trait solver: Include implied outlives assumptions

[Dnreikronos]

Part of rust-lang/project-assumptions-on-binders#19

Split out of #161988 after @BoxyUwU pointed out that these are about which assumptions we keep, not really about reflexive region constraints.

I went back through where each piece comes from and found two gaps. Inside a binder we kept Ty: 'a, but the region relation only knew about explicit region clauses. That means something like &'b T: 'a did not also give us 'b: 'a. At the root it was a slightly different version of the same problem: known_type_outlives has the explicit where clauses, while implied bounds from things like &'b self live in region_bound_pairs, so constraint destructuring never saw them.

Assumptions::new now pulls the free region components out of type outlives clauses and adds those edges to the region relation. I think doing it there is the cleanest spot. All callers get the same view of an assumption, and the original type clauses stay around for placeholder and alias cases. Regions bound inside the type are ignored because they do not name anything we can use outside that binder.

The root path now adds its implied type bounds to the same assumption set before destructuring. The regression uses an implied I: 'b from a receiver and a separate 'b: 'a relation, so it covers this without leaning on the reflexive fix from #161988. There are also binder checks for a reference and a higher-ranked function type. Those caught an easy testing trap here: a green direct constraint could have depended on the other PR, so the checks look at the lifted candidates instead.

Personally, I think splitting this was the right call. It is really a change to how assumption data is built, and that is easier to reason about on its own than under the reflexive constraint fix.

cc @BoxyUwU, this is the pair of changes you asked me to pull out.

[rustbot]

r? @jackh726

rustbot has assigned @jackh726.
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

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 75 candidates
• Random selection from 20 candidates

[BoxyUwU]

r? me

[Dnreikronos]

Pushed some more changes after staring at this a bit longer, quick summary of what I did and why.

Borrowck had the same gap I fixed at the regionck root. Its known_type_outlives_obligations only has the explicit where clauses too, and the implied bounds sit right next to it in region_bound_pairs, so now those get passed in as well. While doing that I noticed GenericKind::to_ty already does what my hand written match did, so both paths now share a small helper that uses it. I folded this into the existing root commit since it's really the same fix, which is why the force push.

I couldn't come up with a test for the borrowck half. As far as I can tell borrowck still gets an empty solver constraint tree, so any test passes with or without this. Once the type-op constraints from #161423 reach borrowck it becomes testable and I'd add the regression over there. If you'd rather I drop this part until then, that's fine too.

The last commit is the one I'm not sure belongs here, so I kept it separate and easy to drop. The free region map stores its edges as 'sub <= 'sup, while the relation in Assumptions gets read the other way, 'longer: 'shorter. That mismatch is older than this PR and nothing notices it today because the root only ever looks at type_outlives. But Assumptions::new now merges the edges it derives from type outlives clauses into that same relation, so the regionck path would end up with edges pointing both ways in one relation. Felt wrong to leave that in once I'd seen it, so I invert the edges before building the assumptions and wrote down the expected direction on the field. No test here for the same reason as above, nothing at the root reads the relation yet.

Small thing I noticed while writing comments is that the destructuring in Assumptions::new is redundant for the solver path, since that one already elaborates before building the relation. It still matters for the callers that build assumptions straight from where clauses, the test harness and the root paths. I'd rather have one place that guarantees the shape than trust every caller to remember to elaborate, but happy to move it into the harness if you prefer.

cc @BoxyUwU :)

[BoxyUwU]

Hmm okay this all makes sense to me 🤔 I think we should definitely wait for #161423 so that the borrowck path can actually be tested.

I'd rather have one place that guarantees the shape than trust every caller to remember to elaborate

this makes complete sense to me 🤔 can you remove the elaboration logic from the solver's logic for computing assumptions and instead rely on Assumptions::new to do it.

The other thing is that I'd actually prefer for the test suite to not rely on elaboration and instead hand write out all of the relevant requirements. When I'm looking at the custom test suite DSL I want to be able to see exactly what assumptions we have and not worry about there being "extra" ones hidden behind the scenes.

Could you make it so that we have an Assumptions::new_unelaborated or something and use that from the test suite :3

[BoxyUwU]

We should probably not be even letting you write such where clauses on forall since they don't do anything without elaboration (and I don't think we want to elaborate these).

cc rust-lang/project-assumptions-on-binders#39

View changes since the review

[BoxyUwU]

why not use elaborate directly here?

View changes since the review

[BoxyUwU]

Do we actually use region outlives assumptions in the root context like ever? I would have expect this to be more blatantly broken if we did, but I guess the only real root handling we do is destructuring type outlives constraints which doesn't care about region outlives assumptions 🤔

View changes since the review