Fix ret > 1 bound if shadowed by const

[JulianKnodt]

Prior to a change, it would only look at types in bounds. When it started looking for consts,
shadowing type variables with a const would cause an ICE, so now defer looking at consts only if
there are no types present.

cc @compiler-errors
Should Fix #93553

[rust-highfive]

r? @matthewjasper

(rust-highfive has picked a reviewer for you, use r? to override)

[JulianKnodt]

This is the main fix, the rest should make it more carefully stick to logic prior associated-const equality change which introduced the ice.

[compiler-errors]

Can you verify that the compiler doesn't ICE in situations like:

trait Baz {
  const Bar: usize;
  const Qux: Self::Bar;
}

[JulianKnodt]

I don't think it should ice in that case since there's no shadowing and no name overlaps, but will add a test in

[JulianKnodt]

oh it does ice, I'll look into it a bit further

[compiler-errors]

I think the stack trace points to an .expect so you can find where it ICEs. I think it's in a method that resolves a path to a ty, maybe you want to check if it resolves to a const instead and emit an error. Maybe something like this:

trait Baz {
  const Bar: usize;
  const Qux: Self::Bar;
//           ^ Expected type, found const
}

[JulianKnodt]

yuh, i'm not sure if the best way to fix it is to turn the expect into an actual error, or handle it earlier, for now I can just turn it into an error and if the reviewer wants to do the other I'll update

[oli-obk]

r? @oli-obk

[oli-obk]

@bors r+

[bors]

📌 Commit 2dfd77d has been approved by oli-obk