Don't substitute a GAT that has mismatched generics in OpaqueTypeCollector

[compiler-errors]

Fixes #111828

I didn't put up minimized UI tests for #112510 or #112873 because they'd minimize to literally the same code, but with different substs on the trait/impl. I don't think that warrants duplicate tests given the nature of the fix.

r? @oli-obk

---

Side-note: I checked, and this isn't fixed by #112652 -- I think we discussed whether or not that PR fixed it either intentionally or by accident. The code here isn't really touched by that PR either as far as I can tell?

Also, sorry, did some other drive-bys. Hope it doesn't make rebasing #112652 too difficult 😅

[oli-obk]

Future nit: please do such moves in separate commits

[oli-obk]

how does that happen?

[compiler-errors]

you can have something like -> <Self as Trait<NotIdentitySubsts...>>::Assoc, though I'm not sure if it ever results in anything but a cycle.

[compiler-errors]

Actually I think I can come up with an example where the old implementation results in surprising behavior.

[compiler-errors]

Yeah, here's an example that should error but does not:

#![feature(impl_trait_in_assoc_type)]

trait Foo<T> {
    type Assoc;

    fn test() -> u32;
}

struct DefinesOpaque;
impl Foo<DefinesOpaque> for () {
    type Assoc = impl Sized;

    fn test() -> <() as Foo<NoOpaques>>::Assoc {
        let _: <Self as Foo<DefinesOpaque>>::Assoc = "";
    
        1
    }
}

struct NoOpaques;
impl Foo<NoOpaques> for () {
    type Assoc = u32;

    fn test() -> u32 { 1 }
}

[oli-obk]

At this point of complexity, should we just figure out how to normalize?

I tried before, but it requires duplicating some of the normalization logic 🙃

Basically I'd want exactly what the normalizer does, but be able to hook into fold_ty

[compiler-errors]

Hm, I don't think there is a good way of doing it right now...

Do you have any examples where this current approach of manually substituting is not enough? Like code that isn't currently passing but we want it to?

[oli-obk]

Nope. I was just hoping to always go the normalization route like I need to with TAITs

[oli-obk]

@bors r+

[bors]

📌 Commit c4cd607 has been approved by oli-obk

It is now in the queue for this repository.