Rollup merge of rust-lang#135331 - fmease:ban-assoc-ty-unbounds, r=lcnr

Reject relaxed bounds inside associated type bounds (ATB)

**Reject** relaxed bounds — most notably `?Sized` — inside associated type bounds `TraitRef<AssocTy: …>`.

This was previously accepted without warning despite being incorrect: ATBs are *not* a place where we perform *sized elaboration*, meaning `TraitRef<AssocTy: …>` does *not* elaborate to `TraitRef<AssocTy: Sized + …>` if `…` doesn't contain `?Sized`. Therefore `?Sized` is meaningless. In no other (stable) place do we (intentionally) allow relaxed bounds where we don't also perform sized elab, this is highly inconsistent and confusing! Another point of comparison: For the desugared `$SelfTy: TraitRef, $SelfTy::AssocTy: …` we don't do sized elab either (and thus also don't allow relaxed bounds).

Moreover — as I've alluded to back in rust-lang#135841 (review) — some later validation steps only happen during sized elaboration during HIR ty lowering[^1]. Namely, rejecting duplicates (e.g., `?Trait + ?Trait`) and ensuring that `Trait` in `?Trait` is equal to `Sized`[^2]. As you can probably guess, on stable/master we don't run these checks for ATBs (so we allow even more nonsensical bounds like `Iterator<Item: ?Copy>` despite T-types's ruling established in the FCP'ed rust-lang#135841).

This PR rectifies all of this. I cratered this back in 2025-01-10 with (allegedly) no regressions found ([report](rust-lang#135331 (comment)), [its analysis](rust-lang#135331 (comment))). [However a contributor manually found two occurrences](rust-lang#135229 (comment)) of `TraitRef<AssocTy: ?Sized>` in small hobby projects (presumably via GH code search). I immediately sent downstream PRs: Gui-Yom/turbo-metrics#14, ireina7/summon#1 (however, the owners have showed no reaction so far).

I'm leaning towards banning these forms **without a FCW** because a FCW isn't worth the maintenance cost[^3]. Note that associated type bounds were stabilized in 1.79.0 (released 2024-06-13 which is 13 months ago), so the proliferation of ATBs shouldn't be that high yet. If you think we should do another crater run since the last one was 6 months ago, I'm fine with that.

Fixes rust-lang#135229.

[^1]: I consider this a flaw in the implementation and [I've already added a huge FIXME](https://github.com/rust-lang/rust/blob/82a02aefe07092c737c852daccebf49ca25507e3/compiler/rustc_hir_analysis/src/hir_ty_lowering/bounds.rs#L195-L207).
[^2]: To be more precise, if the internal flag `-Zexperimental-default-bounds` is provided other "default traits" (needs internal feature `lang_items`) are permitted as well (cc closely related internal feature: `more_maybe_bounds`).
[^3]: Having to track this and adding an entire lint whose remnants would remain in the code base forever (we never *fully* remove lints).

Author: Zalathar

compiler/rustc_ast_lowering/src/lib.rs

@@ -296,6 +296,7 @@ enum RelaxedBoundPolicy<'a> {
 enum RelaxedBoundForbiddenReason {
     TraitObjectTy,
     SuperTrait,
+    AssocTyBounds,
     LateBoundVarsInScope,
 }
 
@@ -1109,9 +1110,11 @@ impl<'a, 'hir> LoweringContext<'a, 'hir> {
                         &*self.arena.alloc(self.ty(constraint.span, hir::TyKind::Err(guar)));
                     hir::AssocItemConstraintKind::Equality { term: err_ty.into() }
                 } else {
-                    // FIXME(#135229): These should be forbidden!
-                    let bounds =
-                        self.lower_param_bounds(bounds, RelaxedBoundPolicy::Allowed, itctx);
+                    let bounds = self.lower_param_bounds(
+                        bounds,
+                        RelaxedBoundPolicy::Forbidden(RelaxedBoundForbiddenReason::AssocTyBounds),
+                        itctx,
+                    );
                     hir::AssocItemConstraintKind::Bound { bounds }
                 }
             }
@@ -2124,7 +2127,8 @@ impl<'a, 'hir> LoweringContext<'a, 'hir> {
                         diag.emit();
                         return;
                     }
-                    RelaxedBoundForbiddenReason::LateBoundVarsInScope => {}
+                    RelaxedBoundForbiddenReason::AssocTyBounds
+                    | RelaxedBoundForbiddenReason::LateBoundVarsInScope => {}
                 };
             }
         }

compiler/rustc_hir_analysis/src/hir_ty_lowering/bounds.rs

@@ -199,12 +199,10 @@ impl<'tcx> dyn HirTyLowerer<'tcx> + '_ {
             //        However, this can easily get out of sync! Ideally, we would perform this step
             //        where we are guaranteed to catch *all* bounds like in
             //        `Self::lower_poly_trait_ref`. List of concrete issues:
-            //        FIXME(more_maybe_bounds): We don't call this for e.g., trait object tys or
-            //                                  supertrait bounds!
+            //        FIXME(more_maybe_bounds): We don't call this for trait object tys, supertrait
+            //                                  bounds or associated type bounds (ATB)!
             //        FIXME(trait_alias, #143122): We don't call it for the RHS. Arguably however,
-            //                                       AST lowering should reject them outright.
-            //        FIXME(associated_type_bounds): We don't call this for them. However, AST
-            //                                       lowering should reject them outright (#135229).
+            //                                     AST lowering should reject them outright.
             let bounds = collect_relaxed_bounds(hir_bounds, self_ty_where_predicates);
             self.check_and_report_invalid_relaxed_bounds(bounds);
         }

