---
title: 2023-11-06 Planning meeting
tags: weekly-meeting, T-types
date: 2023-11-06
discussion: https://rust-lang.zulipchat.com/#narrow/stream/326132-t-types.2Fmeetings/topic/2023-11-06.20planning.20meeting
url: https://hackmd.io/gw4UQcN8Rm-MGqFFF0M6jA
---

# T-types planning meeting agenda

## Updates

### [Tracking issue for shared type library](https://github.com/rust-lang/types-team/issues/16)

@compiler-errors has started to pick this up again, currently a few in progress PRs here. We're working towards using the new solver in r-a, still unclear how long this will take.

### [Tracking issue for chalk](https://github.com/rust-lang/types-team/issues/25)

Should probably close this :)

### [Tracking issue for RPITIT refactoring](https://github.com/rust-lang/types-team/issues/65)

@spastorino is this still happening? ... @spastorino: nope this is finished since a while :).

### [Tracking issue for AFIT and RPITIT](https://github.com/rust-lang/types-team/issues/66)

nikomatsakis: Stable on nightly! As far as I know.

TC: It's scheduled to go out in v1.75 alongside [#116733](https://github.com/rust-lang/rust/pull/116733).

### [Tracking issue for TAITs](https://github.com/rust-lang/types-team/issues/67)

nikomatsakis: Scheduled a [lang-team proposal this Wed, Nov 8](https://github.com/rust-lang/lang-team/issues/233).

TC: We're polishing the document for this.

### [Tracking issue for GATs](https://github.com/rust-lang/types-team/issues/68)

Jack has been digging in https://github.com/rust-lang/rust/issues/112764 and has concluded we should automatically coerce `Option<&'?0 mut (dyn Trait + '?0)` to `Option<&'?1 mut (dyn Trait + '?2>`, which is a subset of what we do for unsizing coercion (which would occur without the `Option`)

### [Tracking issue for a-mir-formality](https://github.com/rust-lang/types-team/issues/69)

nikomatsakis: I did some work on coherence and fuzzing but got quite distracted factoring out formality-core to be re-usable for other languages. Still have a few pending changes but things are much nicer now. You can for example write things like `impl Foo for Bar` instead of `impl<> Foo<> for Bar<> where []`. You can read some docs [on the new mdbook page](https://rust-lang.github.io/a-mir-formality/) -- this is still a WIP, I'm adding in some of the docs from our meetup and adapting them, and trying to write-up a `formality-eg` example that will show how things work for a simpler case.

~~**Need help:** I am still trying to figure out how to get the mdbook rendered to github pages! I think I need to pester somebody. Or perhaps check my GH notifications, which I've not done, and which may have information. =) I opened https://github.com/rust-lang/infra-team/issues/87. Oh, hey look, simulacrum fixed it! Yay.~~

### [Tracking issue for subtyping refactor](https://github.com/rust-lang/types-team/issues/70)

TODO(@lcnr): should we close this

### [Tracking issue for implied bounds refactor](https://github.com/rust-lang/types-team/issues/71)

TODO(@lcnr): should we close this, tracked further down again

### [Tracking issue for trait object upcasting](https://github.com/rust-lang/types-team/issues/72)

@compiler-errors started to handle the type system formalization here, getting closer to stabilization

### [Tracking issue for negative impls](https://github.com/rust-lang/types-team/issues/73)

We were meeting regularly Niko, Lcnr and spastorino but we put this project on-hold due to lack of time.
I (@spastorino) would be happy to resume at some point if Niko/Lcnr have also time for this.

### [Tracking issue for polonius](https://github.com/rust-lang/types-team/issues/74)

nikomatsakis: Amanda Stjerna is joining the discussions with me + lqd. Some work has landed. One thing we discussed was how to incorporate higher-ranked constraints. There is currently a piece of code that walks over the region constraints to construct the strongly connected components. I believe that can be adapted to also insert explicit `T: 'static` bounds equivalent to today's leak check and other processing, which would mean that other parts of the code don't have to think about universes at all. We are hoping Amanda can implement that as a "getting start" project. I promised a write-up to Amanda this week.

Our other goal is to get a nightly version of polonius that works (but is inefficient) this year. We talked it out a bit and it "seems doable" at a high-level I promised a blog post with some thoughts. Uh oh! Better get writing.

### [Tracking issue for rust trait solver refactor](https://github.com/rust-lang/types-team/issues/76)

work continues, dealing with aliases is horrible, currently experimenting with a different approach to `alias-relate`. The approach shows promise, however there are some fun(damental) issues which require some a-mir-formality cooperation.

### [Model coherence in formality](https://github.com/rust-lang/types-team/issues/103)

nikomatsakis: I landed a rewrite that brings formality closer to how compiler works and we reviewed in our last sync meeting. Next steps here are to sync up with student who is interested and also create various starter issues. I got distracted by parser improvements described above. Perhaps this week?

### [Normalize in orphan check](https://github.com/rust-lang/types-team/issues/104)

### [New solver in coherence](https://github.com/rust-lang/types-team/issues/105)

^ see above

### ["Semantic outlives" for opaque aliases](https://github.com/rust-lang/types-team/issues/106)

nikomatsakis: Eep! I had forgotten about this.

