Anonymize inner binders when checking refining by mu001999 · Pull Request #165222 · rust-lang/rust

mu001999 · 2026-10-31T07:58:41Z

The inner binder is missed.

fmease

Thanks for working on this! I got a smallish request :)

fmease · 2026-11-04T00:51:14Z

        T: TypeFoldable<TyCtxt<'tcx>>,
    {
-        self.tcx.anonymize_bound_vars(t)
+        self.tcx.anonymize_bound_vars(t.super_fold_with(self))


This implementation is obviously correct but it's doing a lot unnecessary walking & work. Of course, this part of the code isn't performance critical & I don't think we have any benchmarks containing RPITITs. Still, IMO we should simplify this, it won't take that much more code.

As you know, anonymize_bound_vars only anonymizes bound vars bound by the current binder & under the hood it's also a TypeFolder (with some TypeFlags and other early returns) meaning in total we rewalk the type a lot if there are several binders.

Moreover, the BoundVarReplacer (a folder) inside anonymize_bound_vars performs a lot of checks to ensure it only replaces bound vars that are bound at the correct level & it has to shift the de-Bruijn indices & re-intern the shifted regions etc. produced by anonymize_bound_vars's BoundVarReplacerDelegate.

However, since the idea of our Anonymize here is to fully anonymize the type, most of these things don't need to be done. Fundamentally we simply we want to replace any Bound{Region,Ty,Const}Kind with Anon keeping the de Bruijn index & the BoundVar (index) fully intact.

So this whole operation could just be a TypeFolder that impls fold_{region,type,const}, returns the argument unchanged if it !_.has_bound_vars() otherwise if it's a ty::{,Re,ConstKind::}Bound return a *Bound with *Kind::Anon.

E.g., for code like

#![deny(refining_impl_trait)] trait A { fn f() -> impl B<for<'a> fn(for<'b> fn(&'a (), &'b ()))>; } impl A for () { fn f() -> impl B<for<'a> fn(for<'b> fn(&'a (), &'b ()))> { 0 } } trait B<T: ?Sized> {} impl<T: ?Sized, U: ?Sized> B<T> for U {}

