`-Znext-solver` Ignore region constraints from the nested goals in leakcheck by ShoyuVanilla · Pull Request #155749 · rust-lang/rust

ShoyuVanilla · 2026-04-24T16:56:57Z

Fixes rust-lang/trait-system-refactor-initiative#251
Fixes #140577
Fixes #153596

Canonicalize free regions from inputs as placeholders in root univ Context: The box named *coroutine witness Send lifetime requirements now considered by leakcheck* [this roadmap](https://raw.githubusercontent.com/hexcatnl/roadmap/e380fef94b47c02a056b4c8f05124a9db475b990/next-solver.svg) Fixes (only for the next-solver) #106569 Prerequisite of #155749

ShoyuVanilla · 2026-04-30T03:23:16Z

r? lcnr

rustbot · 2026-04-30T03:23:22Z

This PR changes a file inside tests/crashes. If a crash was fixed, please move into the corresponding ui subdir and add 'Fixes #' to the PR description to autoclose the issue upon merge.

Some changes occurred to the core trait solver

cc @rust-lang/initiative-trait-system-refactor

changes to the core type system

cc @lcnr

lcnr · 2026-04-30T07:17:46Z

-                            .for_each(|c| each_edge(c.sub, c.sup));
+                        region_constraints.data().constraints[i].0.iter_outlives().for_each(
+                            |Constraint { kind: _, sub, sup, visible_for_leak_check }| {
+                                each_edge(sub, sup, visible_for_leak_check)


maybe don't even provide any constraints that are not visible to leak hceck to each_edge, feels cleaner to me,
i.e.

Suggested change

each_edge(sub, sup, visible_for_leak_check)

match visible_for_leak_check {

ty::VisibleForLeakCheck::Yes =>

each_edge(sub, sup),

ty::VisibleForLeakCheck::No => {}

}

View changes since the review

lcnr · 2026-04-30T07:19:43Z

+pub type QueryRegionConstraint<'tcx> =
+    (ty::RegionConstraint<'tcx>, ConstraintCategory<'tcx>, ty::VisibleForLeakCheck);


not in this PR, but mind adding a FIXME: Convert this into a struct

View changes since the review

lcnr · 2026-04-30T07:20:44Z

+        // We ignore constraints from the nested goals in leak check. This is to match
+        // with the old solver's behavior, which has separated evaluation and fulfillment,
+        // and the former doesn't consider outlives obligations from the later.
+        let vis = if is_root_goal {


why special-case root goals here?

View changes since the review

I thought that we should only ignore region constraints from nested goals and if I mark them as VisibleForLeakCheck::No for root goals as well, they would be so in the InferCtxt outside of the solver and then ignored from the leak check for other goals as well, but after thinking twice, they wouldn't cause universe leaks in other goals anyway 😅 Yeah, so might be okay to remove the spacial casing here

lcnr · 2026-04-30T07:24:24Z

            .into_iter()
-            .filter(|&(outlives, _)| seen.insert(outlives))
-            .map(|(outlives, _)| outlives)
+            .filter(|&(outlives, _, vis)| seen.insert((outlives, vis)))


this may cause us to end up with the same constraint with vis: Yes and vis: No 🤔 I guess it's fine as long as that doesn't cause issues

View changes since the review

I did this deliberately as we could already have the same constraints with different visibilities without this and if that would be the case, the resulting visibility would be order-dependent 🤔

could eagerly collect into a list, and use a hashmap pointing to the index of the element, on a hit we Or the visibility

lcnr · 2026-04-30T07:27:25Z

                            t1,
                            r2,
                            constraint_category,
+                            ty::VisibleForLeakCheck::Yes,


slightly concerning 🤔

how hard is it to pass VisibleForLeakCheck to here?

