rework min_choice algorithm of member constraints

[aliemjay]

See this comment for the description of the new algorithm.

Fixes #63033
Fixes #104639

This uses a more general algorithm than #89056 that doesn't treat 'static as a special case. It thus accepts more code. For example:

async fn test2<'s>(_: &'s u8, _: &'_ &'s u8, _: &'_ &'s u8) {}

I claim it's more correct as well because it fixes #104639.

cc @nikomatsakis @lqd @tmandry @eholk @chenyukang @oli-obk

r? types

[compiler-errors]

r? types

[bors]

☔ The latest upstream changes (presumably #104765) made this pull request unmergeable. Please resolve the merge conflicts.

[c410-f3r]

Nice! Looking forward to see #63033 fixed.

[oli-obk]

could you rebase and bless? I'll start a T-types FCP then

[oli-obk]

@rfcbot fcp merge. This PR (similarly, and building on top of #104765) allows RPITs and async fns to untangle their lifetime bounds to figure out the set of lifetimes that actually get caputed instead of falling back to a larger one or even 'static. For most cases it was was already possible for users to manually reorder their lifetimes to make their code compile, this PR just makes that automatic, as the order of generic params and bounds should not matter.

[aliemjay]

This PR modifies the minimal choice algorithm described here. The goal of this algorithm is to pick up the minimum of a set of universal regions related by universal_region_relations. Because universal_region_relations is not a total order relation, a unique minimum choice may not always exist. An example of this is shown in the inline comment where ['static, 'a, 'b, 'c, 'd, 'e] are related by this graph (transitive edges were removed for simplicity):

In this example, the minimal choice is ambiguous with two candidates ['d, 'e]. We used to bail out in this case (#63033). This PR however picks up the least upper bound of the two candidates LUB(['d, 'e]), which happens to be 'c.

Forward compatibility

In the future we may want to use instead a new region that represents the greatest lower bound of the candidates ReGLB(['d, 'e]) (the intersection regions proposal, if I understand it correctly), but this would still be backwards compatible with this PR because we would infer a lower region value GLB(['d, 'e]) <= LUB(['d, 'e]) (at least this is my understanding of region inference).

See test/ui/nll/member-constraints/min-choice-reject-ambiguous.rs for a test that should pass with intersection regions proposal.

Implementation

TODO: explain the implementation

[compiler-errors]

@rfcbot fcp merge

oli's FCP didn't start -- copied from that:

This PR (similarly, and building on top of #104765) allows RPITs and async fns to untangle their lifetime bounds to figure out the set of lifetimes that actually get caputed instead of falling back to a larger one or even 'static. For most cases it was was already possible for users to manually reorder their lifetimes to make their code compile, this PR just makes that automatic, as the order of generic params and bounds should not matter.

[rfcbot]

Team member @compiler-errors has proposed to merge this. The next step is review by the rest of the tagged team members:

• @compiler-errors
• @jackh726
• @lcnr
• @nikomatsakis
• @oli-obk
• @spastorino

No concerns currently listed.

Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

See this document for info about what commands tagged team members can give me.

[Swatinem]

Out of curiosity, I tried this PR on top of #104833 and it solves the bogus lifetime errors I see:

hidden type for impl Future<Output = ...> captures lifetime that does not appear in bounds

Looks like identity_future is an example of point 2 in #63033:

Seemingly insignificant tweaks, like adding a wrapper struct, make it compile.

[oli-obk]

@bors delegate+

r=me once the FCP finishes (@bors r=oli-obk is the invocation needed to approve in my name). Here's me 🤞-ing that bors doesn't pick up the command in backticks

[bors]

✌️ @aliemjay can now approve this pull request

[rfcbot]

The final comment period, with a disposition to merge, as per the review above, is now complete.

As the automated representative of the governance process, I would like to thank the author for their work and everyone else who contributed.

This will be merged soon.

[aliemjay]

@bors r=oli-obk

[bors]

📌 Commit 837c1af has been approved by oli-obk

It is now in the queue for this repository.