tests/ui/trait-bounds/more_maybe_bounds.rs

@@ -1,7 +1,7 @@
 // FIXME(more_maybe_bounds): Even under `more_maybe_bounds` / `-Zexperimental-default-bounds`,
 // trying to relax non-default bounds should still be an error in all contexts! As you can see
-// there are places like supertrait bounds and trait object types where we currently don't perform
-// this check.
+// there are places like supertrait bounds, trait object types or associated type bounds (ATB)
+// where we currently don't perform this check.
 #![feature(auto_traits, more_maybe_bounds, negative_impls)]
 
 trait Trait1 {}
@@ -13,11 +13,15 @@ trait Trait4 where Self: Trait1 {}
 
 // FIXME: `?Trait2` should be rejected, `Trait2` isn't marked `#[lang = "default_traitN"]`.
 fn foo(_: Box<(dyn Trait3 + ?Trait2)>) {}
+
 fn bar<T: ?Sized + ?Trait2 + ?Trait1 + ?Trait4>(_: &T) {}
 //~^ ERROR bound modifier `?` can only be applied to default traits like `Sized`
 //~| ERROR bound modifier `?` can only be applied to default traits like `Sized`
 //~| ERROR bound modifier `?` can only be applied to default traits like `Sized`
 
+// FIXME: `?Trait1` should be rejected, `Trait1` isn't marked `#[lang = "default_traitN"]`.
+fn baz<T>() where T: Iterator<Item: ?Trait1> {}
+
 struct S;
 impl !Trait2 for S {}
 impl Trait1 for S {}

tests/ui/trait-bounds/more_maybe_bounds.stderr

@@ -1,17 +1,17 @@
 error: bound modifier `?` can only be applied to default traits like `Sized`
-  --> $DIR/more_maybe_bounds.rs:16:20
+  --> $DIR/more_maybe_bounds.rs:17:20
    |
 LL | fn bar<T: ?Sized + ?Trait2 + ?Trait1 + ?Trait4>(_: &T) {}
    |                    ^^^^^^^
 
 error: bound modifier `?` can only be applied to default traits like `Sized`
-  --> $DIR/more_maybe_bounds.rs:16:30
+  --> $DIR/more_maybe_bounds.rs:17:30
    |
 LL | fn bar<T: ?Sized + ?Trait2 + ?Trait1 + ?Trait4>(_: &T) {}
    |                              ^^^^^^^
 
 error: bound modifier `?` can only be applied to default traits like `Sized`
-  --> $DIR/more_maybe_bounds.rs:16:40
+  --> $DIR/more_maybe_bounds.rs:17:40
    |
 LL | fn bar<T: ?Sized + ?Trait2 + ?Trait1 + ?Trait4>(_: &T) {}
    |                                        ^^^^^^^

tests/ui/unsized/relaxed-bounds-invalid-places.rs

@@ -22,6 +22,10 @@ impl<T> S1<T> {
     fn f() where T: ?Sized {} //~ ERROR this relaxed bound is not permitted here
 }
 
+// Test associated type bounds (ATB).
+// issue: <https://github.com/rust-lang/rust/issues/135229>
+struct S6<T>(T) where T: Iterator<Item: ?Sized>; //~ ERROR this relaxed bound is not permitted here
+
 trait Tr: ?Sized {} //~ ERROR relaxed bounds are not permitted in supertrait bounds
 
 // Test that relaxed `Sized` bounds are rejected in trait object types:

tests/ui/unsized/relaxed-bounds-invalid-places.stderr

@@ -38,28 +38,36 @@ LL |     fn f() where T: ?Sized {}
    |
    = note: in this context, relaxed bounds are only allowed on type parameters defined by the closest item
 
+error: this relaxed bound is not permitted here
+  --> $DIR/relaxed-bounds-invalid-places.rs:27:41
+   |
+LL | struct S6<T>(T) where T: Iterator<Item: ?Sized>;
+   |                                         ^^^^^^
+   |
+   = note: in this context, relaxed bounds are only allowed on type parameters defined by the closest item
+
 error: relaxed bounds are not permitted in supertrait bounds
-  --> $DIR/relaxed-bounds-invalid-places.rs:25:11
+  --> $DIR/relaxed-bounds-invalid-places.rs:29:11
    |
 LL | trait Tr: ?Sized {}
    |           ^^^^^^
    |
    = note: traits are `?Sized` by default
 
 error: relaxed bounds are not permitted in trait object types
-  --> $DIR/relaxed-bounds-invalid-places.rs:29:20
+  --> $DIR/relaxed-bounds-invalid-places.rs:33:20
    |
 LL | type O1 = dyn Tr + ?Sized;
    |                    ^^^^^^
 
 error: relaxed bounds are not permitted in trait object types
-  --> $DIR/relaxed-bounds-invalid-places.rs:30:15
+  --> $DIR/relaxed-bounds-invalid-places.rs:34:15
    |
 LL | type O2 = dyn ?Sized + ?Sized + Tr;
    |               ^^^^^^
 
 error: relaxed bounds are not permitted in trait object types
-  --> $DIR/relaxed-bounds-invalid-places.rs:30:24
+  --> $DIR/relaxed-bounds-invalid-places.rs:34:24
    |
 LL | type O2 = dyn ?Sized + ?Sized + Tr;
    |                        ^^^^^^
@@ -76,5 +84,5 @@ error: bound modifier `?` can only be applied to `Sized`
 LL | struct S5<T>(*const T) where T: ?Trait<'static> + ?Sized;
    |                                 ^^^^^^^^^^^^^^^
 
-error: aborting due to 11 previous errors
+error: aborting due to 12 previous errors