fix: Function argument type inference with associated type impl trait #16769
Conversation
rustbot
added
the
S-waiting-on-review
| @@ -1055,10 +1055,10 @@ impl<'a> TyLoweringContext<'a> { | |||
|
|
|||
| fn assoc_type_bindings_from_type_bound( | |||
| &'a self, | |||
| bound: &'a TypeBound, | |||
| bound: &'a Interned<TypeBound>, | |||
There was a problem hiding this comment.
Changed the argument type into Interned for faster equality comparision
| .with_type_param_mode(ParamLoweringMode::Variable); | ||
| let ctx = if let GenericDefId::FunctionId(_) = def { | ||
| TyLoweringContext::new(db, &resolver, def.into()) | ||
| .with_impl_trait_mode(ImplTraitLoweringMode::Variable) |
There was a problem hiding this comment.
impl Traits that appear in the function arguments should be handled as generic parameters with trait implementation predicates. So, we handle impl Trait by the same way with the type parameters.
There was a problem hiding this comment.
I wonder if we should just require the modes to specified in the new constructor. Feels like an easy to cause bug by relying on a default.
There was a problem hiding this comment.
That would be better. Actually, I've been considering giving a shot for #16584 while writing this PR and that would require some changes to lower.rs. It would be good to apply your suggestions while making that changes
| .with_type_param_mode(ParamLoweringMode::Variable); | ||
| let ctx = if let GenericDefId::FunctionId(_) = def { | ||
| TyLoweringContext::new(db, &resolver, def.into()) | ||
| .with_impl_trait_mode(ImplTraitLoweringMode::Variable) |
There was a problem hiding this comment.
I wonder if we should just require the modes to specified in the new constructor. Feels like an easy to cause bug by relying on a default.
|
Thanks! |
|
☀️ Test successful - checks-actions |
Fixes #15412