rustc_resolve: improve const generic errors by Unique-Usman · Pull Request #152913 · rust-lang/rust

Unique-Usman · 2026-02-20T14:49:03Z

No description provided.

rustbot · 2026-02-20T14:49:08Z

rustbot has assigned @JonathanBrouwer.
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: compiler
compiler expanded to 68 candidates
Random selection from 14 candidates

Unique-Usman · 2026-02-20T14:49:21Z

r? @estebank

compiler/rustc_resolve/src/late/diagnostics.rs

compiler/rustc_hir_analysis/src/hir_ty_lowering/generics.rs

compiler/rustc_resolve/src/late/diagnostics.rs

tests/ui/const-generics/const-generic-function.stderr

tests/ui/const-generics/const-arg-in-const-arg.min.stderr

compiler/rustc_resolve/src/build_reduced_graph.rs

tests/ui/missing/missing-items/missing-type-parameter2.stderr

estebank · 2026-03-06T17:31:53Z

compiler/rustc_resolve/src/build_reduced_graph.rs

                        .struct_constructors
                        .insert(local_def_id, (ctor_res, ctor_vis.to_def_id(), ret_fields));
                }
+                self.r.struct_generics.insert(local_def_id, generics.clone());


Let's see if we can struct_generics store references. That should keep mem usage to a minimum.

estebank · 2026-03-06T17:32:57Z

compiler/rustc_resolve/src/late.rs

+                if let Some(const_err) = const_err {
+                    err.cancel();
+                    err = const_err;
+                }


This is currently inactive, right? const_err isn't None anywhere right now?

const_err is actually always None except in the case of where we add the error we just introduce see it here -> https://github.com/rust-lang/rust/pull/152913/changes#diff-2c81e6d8048a31a3ce43f30181aaa037e0ad0076e7aee67792ca993f28c75ac3R3440-R3441

it is to prevent the type error from showing up.

tests/ui/const-generics/early/invalid-const-arguments.stderr

rustbot · 2026-03-06T17:35:33Z

HIR ty lowering was modified

cc @fmease

Unique-Usman · 2026-03-10T02:48:12Z

@estebank

estebank · 2026-03-12T04:10:08Z

tests/ui/const-generics/assoc_const_as_type_argument.stderr

@@ -4,18 +4,19 @@ error[E0575]: expected associated type, found associated constant `Trait::ASSOC`
 LL |     bar::<<T as Trait>::ASSOC>();
   |           ^^^^^^^^^^^^^^^^^^^ not a associated type

-error[E0747]: unresolved item provided when a constant was expected
-  --> $DIR/assoc_const_as_type_argument.rs:8:11
+error[E0284]: type annotations needed


It is unfortunate that after silencing E0747 we end up with these inference errors instead...

I think we can split the work into two PRs: one silencing the unnecessary/redundant resolve error ("unresolved item when X was expected") and another for the appropriate const generic suggestions. That would make things easier to review and evolve on isolation.

Yeah, thanks. we can resolve the E0747 error after getting this pr merged. I already update the patch to not silence the E0747 for now.

estebank · 2026-03-12T04:11:45Z

Lets see what impact keeping track of all struct Generics in Resolver has. I don't think we can make the lifetimes work with borrowing &Generics without some more significant changes, but I will try tomorrow.

@bors try @rust-timer queue

rustc_resolve: improve const generic errors

rust-bors · 2026-03-12T06:19:29Z

☀️ Try build successful (CI)
Build commit: 94de249 (94de2490ee19b3ff6f37c252af545a638a79ec5d, parent: d1ee5e59a964a419b84b760812a35075034f4861)

rust-timer · 2026-03-12T06:59:20Z

Finished benchmarking commit (94de249): comparison URL.

Overall result: no relevant changes - no action needed

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

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

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results (secondary 0.6%)

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)	1.4%	[0.9%, 1.9%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-1.0%	[-1.0%, -1.0%]	1
All ❌✅ (primary)	-	-	0

Cycles

This benchmark run did not return any relevant results for this metric.

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 480.488s -> 481.825s (0.28%)
Artifact size: 394.87 MiB -> 394.91 MiB (0.01%)

Signed-off-by: Usman Akinyemi <usmanakinyemi202@gmail.com>

estebank

Will continue reading the logic in diagnostics.rs.

I tried today to silence the errors. I'd found one way that works for almost every case, except one where an outer const parameter is used in a nested item, which ICEs, so I'll look for another alternative.