### [Minimal version of TAITs](https://github.com/rust-lang/types-team/issues/107)

I believe this has overlap with general TAITs above?

### [Prototype of Polonius](https://github.com/rust-lang/types-team/issues/108)

nikomatsakis: See above?

### [Dyn upcasting](https://github.com/rust-lang/types-team/issues/109)

^ duplicate

### [RPIT capture all lifetimes](https://github.com/rust-lang/types-team/issues/110)

TC: We've merged [RFC 3498](https://github.com/rust-lang/rfcs/pull/3498) and opened a [tracking issue](https://github.com/rust-lang/rust/issues/117587). CE has an [implementation](https://github.com/rust-lang/rust/pull/116952).

nikomatsakis: :tada: Does the new implementation show any kind of concerns?

TC/CE: The RFC specifies that RPITs will capture late-bound lifetime parameters in binders. This doesn't work yet. Previously these opaque types did not capture these parameters in `for<..>` binders. alliemjay suggests that we resolve:`

https://github.com/rust-lang/rust/issues/100503

### [Enforce where-clauses on type aliases via alias type](https://github.com/rust-lang/types-team/issues/111)

nikomatsakis: We scheduled a lang-team meeting related to this, I am on the hook to prepare an RFC for reading.

### [Use the new trait solver in all the places](https://github.com/rust-lang/types-team/issues/112)

### [Support negative impls in coherence](https://github.com/rust-lang/types-team/issues/113)

### [Location-sensitive polonius](https://github.com/rust-lang/types-team/issues/114)

### [UnlimiTAIT](https://github.com/rust-lang/types-team/issues/115)

TC: We need to stabilize Mini-TAIT first, then the new solver.

### [Coinduction](https://github.com/rust-lang/types-team/issues/116)

blocked on new solver

### [Instantiate implied bounds explicitly](https://github.com/rust-lang/types-team/issues/117)

blocked on coinduction

### [Perfect derive](https://github.com/rust-lang/types-team/issues/118)

blocked on coinduction

### [Specialization](https://github.com/rust-lang/types-team/issues/119)

blocked on the heat death of the universe

### [Avoid inference guessing in trait solver (e.g., favoring where-clauses over impls)](https://github.com/rust-lang/types-team/issues/120)

### [Avoid bounding impl trait return type by all input type parameters using existential lifetimes](https://github.com/rust-lang/types-team/issues/121)

### [Implied bounds on higher-ranked binders](https://github.com/rust-lang/types-team/issues/122)

new solver?

### [Full triage of all soundness issues](https://github.com/rust-lang/types-team/issues/123)

## Nominated issues

### [Boxed recursive function should not require trait object](https://github.com/rust-lang/rust/issues/73625)

Seems useful, but not a priority for the team. Michael might take a look, but otherwise @lcnr is going to unnominate.

### [`.await` does not perform autoref or autoderef](https://github.com/rust-lang/rust/issues/111546)

Seems like a T-lang thing for now - maybe T-types could help with evaluating breakage, but otherwise we don't feel the need to be involved.

### [Code no longer compiles after -Zdrop-tracking-mir was enabled by default](https://github.com/rust-lang/rust/issues/116242)

Fixed, just needs backport. Needs to be unnominated.

### [regression: change in async capture rules?](https://github.com/rust-lang/rust/issues/117059)

Don't have time to fix. Should keep investigating.

## Types FCPs

### [Tracking issue for dyn upcasting coercion](https://github.com/rust-lang/rust/issues/65991)

### [relax leak-check](https://github.com/rust-lang/rust/pull/112999)

### [generalize: handle occurs check failure in aliases](https://github.com/rust-lang/rust/pull/117088)

## Major change proposals

### [Add experimental support for implication predicates](https://github.com/rust-lang/types-team/issues/80)

### [Type system refactorings for further `rustc_type_ir`-ification](https://github.com/rust-lang/types-team/issues/124)

## Deep dive planning

### [Variance and Rust](https://github.com/rust-lang/types-team/issues/45)

### [discuss the `trait_alias` feature](https://github.com/rust-lang/types-team/issues/49)

### [Types team roadmap](https://github.com/rust-lang/types-team/issues/53)

### [Closure return type outlives guarantees](https://github.com/rust-lang/types-team/issues/57)

### [Account for late-bound lifetimes in generics #103448 ](https://github.com/rust-lang/types-team/issues/62)

### [Negative trait impls check-in](https://github.com/rust-lang/types-team/issues/82)

### [Coinductive trait semantics and normalization](https://github.com/rust-lang/types-team/issues/83)

### [MIR Formality progress](https://github.com/rust-lang/types-team/issues/86)

### [Specialization](https://github.com/rust-lang/types-team/issues/89)

### [Deep dive for leak check/higher ranked fn ptr subtyping](https://github.com/rust-lang/types-team/issues/96)

### [discuss "incorrect implied bounds in wfcheck"](https://github.com/rust-lang/types-team/issues/99)

### [dyn safety and coherence](https://github.com/rust-lang/types-team/issues/100)

### [`'erased` during analysis](https://github.com/rust-lang/types-team/issues/101)