Make region equality emits Eq constraints by ShoyuVanilla · Pull Request #155258 · rust-lang/rust

ShoyuVanilla · 2026-04-13T20:48:07Z

View all comments

For context, see..

The box named coroutine witness Send lifetime requirements now considered by leakcheck from this roadmap
#t-types/trait-system-refactor > A question on #251 @ 💬
Comments on rustc_type_ir::predicate::RegionEqPredicate

ShoyuVanilla · 2026-04-13T20:52:11Z

r? lcnr

rustbot · 2026-04-13T20:52:18Z

Some changes occurred to the core trait solver

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

ShoyuVanilla · 2026-04-13T20:54:28Z

tests/ui/lifetimes/explicit-self-lifetime-mismatch.rs

+        //~| ERROR mismatched `self` parameter type
+        //~| NOTE expected struct `Foo<'a, 'b>`
+        //~| NOTE found struct `Foo<'b, 'a>`
+        //~| NOTE lifetime mismatch


View changes since the review

Other tests are untouched but this has been affected somehow (I guess this might be due to having some duplication with both 'a: 'b, 'b: 'a and 'a == 'b)
I think this might be okay as it's unchanged anyway modulo diagnostics deduplication

seems fine 🤷

This somehow got returned back again while destructuring eq bounds 😂

ShoyuVanilla · 2026-04-14T03:36:52Z

@bors try @rust-timer queue

Make region equality emits Eq constraints

rust-bors · 2026-04-14T05:44:56Z

☀️ Try build successful (CI)
Build commit: 52727e5 (52727e55499277ba04069f2dfd8ca8cbbc4e76b6, parent: 338dff3e3a375cb4a3c68be825058c582262655a)

rust-timer · 2026-04-14T06:25:09Z

Finished benchmarking commit (52727e5): comparison URL.

Overall result: ❌✅ regressions and improvements - please read:

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

Next, please: If you can, justify the regressions found in this try perf run in writing along with @rustbot label: +perf-regression-triaged. If not, fix the regressions and do another perf run. Neutral or positive results will clear the label automatically.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.2%	[0.0%, 0.3%]	6
Improvements ✅ (primary)	-0.3%	[-0.5%, -0.2%]	23
Improvements ✅ (secondary)	-0.5%	[-1.3%, -0.0%]	27
All ❌✅ (primary)	-0.3%	[-0.5%, -0.2%]	23

Max RSS (memory usage)

Results (secondary -0.9%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	3.3%	[0.9%, 5.7%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.1%	[-3.6%, -2.8%]	4
All ❌✅ (primary)	-	-	0

Cycles

Results (primary -2.4%, secondary -5.3%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.4%	[-2.4%, -2.4%]	1
Improvements ✅ (secondary)	-5.3%	[-8.9%, -1.9%]	10
All ❌✅ (primary)	-2.4%	[-2.4%, -2.4%]	1

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 489.33s -> 490.845s (0.31%)
Artifact size: 394.21 MiB -> 394.03 MiB (-0.04%)

ShoyuVanilla · 2026-04-14T06:28:04Z

Oh, looks like having less number of constraints by collapsing into Eq helps the perf a bit

compiler/rustc_borrowck/src/type_check/constraint_conversion.rs

compiler/rustc_infer/src/infer/canonical/query_response.rs

lcnr

I would keep assumptions as bidirectional outlives and never have Eq in there

can you make sure we always exhaustively match on ConstraintKind 🤔 I do dislike constraint kind as a concept and feel like we should just match on the actual region instead of separately storing that in the ConstraintKind

View changes since this review

compiler/rustc_middle/src/infer/canonical.rs

compiler/rustc_infer/src/infer/outlives/mod.rs

compiler/rustc_infer/src/infer/region_constraints/leak_check.rs

compiler/rustc_type_ir/src/elaborate.rs

lcnr · 2026-04-14T09:48:12Z

compiler/rustc_type_ir/src/predicate.rs

+/// - We want to uplift bidirectional constraints to the caller instead of unifying them
+///   when solving nested goals, otherwise we often lose implied bounds.
+/// - But we still want to know they are equal from the caller. This is crucial when we
+///   are proving other nested goals which are sensitive to region equalities.


this comment is slightly confusing/unhelpful to me right now. It is hard to express why exactly we need this, because really, it is a hack required by the better solution not being possible

"otherwise we often lose implied bounds", what do you mean, what is often, link to a relevant example

I think the framing here isn't quite right. For me the issue is that having existential regions which aren't actually existential is not something which we can support.