the execution trace under your PR looks something like

fold_binder() # arg: Binder { value: TraitClause(<Opaque(DefId(0:10 ~ b[656a]::{impl#0}::f::{opaque#0}), []) as B<for<Region(BrNamed(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)))> fn(for<Region(BrNamed(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)))> fn(&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) ()))>>, polarity:Positive), bound_vars: [] } # arg has bound vars; calling super_fold_with()... fold_ty() # arg: Alias(Yes, Alias { kind: Opaque { def_id: DefId(0:10 ~ b[656a]::{impl#0}::f::{opaque#0}) }, args: [], .. }) # arg has no bound vars; ret arg as is... fold_ty() # arg: Binder { value: fn(Binder { value: fn(&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) ()), bound_vars: [Region(BrNamed(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)))] }), bound_vars: [Region(BrNamed(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)))] } # arg has bound vars; calling super_fold_with()... fold_binder() # arg: Binder { value: FnSigTys { inputs_and_output: [Binder { value: fn(&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) ()), bound_vars: [Region(BrNamed(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)))] }, ()] }, bound_vars: [Region(BrNamed(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)))] } # arg has bound vars; calling super_fold_with()... fold_ty() # arg: Binder { value: fn(&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) ()), bound_vars: [Region(BrNamed(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)))] } # arg has bound vars; calling super_fold_with()... fold_binder() # arg: Binder { value: FnSigTys { inputs_and_output: [&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) (), ()] }, bound_vars: [Region(BrNamed(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)))] } # arg has bound vars; calling super_fold_with()... fold_ty() # arg: &'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) () # arg has bound vars; calling super_fold_with()... fold_ty() # arg: () # arg has no bound vars; ret arg as is... fold_ty() # arg: &'^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) () # arg has bound vars; calling super_fold_with()... fold_ty() # arg: () # arg has no bound vars; ret arg as is... fold_ty() # arg: () # arg has no bound vars; ret arg as is... anonymize_bound_vars() # arg: Binder { value: FnSigTys { inputs_and_output: [&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) (), ()] }, bound_vars: [Region(BrNamed(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)))] } # has esc bound vars; calling arg.fold_with(BVR)... BVR::fold_ty() # arg: &'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) () # has vars bound at/above current index # calling super_fold_with()... BVR::fold_region() # arg: '^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) # not ReBound with matching index; returning arg as is... BVR::fold_ty() # arg: () # no vars bound at/above current index; ret arg as is... BVR::fold_ty() # arg: &'^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) () # has vars bound at/above current index # calling super_fold_with()... BVR::fold_region() # arg: '^0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)) # is ReBound with matching index # calling delegate.replace_region(0.Named(DefId(0:7 ~ b[656a]::A::f::{opaque#0}::'b)))... # ...which yielded '^0 # adjusting the de Bruijn index... # ...giving us '^0 BVR::fold_ty() # arg: () # no vars bound at/above current index; ret arg as is... BVR::fold_ty() # arg: () # no vars bound at/above current index; ret arg as is... fold_ty() # arg: () # arg has no bound vars; ret arg as is... anonymize_bound_vars() # arg: Binder { value: FnSigTys { inputs_and_output: [Binder { value: fn(&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0 ()), bound_vars: [Region(BrAnon)] }, ()] }, bound_vars: [Region(BrNamed(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)))] } # has esc bound vars; calling arg.fold_with(BVR)... BVR::fold_ty() # arg: Binder { value: fn(&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0 ()), bound_vars: [Region(BrAnon)] } # has vars bound at/above current index # calling super_fold_with()... BVR::fold_binder() # arg: Binder { value: FnSigTys { inputs_and_output: [&'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) (), &'^0 (), ()] }, bound_vars: [Region(BrAnon)] } # calling super_fold_with()... BVR::fold_ty() # arg: &'^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) () # has vars bound at/above current index # calling super_fold_with()... BVR::fold_region() # arg: '^1_0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)) # is ReBound with matching index # calling delegate.replace_region(0.Named(DefId(0:6 ~ b[656a]::A::f::{opaque#0}::'a)))... # ...which yielded '^0 # adjusting the de Bruijn index... # ...giving us '^1_0 BVR::fold_ty() # arg: () # no vars bound at/above current index; ret arg as is... BVR::fold_ty() # arg: &'^0 () # no vars bound at/above current index; ret arg as is... BVR::fold_ty() # arg: () # no vars bound at/above current index; ret arg as is... BVR::fold_ty() # arg: () # no vars bound at/above current index; ret arg as is... anonymize_bound_vars() # arg: Binder { value: TraitClause(<Opaque(DefId(0:10 ~ b[656a]::{impl#0}::f::{opaque#0}), []) as B<for<Region(BrAnon)> fn(for<Region(BrAnon)> fn(&'^1_0 (), &'^0 ()))>>, polarity:Positive), bound_vars: [] } # no esc bound vars; ret arg...

See how it walks into parts of the type just to walk out and reenter parts it's already visited & partially anonymized (but now under different Binder.bound_vars / at a different binder depth)?

rustbot · 2026-11-05T15:23:05Z

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.

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Oct 31, 2026

mu001999 marked this pull request as ready for review October 31, 2026 13:47

rustbot added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 31, 2026

rustbot assigned beetrees Oct 31, 2026

rustbot removed the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Oct 31, 2026

This comment was marked as outdated.

Sign in to view

fmease assigned fmease and unassigned beetrees Nov 3, 2026

fmease reviewed Nov 4, 2026

View reviewed changes

fmease 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 Nov 4, 2026

Anonymize bound vars fully

59a2560

mu001999 force-pushed the fix-165091 branch from 39cf2f0 to 59a2560 Compare November 5, 2026 15:23

Navigation Menu

Repository navigation

Anonymize inner binders when checking refining - #165222

Anonymize inner binders when checking refining#165222
mu001999 wants to merge 1 commit into
rust-lang:mainfrom
mu001999-contrib:fix-165091

mu001999 commented Oct 31, 2026 •

edited by rustbot

Loading

Uh oh!

This comment was marked as outdated.

fmease left a comment •

edited

Loading

Uh oh!

Uh oh!

fmease Nov 4, 2026 •

edited

Loading

Uh oh!

fmease Nov 4, 2026 •

edited

Loading

Uh oh!

rustbot commented Nov 5, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

mu001999 commented Oct 31, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as outdated.

fmease left a comment • 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!

Uh oh!

fmease Nov 4, 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!

fmease Nov 4, 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!

rustbot commented Nov 5, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

mu001999 commented Oct 31, 2026 •

edited by rustbot

Loading

fmease left a comment •

edited

Loading

fmease Nov 4, 2026 •

edited

Loading

fmease Nov 4, 2026 •

edited

Loading