borrowck: Normalize non-rigid aliases in NLL type relating by Dnreikronos · Pull Request #161012 · rust-lang/rust

Dnreikronos · 2026-08-13T00:09:03Z

With -Znext-solver=globally, yielding from an impl Iterator without an explicit Item bound ICEs in borrowck. The coroutine defining type returned by type_of is unnormalized, so its yield type remains <impl Iterator as Iterator>::Item. Skipping normalization propagates that non-rigid alias into both MIR's CoroutineInfo and borrowck's UniversalRegions; NLL type relating then hits its invariant that non-rigid aliases must already have been normalized.

Deeply normalize the instantiated defining type when MIR construction creates CoroutineInfo and when borrowck reconstructs DefiningTy. That makes the coroutine yield and resume types rigid before NLL compares them.

This is intentionally gated to the next solver. The old solver keeps the existing skip-normalization path because deeply normalizing defining types there causes regressions.

rustbot · 2026-08-13T00:09:08Z

r? @camelid

rustbot has assigned @camelid.
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: borrowck, compiler
borrowck, compiler expanded to 75 candidates
Random selection from 17 candidates

adwinwhite · 2026-08-13T08:07:45Z

r? me

We really shouldn't have non-rigid aliases here. I would guess that we lack a normalization call earlier, likely yield_ty in construct_fn.

Dnreikronos · 2026-08-13T14:15:54Z

Imo we really shouldn't have non-rigid aliases in NLL relating. You were right, copying that into relate_tys was tempting and it's the wrong move. Dropping it.

construct_fn was the right guess for body.yield_ty. Writeback leaves <impl Iterator as Iterator>::Item on it sometimes. Normalizing makes the alias rigid, which is what stops the span-bug. Fyi dumps can still print the projection.

That still doesn't cover the ICE though. It's sub_types(value.ty(), yield_ty) and value.ty() is the local i, still a non-rigid alias. universal_regions.yield_ty comes off defining_ty unnormalized, same as unnormalized_output_ty.

So I'm also normalizing the operand at the Yield terminator and the defining_ty yield in equate_inputs_and_outputs. Next-solver only. Btw not touching resume_ty, this bug is yield.

Idk if this is the long-term answer once next-solver is default irl. Maybe NLL just grows real lazy norm later. Pushing asap, ltm if you'd rather I skip the operand/defining_ty bits.

adwinwhite · 2026-08-24T01:33:15Z

Sorry for the delay :>

Dnreikronos · 2026-08-24T12:43:24Z

@rustbot ready

Dnreikronos · 2026-08-24T12:45:45Z

Sup, @adwinwhite!
I've made the adjustments. Can you re-validate, please?

adwinwhite

nits nits :>

View changes since this review

Dnreikronos · 2026-08-25T14:39:34Z

Ha, good nits. All four applied, answered each one inline.

adwinwhite · 2026-08-26T00:42:33Z

@bors try @rust-timer queue

…=<try> borrowck: Normalize non-rigid aliases in NLL type relating

Dnreikronos · 2026-08-26T13:44:41Z

Unfortunately the old solver does have some regressions. Let's add the old solver early skip back. Sorry for the back and forth. And please update the PR description and squash the commits. Thanks! @rustbot author

Sup, @adwinwhite :)

I think I addressed what you want. can you re-validate, please?

@rustbot ready

adwinwhite · 2026-08-27T02:07:56Z

@bors r+ rollup

rust-bors · 2026-08-27T02:07:59Z

📌 Commit 2701241 has been approved by adwinwhite

It is now in the queue for this repository.

…_aliases, r=adwinwhite borrowck: Normalize non-rigid aliases in NLL type relating Fixes rust-lang#160652 With `-Znext-solver=globally`, yielding from an `impl Iterator` without an explicit `Item` bound ICEs in borrowck. The coroutine defining type returned by `type_of` is unnormalized, so its yield type remains `<impl Iterator as Iterator>::Item`. Skipping normalization propagates that non-rigid alias into both MIR's `CoroutineInfo` and borrowck's `UniversalRegions`; NLL type relating then hits its invariant that non-rigid aliases must already have been normalized. Deeply normalize the instantiated defining type when MIR construction creates `CoroutineInfo` and when borrowck reconstructs `DefiningTy`. That makes the coroutine yield and resume types rigid before NLL compares them. This is intentionally gated to the next solver. The old solver keeps the existing skip-normalization path because deeply normalizing defining types there causes regressions.

