trait_selection: fix type-const eval ICE by JohnTitor · Pull Request #152040 · rust-lang/rust

JohnTitor · 2026-02-03T12:35:04Z

I'd recommend reviewing commit-by-commit, the diff is less-readable to address a cyclic issue.

rustbot · 2026-02-03T12:35:09Z

fmease is currently at their maximum review capacity.
They may take a while to respond.

JohnTitor · 2026-02-03T12:36:29Z

compiler/rustc_trait_selection/src/traits/mod.rs

+                    if !type_const_stack.insert(uv.def) {
+                        let guar = tcx.dcx().span_err(
+                            tcx.def_span(uv.def),
+                            "cycle detected when evaluating type-level constant",


I'm not sure if we should detect a cyclic error in this place though, maybe we can handle it better?

I'm don't know how/where/when we invoke try_evaluate_const (during normalization of course, but …) but ideally (?) either the query system would detect cycles / overflows for us (but I guess no queries are involved here) or we have a simple depth param similar to what type normalization does instead of storing visited definitions. idk, imma wait for Boxy

Human9000-bit · 2026-02-05T13:34:15Z

Does this code compiles under your impl?

JohnTitor · 2026-02-07T08:06:49Z

@Human9000-bit Yeah, it should. Added a rev so that we can make sure it.

fmease

Some preliminary comments & suggestions. I'm now gonna pass the assignment to Boxy since I don't feel comfortable with this part of the code (yet).

View changes since this review

tests/ui/const-generics/type-const-ice-issue-151631.rs

fmease · 2026-02-10T19:51:04Z

compiler/rustc_trait_selection/src/traits/mod.rs

                }
            };

+            if tcx.is_type_const(uv.def) {


I don't know if this change makes sense, I'm not an expert in mGCA. Therefore I'll assign @BoxyUwU instead.

fmease · 2026-02-10T19:58:41Z

tests/ui/const-generics/type-const-cycle.rs

+trait Owner {
+    #[type_const]
+    const C<const N: u32>: u32;
+    //~^ ERROR cycle detected when evaluating type-level constant


Ideally this error would point to the C in the trait impl.

fmease · 2026-02-10T20:02:25Z

compiler/rustc_trait_selection/src/traits/mod.rs

+                    if !type_const_stack.insert(uv.def) {
+                        let guar = tcx.dcx().span_err(
+                            tcx.def_span(uv.def),
+                            "cycle detected when evaluating type-level constant",


I'm don't know how/where/when we invoke try_evaluate_const (during normalization of course, but …) but ideally (?) either the query system would detect cycles / overflows for us (but I guess no queries are involved here) or we have a simple depth param similar to what type normalization does instead of storing visited definitions. idk, imma wait for Boxy

rustbot · 2026-02-11T11:34:13Z

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.

BoxyUwU · 2026-02-14T19:43:28Z

compiler/rustc_trait_selection/src/traits/mod.rs

+                    }
+                };
+
+                if tcx.is_type_const(uv.def) {


We're not supposed to be calling try_evaluate_const on a type const in the first place 🤔 Whatever code is causing this to be reachable is the problematic place.

Can you look into that a bit and then explain to me what's going on there?

I see!

Can you look into that a bit and then explain to me what's going on there?

IIUC:

we generate ConstEvaluatable obligation in const_evaluatable_predicates_of, including type consts.

the next solver handles that in compute_const_evaluatable_goal by calling delegate.evaluate_const, which forwards to try_evaluate_const.

Do you have any idea for proper fix?

Ah interesting, follow up question: Where does the ConstEvaluatable obligation come from? I'm not sure why const_evaluatable_predicates_of would be returning anything here

Sorry, I misunderstood.

WfPredicates::visit_const emits ClauseKind::ConstEvaluatable(c) for ConstKind::Unevaluated:

rust/compiler/rustc_trait_selection/src/traits/wf.rs

Lines 1019 to 1021 in fbd6934

let predicate = ty::Binder::dummy(ty::PredicateKind::Clause(

ty::ClauseKind::ConstEvaluatable(c),

));

There, IIUC we don't have a check for type const.

And delegate calls unnormalized_obligations:

rust/compiler/rustc_trait_selection/src/solve/delegate.rs

Line 197 in fbd6934

crate::traits::wf::unnormalized_obligations(

This place should register ConstEvaluatable obligations as goals (right? sorry if I'm wrong).

Then compute_const_evaluatable_goal calls evaluate_const -> try_evaluate_const

BoxyUwU · 2026-02-14T19:43:36Z

@rustbot author

rustbot assigned fmease Feb 3, 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. labels Feb 3, 2026

JohnTitor commented Feb 3, 2026

View reviewed changes

JohnTitor force-pushed the issue-151631 branch from 3fc70e5 to e853189 Compare February 7, 2026 08:04

This comment has been minimized.

Sign in to view

fmease reviewed Feb 10, 2026

View reviewed changes

fmease assigned BoxyUwU and unassigned fmease Feb 10, 2026

JohnTitor added 2 commits February 11, 2026 20:10

trait_selection: fix type-const eval ICE

3587bf8

guard recursive type-const evaluation

1641602

JohnTitor force-pushed the issue-151631 branch from e853189 to 1641602 Compare February 11, 2026 11:34

BoxyUwU reviewed Feb 14, 2026

View reviewed changes

rustbot 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 Feb 14, 2026

	let predicate = ty::Binder::dummy(ty::PredicateKind::Clause(
	ty::ClauseKind::ConstEvaluatable(c),
	));

Uh oh!

Comments

Conversation

JohnTitor commented Feb 3, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Feb 3, 2026

Uh oh!

JohnTitor Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

fmease Feb 10, 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!

Human9000-bit commented Feb 5, 2026

Uh oh!

This comment has been minimized.

JohnTitor commented Feb 7, 2026

Uh oh!

This comment has been minimized.

fmease 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!

fmease Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

fmease Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

fmease Feb 10, 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 Feb 11, 2026

Uh oh!

BoxyUwU Feb 14, 2026

Choose a reason for hiding this comment

Uh oh!

JohnTitor Feb 18, 2026

Choose a reason for hiding this comment

Uh oh!

BoxyUwU Feb 18, 2026

Choose a reason for hiding this comment

Uh oh!

JohnTitor Feb 19, 2026

Choose a reason for hiding this comment

Uh oh!

BoxyUwU commented Feb 14, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

JohnTitor commented Feb 3, 2026 •

edited by rustbot

Loading

fmease Feb 10, 2026 •

edited

Loading

fmease left a comment •

edited by rustbot

Loading

fmease Feb 10, 2026 •

edited

Loading