fix APITIT being treated as a normal generic parameter in suggestions

[Qelxiros]

closes #126395

[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

[BoxyUwU]

I think this function fundamentally works the wrong way (not ur fault, it's prexisting).

It seems to:

1. Iterate over the list of where clauses on an item
2. Look at where clauses that are either Type<...>: Trait<..> or Type<...>: Trait<..., Assoc<...> = ...>
3. Collect a map (types) of all those Type<...> to the trait on the rhs of the :. e.g. if we have where u32: Trait, u64: OtherTrait, T: Trait + OtherTrait, we'd wind up with [u32 => Trait, u64 => OtherTrait, T => { Trait, OtherTrait }]
4. We then iterate that map filtering to only keys that are type parameters, and use that to reconstruct the list of generic parameters defined on the item.

this is completely incoherent. it assumes multiple things that aren't true:

• Type parameters are the only kind of generic parameters. You can't have const N: usize as the self type of a trait bound, so it never winds up in types so we'll never wind up putting const generics in the final generics string. playground (run test)
• Type parameters are always used in the self type of a trait bound. This was not true on stable until very recently when we made T: ?Sized be sugar for T: MetaSized. You can still observe this with the rustc_no_implicit_bounds attr: playground (run test). Also this means that right now T: ?Sized on stable gets re-sugared as T: MetaSized which is wrong

---

Instead, we should unconditionally accept an AssocItem so that we can always call generics_of to get the list of generic parameters defined on the item.

You can then independently construct the types_str by iterating over the generic parameters from generics_of and fetching any bounds on them from the types map.

The existing for loop over types can be left mostly as-is, just removing the parts of it that mutate types_str.

View changes since this review

[Qelxiros]

@rustbot ready

[BoxyUwU]

@bors r+ rollup

thanks :3 if you want to fix some of the other diagnostic bugs in this fn (ignoring non-type parameters, or us resugaring T: ?Sized as T: MetaSized) then feel free to r? me on a PR doing that

[bors]

📌 Commit 2f0c103 has been approved by BoxyUwU

It is now in the queue for this repository.

[rustbot]

Failed to set assignee to me: invalid assignee

Note: Only org members with at least the repository "read" role, users with write permissions, or people who have commented on the PR may be assigned.

[Zalathar]

Bors, this has already been merged.

@bors r-