View changes since this review

estebank · 2026-03-12T17:12:27Z

tests/ui/delegation/generics/generics-gen-args-errors.rs

        //~| ERROR: cannot find type `asd` in this scope
-        //~| ERROR: unresolved item provided when a constant was expected
+        //~| ERROR unresolved item provided when a constant was expected

        reuse foo::<A, B, C> as xd;
        //~^ ERROR can't use generic parameters from outer item
        //~| ERROR can't use generic parameters from outer item
        //~| ERROR can't use generic parameters from outer item
-        //~| ERROR: unresolved item provided when a constant was expected
+        //~| ERROR unresolved item provided when a constant was expected


Revert these trivial changes to the test.

estebank · 2026-03-12T22:08:06Z

compiler/rustc_resolve/src/errors.rs


+#[derive(Subdiagnostic)]
+#[suggestion(
+    "you might have meant to introduce a const parameter on the `{$item_name}`",


Suggested change

"you might have meant to introduce a const parameter on the `{$item_name}`",

"you might have meant to introduce a const parameter on `{$item_name}`",

estebank · 2026-03-12T22:14:06Z

tests/ui/const-generics/early/invalid-const-arguments.stderr

  --> $DIR/invalid-const-arguments.rs:14:32
   |
-LL | struct A<const N: u8>;
-   | ---------------------- similarly named struct `A` defined here
-...
+LL | struct C<const C: u8, const N: u8>;
+   |                       ----------- const parameter `T` is defined on the type
 LL | impl<const N: u8> Foo for C<N, T> {}
-   |                                ^
+   |                                ^ not found in this scope


This reads a bit weirdly because the name of the const in the struct is different than on the impl. Let's think about how we can make it clearer, specially in the first span label.

estebank · 2026-03-12T22:15:21Z

compiler/rustc_resolve/src/late/diagnostics.rs

+    // Detects missing const parameters in `impl` blocks and suggests adding them.
+    //
+    // When a const parameter is used in the self type of an `impl` but not declared
+    // in the `impl`'s own generic parameter list, this function emits a targeted
+    // diagnostic with a suggestion to add it at the correct position.
+    //
+    // Example:
+    //
+    // struct C<const A: u8, const X: u8, const P: u32>;
+    //
+    // impl Foo for C<A, X, P> {}
+    // //           ^ the struct `C` in `C<A, X, P>` is used as the self type
+    // //             ^ ^ ^ but A, X and P are not declared on the impl
+    //
+    // Suggested fix:
+    //
+    // impl<const A: u8, const X: u8, const P: u32> Foo for C<A, X, P> {}
+    //
+    // Current behavior (suggestions are emitted one-by-one):
+    //
+    // impl<const A: u8> Foo for C<A, X, P> {}
+    // impl<const X: u8> Foo for C<A, X, P> {}
+    // impl<const P: u32> Foo for C<A, X, P> {}
+    //
+    // Ideally the suggestion should aggregate them into a single line:
+    //
+    // impl<const A: u8, const X: u8, const P: u32> Foo for C<A, X, P> {}


Turn this into a docstring with ///.

estebank · 2026-03-12T22:17:04Z

compiler/rustc_resolve/src/late/diagnostics.rs

+        if let Some(item) = self.diag_metadata.current_item
+            && let ItemKind::Impl(impl_) = &item.kind
+        {


You can use early return to avoid one level of indentation:

let Some(item) = self.diag_metadata.current_item else { return None }; let ItemKind::Impl(impl_) = &item.kind else { return None };

rustbot assigned JonathanBrouwer Feb 20, 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 20, 2026

rustbot assigned estebank and unassigned JonathanBrouwer Feb 20, 2026

Unique-Usman marked this pull request as draft February 20, 2026 14:49

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 20, 2026

estebank reviewed Feb 23, 2026

View reviewed changes

compiler/rustc_resolve/src/late/diagnostics.rs Show resolved Hide resolved

compiler/rustc_resolve/src/late/diagnostics.rs Show resolved Hide resolved

Unique-Usman force-pushed the ua/constnottype branch from 00fa325 to 98d6219 Compare February 28, 2026 21:28

Unique-Usman requested a review from estebank February 28, 2026 21:31

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 Feb 28, 2026

Unique-Usman commented Feb 28, 2026

View reviewed changes

compiler/rustc_hir_analysis/src/hir_ty_lowering/generics.rs Outdated Show resolved Hide resolved