Also, "know they are equal" seems slightly odd to me. Know they are equal means "unify existential regions with regions they are equal to, to resolve them to the same thing" here I think? Maybe make that more clear

I also think our conversation on zulip has been useful here, so maybe just link to that 😁

ShoyuVanilla · 2026-04-14T09:57:54Z

I would keep assumptions as bidirectional outlives and never have Eq in there

Okay, I'll keep them as they are.

can you make sure we always exhaustively match on ConstraintKind 🤔 I do dislike constraint kind as a concept and feel like we should just match on the actual region instead of separately storing that in the ConstraintKind

Yeah, I also thought ConstraintKind is very iffy while working on this 😅 Maybe it would be great to clean them up as a follow up. For now, I'll make sure they are exhaustively matched

lcnr · 2026-04-14T10:07:40Z

yeah, I feel like ConstraintKind just shouldn't exist :3 the fact that we currently use it instead of matching on the RegionKind at the relevant points is a mess imo, though generally the way lexical region resolution works is a mess and should be replaced at some point

lcnr · 2026-04-14T15:07:32Z

compiler/rustc_infer/src/infer/lexical_region_resolve/indexed_edges.rs

        }

        // Index in-edges in reverse order, to match what current tests expect.
        // (It's unclear whether this is important or not.)


View changes since the review

rustc is such a good codebase 🙃

can you add a fixme to merge these two loops?

I'm afraid I might be the one who often writes such good codes 😂

lcnr · 2026-04-14T15:15:00Z

compiler/rustc_infer/src/infer/lexical_region_resolve/mod.rs

-                ConstraintKind::RegSubVar => {
-                    let sup_vid = c.sup.as_var();
+
+        let reg_sub_var = |sub: Region<'tcx>,


View changes since the review

instead of making this closures, can you change the iterators to iter over self.data.contraints.iter().flat_map(|constraints| constraints.convert_to_outlives()) which does the eq to bidir outlives mapping.

And then use that everywhere u eq should just be handled as outlives

lcnr · 2026-04-14T15:16:31Z

compiler/rustc_infer/src/infer/outlives/mod.rs

        // coroutine well-formedness.
        if self.tcx.sess.opts.unstable_opts.higher_ranked_assumptions {
-            storage.data.constraints.retain(|(c, _)| match c.kind {
+            storage.data.constraints.retain_mut(|(c, _)| match c.kind {


View changes since the review

hmm, alternatively, can you already do this eq to bidir outlives when constructing the storage and then just always assert we only have outlives constraints everywhere else?

lcnr · 2026-04-14T15:17:03Z

compiler/rustc_infer/src/infer/outlives/obligations.rs

                    && outlives_env
                        .higher_ranked_assumptions()
-                        .contains(&ty::OutlivesPredicate(sup_type.into(), sub_region))
+                        .contains(&ty::OutlivesPredicate(sup_type.into(), sub_region).into())


View changes since the review

these new .into() calls should be unnecessary, are they?

lcnr · 2026-04-14T15:17:52Z

compiler/rustc_infer/src/infer/region_constraints/leak_check.rs

+                            ConstraintKind::VarSubVar
+                            | ConstraintKind::RegSubVar
+                            | ConstraintKind::VarSubReg
+                            | ConstraintKind::RegSubReg => {}


View changes since the review

here we could also change the iterator to do a flat_map on map to bidir?

lcnr · 2026-04-14T15:18:01Z

compiler/rustc_infer/src/infer/region_constraints/leak_check.rs

            }
        } else {
-            region_constraints.data().constraints.iter().for_each(|(c, _)| each_edge(c.sub, c.sup))
+            region_constraints.data().constraints.iter().for_each(|(c, _)| {


