Don't Fold Self-Referential ADTs in check_inferred_predicates()

[veera-sivarajan]

Fixes #118163, #118449

The I-hang was caused by the compiler trying to instantiate a predicate for self-referential ADTs.

This PR fixes it by not instantiating a predicate when an ADT's field refers to the ADT itself.

[rustbot]

r? @BoxyUwU

rustbot has assigned @BoxyUwU.
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

[compiler-errors]

Can you please find what PR introduced this regression?

I'm also suspicious about if this is the right way to solve this problem -- there's no explanation of the side-effects of the change, nor any reasoning about if this affects soundness or anything like that. In general, I think this needs more explanation.

[veera-sivarajan]

Sure! The issue bisects to #53793.

Will analyze more about soundness and update with a detailed explanation.

[veera-sivarajan]

On second thought, this approach doesn't work for cases like:

trait Trait {
    type Assoc: Trait;
}

struct Outer<'a, T: Trait> {
    value: &'a Inner<'a, T>,
}

struct Inner<'a, T: Trait> {
    value: &'a Outer<'a, T::Assoc>,
}

Thanks for pointing it out.