rust-analyzer fails to resolve some of the bounds on traits with associated types

[klkvr]

rust-analyzer version: 0.4.2336-standalone (5e7dd31 2025-03-09)

rustc version: rustc 1.87.0-nightly (f4a216d28 2025-03-02)

editor or extension: VSCode 0.4.2336

repository link (if public, optional): https://github.com/paradigmxyz/reth

code snippet to reproduce:

trait Bar {
    type Baz;

    fn baz(&self) -> Self::Baz;
}

trait Foo {
    type Bar;

    fn bar(&self) -> Self::Bar;
}

trait FooFactory {
    type Output: Foo<Bar: Bar<Baz = u8>>;

    fn foo(&self) -> Self::Output;

    fn foo_rpit(&self) -> impl Foo<Bar: Bar<Baz = u8>>;
}

fn test1(foo: impl Foo<Bar: Bar<Baz = u8>>) {
    // baz: u8 inferred
    let baz = foo.bar().baz();
}

fn test2<T: FooFactory>(factory: T) {
    // type unknown
    let baz = factory.foo().bar().baz();
    let baz = factory.foo_rpit().bar().baz();
}

It seems that in some cases rust-analyzer is failing to resolve more complex bounds while simpler bounds involving the same types are getting resolved correctlty

[ShoyuVanilla]

We have so many inference failures around associated types and I think that migration to next-gen trait solver would fix most of them

[ChayimFriedman2]

So yes, this is chalk not able to represent shorthand associated type bounds (because of InlineBound). Switching to the next solver will fix that.

[ChayimFriedman2]

I created a new label for things that are blocked on the new trait solver.