View changes since the review

and here :3

lcnr · 2026-04-14T15:20:20Z

compiler/rustc_trait_selection/src/traits/auto_trait.rs

+                        deps1.larger.insert(RegionTarget::RegionVid(vid_b));
+                        deps1.smaller.insert(RegionTarget::RegionVid(vid_b));
+                    }
+


View changes since the review

could this also just destructure eq to bidir outlives to avoid this code?

lcnr · 2026-04-14T15:21:28Z

compiler/rustc_traits/src/coroutine_witnesses.rs

+                .flat_map(|(constraint, _)| match constraint {
+                    ty::RegionConstraint::Outlives(outlives) => iter::once(outlives).chain(None),
+                    ty::RegionConstraint::Eq(eq) => {
+                        let [outlives1, outlives2] = eq.into_bidirectional_outlives();
+                        iter::once(outlives1).chain(Some(outlives2))
+                    }
+                })


View changes since the review

this match should be a method given that we should use it a bunch :3

lcnr · 2026-04-14T15:22:40Z

src/librustdoc/clean/auto_trait.rs

+                        .or_default()
+                        .push(c.sup);
+                }
+            }


View changes since the review

this can also act on the destructured eq bounds :3

lcnr · 2026-04-14T15:22:55Z

tests/ui/lifetimes/explicit-self-lifetime-mismatch.rs

+        //~| ERROR mismatched `self` parameter type
+        //~| NOTE expected struct `Foo<'a, 'b>`
+        //~| NOTE found struct `Foo<'b, 'a>`
+        //~| NOTE lifetime mismatch


seems fine 🤷

ShoyuVanilla · 2026-04-14T17:22:40Z

@bors try @rust-timer queue

Make region equality emits Eq constraints

rust-bors · 2026-04-14T19:29:55Z

☀️ Try build successful (CI)
Build commit: d50a4bb (d50a4bb74fc592f8e09fcb66ace3b73da8846142, parent: 12f35ad39ed3e39df4d953c46d4f6cc6c82adc96)

rust-timer · 2026-04-14T20:10:16Z

Finished benchmarking commit (d50a4bb): comparison URL.

Overall result: ❌✅ regressions and improvements - please read:

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

Next, please: If you can, justify the regressions found in this try perf run in writing along with @rustbot label: +perf-regression-triaged. If not, fix the regressions and do another perf run. Neutral or positive results will clear the label automatically.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.2%	[0.0%, 0.3%]	6
Improvements ✅ (primary)	-0.3%	[-0.4%, -0.2%]	17
Improvements ✅ (secondary)	-0.6%	[-1.3%, -0.0%]	21
All ❌✅ (primary)	-0.3%	[-0.4%, -0.2%]	17

Max RSS (memory usage)

Results (secondary 1.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	4.7%	[3.7%, 5.9%]	4
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.6%	[-3.1%, -1.4%]	4
All ❌✅ (primary)	-	-	0

Cycles

Results (primary 6.8%, secondary -3.5%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	6.8%	[2.0%, 10.5%]	6
Regressions ❌ (secondary)	3.1%	[2.3%, 4.0%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-5.7%	[-9.0%, -1.7%]	6
All ❌✅ (primary)	6.8%	[2.0%, 10.5%]	6

Binary size

Results (primary -0.0%, secondary -0.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.0%	[-0.0%, -0.0%]	11
Improvements ✅ (secondary)	-0.1%	[-0.1%, -0.1%]	4
All ❌✅ (primary)	-0.0%	[-0.0%, -0.0%]	11

Bootstrap: 489.936s -> 490.267s (0.07%)
Artifact size: 394.18 MiB -> 394.06 MiB (-0.03%)

lcnr · 2026-04-15T09:08:07Z

compiler/rustc_infer/src/infer/lexical_region_resolve/indexed_edges.rs

+                ConstraintKind::RegSubVar
+                | ConstraintKind::RegSubReg
+                | ConstraintKind::RegEqReg => {}
            }


can you also map use bidir for both of the loops here?