Rollup of 7 pull requests Successful merges: - #150075 (Implement clamp_to) - #161034 (Add SVE-accelerated Vec::retain_mut for aarch64) - #161628 (interpret: ensure that calls via no-unwind ABIs do not unwind) - #161012 (borrowck: Normalize non-rigid aliases in NLL type relating) - #161702 (Use `drop_guard` in some places in {core,alloc,std}) - #161813 (Change `is_eligible_for_coverage` from a hook to a query) - #161842 (chore: fix cargo lints)

…_aliases, r=adwinwhite borrowck: Normalize non-rigid aliases in NLL type relating Fixes rust-lang#160652 With `-Znext-solver=globally`, yielding from an `impl Iterator` without an explicit `Item` bound ICEs in borrowck. The coroutine defining type returned by `type_of` is unnormalized, so its yield type remains `<impl Iterator as Iterator>::Item`. Skipping normalization propagates that non-rigid alias into both MIR's `CoroutineInfo` and borrowck's `UniversalRegions`; NLL type relating then hits its invariant that non-rigid aliases must already have been normalized. Deeply normalize the instantiated defining type when MIR construction creates `CoroutineInfo` and when borrowck reconstructs `DefiningTy`. That makes the coroutine yield and resume types rigid before NLL compares them. This is intentionally gated to the next solver. The old solver keeps the existing skip-normalization path because deeply normalizing defining types there causes regressions.

Rollup of 6 pull requests Successful merges: - #161034 (Add SVE-accelerated Vec::retain_mut for aarch64) - #161628 (interpret: ensure that calls via no-unwind ABIs do not unwind) - #161012 (borrowck: Normalize non-rigid aliases in NLL type relating) - #161702 (Use `drop_guard` in some places in {core,alloc,std}) - #161813 (Change `is_eligible_for_coverage` from a hook to a query) - #161842 (chore: fix cargo lints)

Rollup of 7 pull requests Successful merges: - #161034 (Add SVE-accelerated Vec::retain_mut for aarch64) - #161628 (interpret: ensure that calls via no-unwind ABIs do not unwind) - #161012 (borrowck: Normalize non-rigid aliases in NLL type relating) - #161691 (Assorted bootstrap config refactors (part 1/N)) - #161813 (Change `is_eligible_for_coverage` from a hook to a query) - #161842 (chore: fix cargo lints) - #161843 (rustdoc: fix lint `cargo::non_kebab_case_bins`)

Rollup merge of #161012 - Dnreikronos:borrowck/normalize_nll_aliases, r=adwinwhite borrowck: Normalize non-rigid aliases in NLL type relating Fixes #160652 With `-Znext-solver=globally`, yielding from an `impl Iterator` without an explicit `Item` bound ICEs in borrowck. The coroutine defining type returned by `type_of` is unnormalized, so its yield type remains `<impl Iterator as Iterator>::Item`. Skipping normalization propagates that non-rigid alias into both MIR's `CoroutineInfo` and borrowck's `UniversalRegions`; NLL type relating then hits its invariant that non-rigid aliases must already have been normalized. Deeply normalize the instantiated defining type when MIR construction creates `CoroutineInfo` and when borrowck reconstructs `DefiningTy`. That makes the coroutine yield and resume types rigid before NLL compares them. This is intentionally gated to the next solver. The old solver keeps the existing skip-normalization path because deeply normalizing defining types there causes regressions.

lcnr · 2026-08-27T08:10:18Z

+    let ocx = ObligationCtxt::new(&infcx);
+    let span = tcx.def_span(def_id);
+    let cause = ObligationCause::misc(span, def_id);
+    ocx.deeply_normalize(&cause, tcx.param_env(def_id), ty).unwrap()


that drops region constraints from normalization 🤔

I assume that's fine because the yield type has already been normalized in HIR typeck, so this should just replace the non-rigid aliases with rigid ones in the closure? Could we do that explicitly with a comment explaining what's going on?

Using a local ocx during borrowck feels brittle to me

sup o/

Yeah, I think you're right about the local ocx. Basically when I wrote this I was mostly following the failure backwards from borrowck..it was getting a non-rigid alias where it expected a rigid one, so normalizing the type there fixed the ICE. It worked, but I don't think it's a great place to do it. Any region constraints created by that normalization live in a temporary inference context and disappear right after.

