Associated type position impl trait

[nikomatsakis]

Metadata
Owner(s)	@oli-obk
Team(s)	lang, types
Goal document	2024h2/ATPIT

Summary

Stable support for impl Trait in the values of associated types (aka "associated type position impl trait" or ATPIT)

Tasks and status

• [ ] Implementation (@oli-obk)
• [ ] FCP decisions (types )
• [ ] Stabilization decision (types, lang )

[nikomatsakis]

This issue is intended for status updates only.

For general questions or comments, please contact the owner(s) directly.

[oli-obk]

The remaining blockers for this feature are

• make a lang decision on whether methods with the ATPIT in their signature need to actually define the ATPIT (https://github.com/rust-lang/rust/pull/123046)
• Switch remaining DefineOpaqueTypes::No to Yes
  • horrible unreviewable PR: https://github.com/rust-lang/rust/pull/127034 instead try one of the following
    • try revealing opaque type in candidate assembly and bail as ambiguous while no concrete hidden type available https://github.com/oli-obk/rust/tree/infer instead (see https://rust-lang.zulipchat.com/#narrow/stream/144729-t-types/topic/opaque.20types.20and.20ATPIT.20oh.20my/near/451059084 for details)
    • or try normalizing opaques to infer vars eagerly https://github.com/rust-lang/rust/compare/master...oli-obk:infer2?expand=1 (may need AliasRelate predicate support in the old solver? not sure) Read https://hackmd.io/lXnLt7nsSS2NUFGHKW4B3Q to see what we can do to make old solver's opaques more similar to the new solver
  • unreviewed PR: https://github.com/rust-lang/rust/pull/126316
  • unopened PR: https://github.com/oli-obk/rust/pull/new/define_opaque_types22

Not necessary for stabilization, but would be cool if #[defines()] would make a method not define the opaque type, even if it is in the signature (see https://github.com/rust-lang/rust/pull/128440 for an impl of the #[defines] attribute)

[oli-obk]

No progress (that I know of) since last update