compiler: report error when trait object type param reference self

[xtexx]

Fixes #139082.

Emits an error when Self is found in the projection bounds of a trait
object. In type aliases, Self has no meaning, so type A = &'static dyn B where trait B = Fn() -> Self will expands to type A = &'static Fn() -> Self which is illegal, causing the region solver to bail out
when hitting the uninferred Self.

r? @compiler-errors @fee1-dead

[rustbot]

HIR ty lowering was modified

cc @fmease

[compiler-errors]

Please add a UI test.

Also, there's no explanation for why this fixes the issue? Specifically, I'm not convinced this is the right fix for the issue yet, since it doesn't really explain why this doesn't ICE for small permutations of the original reproducer like:

trait B<C> = Fn() -> Self;
type D<'a> = Vec<&'a dyn B<()>>;

And why this ends up fixing an ICE in the region solving code.

@rustbot author

[compiler-errors]

I thought more about this, and I think this may be the right fix, though currently it duplicates some code. Please add a test still, and it would be nice to add some sort of explanation about why this is necessary in the PR.

[xtexx]

type D<'a> = Vec<&'a dyn B<()>>; is not a lazy type alias thus does not trigger the well-formed check (

rust/compiler/rustc_hir_analysis/src/check/wfcheck.rs

Line 301 in 2196aff

if tcx.type_alias_is_lazy(item.owner_id) =>

, wfcheck::check_item -> check_where_clauses -> implicit infer, infer_predicates)

However, when the type D is actually used, it leads to a ICE, as the type inference has been called.

trait B<C> = Fn() -> Self;
type D = &'static dyn B<()>;

fn main() {
    let a = D::new();
    _ = a;
}

(note that, the ICE from the above code is not the one in the issue but another one, compiler/rustc_infer/src/infer/canonical/canonicalizer.rs:404:17: encountered a fresh type during canonicalization, but they are both due to not-inferred types passed down)

( I am not sure if Self here should be allowed or rejected. )

[xtexx]

@rustbot review

[compiler-errors]

One last requested change, sorry for the back and forth

@rustbot author

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[xtexx]

@rustbot ready

[compiler-errors]

@bors r+ rollup

[bors]

📌 Commit fdefffe has been approved by compiler-errors

It is now in the queue for this repository.