The part I missed is that HIR typeck had already normalized the closure type. type_of then puts it into an EarlyBinder, which marks the aliases as non-rigid again. So when borrowck gets the type, it looks like it needs normalization even though only the marker changed.

My feeling is that set_aliases_to_rigid says what we actually mean here and avoids running the solver again for no real reason. A comment would help too, because this path is pretty easy to misunderstand. I definitely misunderstood it the first time.

As this PR has already merged, what do you think makes sense as a follow-up? I could open another PR that changes this in borrowck and the MIR builder, since both read the same closure type, but I'm not sure if you'd prefer to keep it smaller.

I assume that's fine because the yield type has already been normalized in HIR typeck, so this should just replace the non-rigid aliases with rigid ones in the closure?

Oh, I missed that rigidness can be shared between typeck and borrowck.

@Dnreikronos a follow-up would be good, thank you!

@Dnreikronos can you add that fact as a comment on IsRigid I think? Unsure where we currently document the way rigidness works

sup o/

ofc, i will work on this today!

Rollup of 7 pull requests Successful merges: - rust-lang/rust#161034 (Add SVE-accelerated Vec::retain_mut for aarch64) - rust-lang/rust#161628 (interpret: ensure that calls via no-unwind ABIs do not unwind) - rust-lang/rust#161012 (borrowck: Normalize non-rigid aliases in NLL type relating) - rust-lang/rust#161691 (Assorted bootstrap config refactors (part 1/N)) - rust-lang/rust#161813 (Change `is_eligible_for_coverage` from a hook to a query) - rust-lang/rust#161842 (chore: fix cargo lints) - rust-lang/rust#161843 (rustdoc: fix lint `cargo::non_kebab_case_bins`)

@adwinwhite

…_rigidity, r=adwinwhite borrowck: Restore alias rigidity from HIR typeck Follow-up to rust-lang#161012. This came out of the review thread here: rust-lang#161012 (comment) Borrowck was normalizing the closure type again in a fresh inference context. HIR writeback had already done that work. EarlyBinder only marked the aliases as non-rigid again because it has to be conservative. If the second normalization creates region constraints, they disappear with the temporary context. I first thought normalizing again here was fine. After tracing the type back through writeback, I think it makes more sense to trust the result from HIR typeck. Borrowck now restores the rigid flag, and the IsRigid docs explain that this state can carry into borrowck and when it needs to be reset. Tested with the coroutine regression and tidy. cc @adwinwhite @lcnr

@adwinwhite

…_rigidity, r=adwinwhite borrowck: Restore alias rigidity from HIR typeck Follow-up to rust-lang#161012. This came out of the review thread here: rust-lang#161012 (comment) Borrowck was normalizing the closure type again in a fresh inference context. HIR writeback had already done that work. EarlyBinder only marked the aliases as non-rigid again because it has to be conservative. If the second normalization creates region constraints, they disappear with the temporary context. I first thought normalizing again here was fine. After tracing the type back through writeback, I think it makes more sense to trust the result from HIR typeck. Borrowck now restores the rigid flag, and the IsRigid docs explain that this state can carry into borrowck and when it needs to be reset. Tested with the coroutine regression and tidy. cc @adwinwhite @lcnr

@adwinwhite

Rollup merge of #161926 - Dnreikronos:borrowck/restore_alias_rigidity, r=adwinwhite borrowck: Restore alias rigidity from HIR typeck Follow-up to #161012. This came out of the review thread here: #161012 (comment) Borrowck was normalizing the closure type again in a fresh inference context. HIR writeback had already done that work. EarlyBinder only marked the aliases as non-rigid again because it has to be conservative. If the second normalization creates region constraints, they disappear with the temporary context. I first thought normalizing again here was fine. After tracing the type back through writeback, I think it makes more sense to trust the result from HIR typeck. Borrowck now restores the rigid flag, and the IsRigid docs explain that this state can carry into borrowck and when it needs to be reset. Tested with the coroutine regression and tidy. cc @adwinwhite @lcnr

rustbot assigned camelid Aug 13, 2026

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. WG-trait-system-refactor The Rustc Trait System Refactor Initiative (-Znext-solver) labels Aug 13, 2026

rustbot assigned adwinwhite and unassigned camelid Aug 13, 2026

This comment was marked as resolved.