View changes since the review

lcnr · 2026-04-15T09:12:33Z

compiler/rustc_infer/src/infer/region_constraints/mod.rs

+                    self.add_constraint(
+                        Constraint { kind: ConstraintKind::VarEqVar, sub: a, sup: b },
+                        origin,
+                    );


mind adding a fixme that we could only emit this constraints if unify_var_var returns an error?

View changes since the review

lcnr

r=me after these nits

View changes since this review

rustbot assigned lcnr Apr 13, 2026

ShoyuVanilla marked this pull request as ready for review April 13, 2026 20:52

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 13, 2026

ShoyuVanilla commented Apr 13, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

ShoyuVanilla force-pushed the eq-constraint branch from 49b0929 to 8d8e4f3 Compare April 13, 2026 20:56

This comment has been minimized.

Sign in to view

ShoyuVanilla force-pushed the eq-constraint branch from 8d8e4f3 to 4119383 Compare April 14, 2026 03:34

This comment has been minimized.

Sign in to view

rust-bors bot pushed a commit that referenced this pull request Apr 14, 2026

Auto merge of #155258 - ShoyuVanilla:eq-constraint, r=<try>

52727e5

Make region equality emits Eq constraints

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Apr 14, 2026

This comment has been minimized.

Sign in to view

rustbot added perf-regression Performance regression. and removed S-waiting-on-perf Status: Waiting on a perf run to be completed. labels Apr 14, 2026

lcnr reviewed Apr 14, 2026

View reviewed changes

compiler/rustc_borrowck/src/type_check/constraint_conversion.rs Outdated Show resolved Hide resolved

lcnr reviewed Apr 14, 2026

View reviewed changes

compiler/rustc_infer/src/infer/canonical/query_response.rs Outdated Show resolved Hide resolved

lcnr reviewed Apr 14, 2026

View reviewed changes

ShoyuVanilla force-pushed the eq-constraint branch 2 times, most recently from 35c064f to 0f3c1ee Compare April 14, 2026 13:26

lcnr reviewed Apr 14, 2026

View reviewed changes

ShoyuVanilla force-pushed the eq-constraint branch from 0f3c1ee to f36814a Compare April 14, 2026 16:25

This comment has been minimized.

Sign in to view

ShoyuVanilla force-pushed the eq-constraint branch 2 times, most recently from 1cce45b to ee76953 Compare April 14, 2026 17:18

This comment has been minimized.

Sign in to view

rust-bors bot pushed a commit that referenced this pull request Apr 14, 2026

Auto merge of #155258 - ShoyuVanilla:eq-constraint, r=<try>

d50a4bb

Make region equality emits Eq constraints

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Apr 14, 2026

This comment has been minimized.

Sign in to view

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Apr 14, 2026

Make region equality emits Eq constraints

91fe2b7

ShoyuVanilla force-pushed the eq-constraint branch from ee76953 to 91fe2b7 Compare April 15, 2026 01:55

lcnr reviewed Apr 15, 2026

View reviewed changes

Uh oh!

Conversation

ShoyuVanilla commented Apr 13, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ShoyuVanilla commented Apr 13, 2026

Uh oh!

rustbot commented Apr 13, 2026

Uh oh!

ShoyuVanilla Apr 13, 2026 • 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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

This comment has been minimized.

ShoyuVanilla commented Apr 14, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

rust-bors bot commented Apr 14, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Apr 14, 2026

Overall result: ❌✅ regressions and improvements - please read:

Uh oh!

ShoyuVanilla commented Apr 14, 2026

Uh oh!

Uh oh!

Uh oh!

lcnr left a comment • 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!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ShoyuVanilla commented Apr 14, 2026

Uh oh!

lcnr commented Apr 14, 2026

Uh oh!

lcnr Apr 14, 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 14, 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 14, 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 14, 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 14, 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 14, 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 14, 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 14, 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

ShoyuVanilla commented Apr 13, 2026 •

edited by rustbot

Loading

ShoyuVanilla Apr 13, 2026 •

edited

Loading

lcnr left a comment •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 14, 2026 •

edited by rustbot

Loading

lcnr Apr 15, 2026 •

edited by rustbot

Loading

lcnr Apr 15, 2026 •

edited by rustbot

Loading

lcnr left a comment •

edited by rustbot

Loading