathas · 2026-10-18T14:50:50Z

I suspect that this PR broke the latest version of generic-array. For example: https://crater-reports.s3.amazonaws.com/pr-164148-crater-rollup/c60b92c03ca8a3b101eb9d57f51e2232690638c7/reg/pep_engine_sequoia_backend-1.1.0/log.txt

Edit: And also the latest version of konst. For example: https://crater-reports.s3.amazonaws.com/pr-164148-crater-rollup/c60b92c03ca8a3b101eb9d57f51e2232690638c7/reg/matrix-sdk-test-0.19.1/log.txt

Edit: And also the latest version of pest. For example: https://crater-reports.s3.amazonaws.com/pr-164148-crater-rollup/c60b92c03ca8a3b101eb9d57f51e2232690638c7/reg/matrix-sdk-test-0.19.1/log.txt

tmiasko · 2026-10-18T17:55:47Z

In rollup #164148 (comment) we have 119 regressed root results with error E0493: "A value with a custom Drop implementation may be dropped during const-eval".

RalfJung · 2026-10-18T18:19:30Z

Sounds like it may be worth exploring a smarter analysis that's sound (unlike the current one) but can accept more code? We should look at some of the failing examples to judge how feasible it would be to keep accepting them.

craterbot · 2026-10-19T22:49:23Z

🚧 Experiment pr-164081 is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

craterbot · 2026-10-20T13:11:38Z

🎉 Experiment pr-164081 is completed!
📊 4582 regressed and 0 fixed (28882 total)
📊 938 spurious results on the retry-regressed-list.txt, consider a retry¹ if this is a significant amount.
📰 Open the summary report.

⚠️ If you notice any spurious failure please add them to the denylist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

re-run the experiment with crates=https://crater-reports.s3.amazonaws.com/pr-164081/retry-regressed-list.txt ↩

tmiasko · 2026-10-20T17:04:39Z

Committing to MIR move elimination semantics and detecting undefined behavior associated with deallocation on a bare local moves seems most realistic approach to fixing this. I think we should do this regardless, so there seems little point to having this kind of breakage in the meantime.

RalfJung · 2026-10-20T20:06:24Z

I think the vast majority of these can be handled with our current MIR semantics. We just need to

clear the qualif on move from the local
when a local gets initialized that was previously borrowed, skip the value and do only type-based reasoning

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. T-libs Relevant to the library team, which will review and decide on the PR/issue. labels Oct 10, 2026

rustbot assigned petrochenkov Oct 10, 2026

tmiasko mentioned this pull request Oct 10, 2026

Improve live drop precision #161627

Closed