Sign in to view

oli-obk closed this Aug 13, 2026

rustbot removed the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Aug 13, 2026

oli-obk reopened this Aug 13, 2026

rustbot added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Aug 13, 2026

adwinwhite reviewed Aug 17, 2026

View reviewed changes

Comment thread compiler/rustc_mir_build/src/builder/mod.rs Outdated

adwinwhite reviewed Aug 24, 2026

View reviewed changes

Comment thread compiler/rustc_borrowck/src/universal_regions.rs Outdated

Comment thread compiler/rustc_mir_build/src/builder/mod.rs Outdated

Comment thread compiler/rustc_mir_build/src/builder/mod.rs

Comment thread compiler/rustc_mir_build/src/builder/mod.rs Outdated

adwinwhite 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 Aug 24, 2026

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 Aug 24, 2026

adwinwhite reviewed Aug 25, 2026

View reviewed changes

Comment thread compiler/rustc_borrowck/src/universal_regions.rs Outdated

Comment thread compiler/rustc_mir_build/src/builder/mod.rs Outdated

Comment thread compiler/rustc_mir_build/src/builder/mod.rs Outdated

Comment thread compiler/rustc_borrowck/src/universal_regions.rs Outdated

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Aug 26, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request Aug 26, 2026

Auto merge of #161012 - Dnreikronos:borrowck/normalize_nll_aliases, r…

e3f11ef

…=<try> borrowck: Normalize non-rigid aliases in NLL type relating

Dnreikronos force-pushed the borrowck/normalize_nll_aliases branch from 1e47fcc to f8dd6a8 Compare August 26, 2026 12:57

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 Aug 26, 2026

This comment has been minimized.

Sign in to view

borrowck: Normalize non-rigid aliases in NLL type relating

2701241

Dnreikronos force-pushed the borrowck/normalize_nll_aliases branch from f8dd6a8 to 2701241 Compare August 26, 2026 14:04

rust-bors Bot added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Aug 27, 2026

jhpratt mentioned this pull request Aug 27, 2026

Rollup of 7 pull requests #161844

Closed

jhpratt mentioned this pull request Aug 27, 2026

Rollup of 6 pull requests #161846

Closed

jhpratt mentioned this pull request Aug 27, 2026

Rollup of 7 pull requests #161849

Merged

rust-bors Bot merged commit 04eae11 into rust-lang:main Aug 27, 2026
13 checks passed

rustbot added this to the 1.100.0 milestone Aug 27, 2026

lcnr reviewed Aug 27, 2026

View reviewed changes

This was referenced Aug 27, 2026

stabilize -Znext-solver=coherence again #130654

Merged

Stabilize the next-generation trait solver rust-lang/goals#113

Open

Dnreikronos mentioned this pull request Aug 28, 2026

borrowck: Restore alias rigidity from HIR typeck #161926

Merged

Uh oh!

Uh oh!

Conversation

Dnreikronos commented Aug 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Aug 13, 2026

Uh oh!

adwinwhite commented Aug 13, 2026

Uh oh!

Dnreikronos commented Aug 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as resolved.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

adwinwhite commented Aug 24, 2026

Uh oh!

Dnreikronos commented Aug 24, 2026

Uh oh!

Dnreikronos commented Aug 24, 2026

Uh oh!

adwinwhite 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!

Dnreikronos commented Aug 25, 2026

Uh oh!

adwinwhite commented Aug 26, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

Dnreikronos commented Aug 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

adwinwhite commented Aug 27, 2026

Uh oh!

rust-bors Bot commented Aug 27, 2026

Uh oh!

Uh oh!

lcnr Aug 27, 2026

Choose a reason for hiding this comment

Uh oh!

Dnreikronos Aug 27, 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!

adwinwhite Aug 28, 2026

Choose a reason for hiding this comment

Uh oh!

lcnr Aug 28, 2026

Choose a reason for hiding this comment

Uh oh!

Dnreikronos Aug 28, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

Dnreikronos commented Aug 13, 2026 •

edited

Loading

Dnreikronos commented Aug 13, 2026 •

edited

Loading

adwinwhite left a comment •

edited by rustbot

Loading

Dnreikronos commented Aug 26, 2026 •

edited

Loading

Dnreikronos Aug 27, 2026 •

edited

Loading