View changes since the review

I guess it wouldn't be that hard but I thought visibilities are not real concerns here in borrowck

they are not 😅 I would love us to somehow defensively guard against this... as rn u need to globally reason about this code for whether it's correct

what do you think you make it into a three-state enum with Yes, No, Unreachable and have the leak_check ICE when encountering Unreachable

Ah, yes, such global context would be very confusing and fragile to future changes 😅 That would be more desirable 👍

lcnr · 2026-04-30T07:28:49Z

        ty: Ty<'tcx>,
        region: ty::Region<'tcx>,
        category: ConstraintCategory<'tcx>,
+        vis: ty::VisibleForLeakCheck,


components_must_outlive can always be ty::VisibleForLeakcheck::No/ for TypeOutlives in general 🤔

View changes since the review

I guess they might be decomposed into RegionOutlivesPredicate with component-wise thingy. Would it be okay?

it matches the old solver which currently never considers type outlives in the leak check 🤔

but also, we only ever use this in places which won't ever leak check anymore, so 🤷

Maybe we could try VisibleForLeakCheck::Unreachable here as well?

rustbot · 2026-04-30T15:48:17Z

This PR was rebased onto a different main 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.

…akcheck

ShoyuVanilla changed the title ~~-Znext-solver Ignore region constraints from the nested goals in leakcheck~~ -Znext-solver Ignore region constraints from the nested goals in leakcheck Apr 24, 2026

ShoyuVanilla force-pushed the leakcheck-vis branch from 589add8 to e4719c4 Compare April 25, 2026 02:50

ShoyuVanilla mentioned this pull request Apr 25, 2026

Canonicalize free regions from inputs as placeholders in root univ #155487

Merged

This comment has been minimized.

Sign in to view

ShoyuVanilla force-pushed the leakcheck-vis branch 2 times, most recently from e2650aa to 54588cf Compare April 27, 2026 16:23

This comment has been minimized.

Sign in to view

ShoyuVanilla force-pushed the leakcheck-vis branch from 54588cf to 29dab0f Compare April 30, 2026 03:22

rustbot assigned lcnr Apr 30, 2026

ShoyuVanilla marked this pull request as ready for review April 30, 2026 03:23

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 Apr 30, 2026

lcnr reviewed Apr 30, 2026

View reviewed changes

ShoyuVanilla force-pushed the leakcheck-vis branch from 29dab0f to cff8c42 Compare April 30, 2026 15:48

-Znext-solver Ignore region constraints from the nested goals in le…

7be9fb6

…akcheck

ShoyuVanilla force-pushed the leakcheck-vis branch from cff8c42 to 7be9fb6 Compare April 30, 2026 17:49

-                                each_edge(sub, sup, visible_for_leak_check)
+                                match visible_for_leak_check {
+                                    ty::VisibleForLeakCheck::Yes =>
+                                each_edge(sub, sup),
+                                    ty::VisibleForLeakCheck::No => {}
+                                }

		pub type QueryRegionConstraint<'tcx> =
		(ty::RegionConstraint<'tcx>, ConstraintCategory<'tcx>, ty::VisibleForLeakCheck);

Uh oh!

Conversation

ShoyuVanilla commented Apr 24, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

ShoyuVanilla commented Apr 30, 2026

Uh oh!

rustbot commented Apr 30, 2026

Uh oh!

lcnr Apr 30, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lcnr Apr 30, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lcnr Apr 30, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lcnr Apr 30, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lcnr Apr 30, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lcnr Apr 30, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rustbot commented Apr 30, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

ShoyuVanilla commented Apr 24, 2026 •

edited by rustbot

Loading

lcnr Apr 30, 2026 •

edited by rustbot

Loading

lcnr Apr 30, 2026 •

edited by rustbot

Loading

lcnr Apr 30, 2026 •

edited by rustbot

Loading

lcnr Apr 30, 2026 •

edited by rustbot

Loading

lcnr Apr 30, 2026 •

edited by rustbot

Loading

lcnr Apr 30, 2026 •

edited by rustbot

Loading