assumptions_on_binders: Test mixed ambiguous ORs

[Dnreikronos]

Follow up to #162442. Boxy pointed out that one got superseded by #162935, and yeah, once adwin's PR removes propagate_ambiguity there's nothing left for my code change to do. The test still covers a case nothing else does though, so this just moves the test over.

The idea is an OR where one side is an alias outlives with a non-lifetime binder, which ends up ambiguous, and the other side is something we can prove from the root assumptions. That should hold no matter which order the OR is in. I also kept the cases that should still error, an AND where the ambiguous part is still required and an OR where every side is ambiguous. adwin mentioned in #162935 they weren't sure what to do with the non-lifetime binder ambiguity yet, so I think it's nice to have something pinning down how it behaves today.

cc @BoxyUwU

[rustbot]

r? @mejrs

rustbot has assigned @mejrs.
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 77 candidates
• Random selection from 20 candidates

[mejrs]

r? @adwinwhite

[rustbot]

adwinwhite is currently at their maximum review capacity.
They may take a while to respond.

[mejrs]

oops sorry missed that it was assigned already

r? boxyuwu

[BoxyUwU]

:3

[BoxyUwU]

with one of @khyperia's PRs it should now be possible to just write ambiguity instead of using an alias outlives which indirectly gets rewritten to ambiguity. Can you rebase and then remove the non_lifetime_binder dependency from the tests?

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[Dnreikronos]

with one of @khyperia's PRs it should now be possible to just write ambiguity instead of using an alias outlives which indirectly gets rewritten to ambiguity. Can you rebase and then remove the non_lifetime_binder dependency from the tests?

yeah, rebased and switched the ambiguous branches to ambiguity. Dropped the extra bound type parameters and non_lifetime_binders too.

The old setup got ambiguity from an alias whose type placeholder couldn't leave the binder. Now that we can write ambiguity directly, the test can skip that whole path. I like this version more because you can see the ambiguous branch right away without having to follow the alias rewrite first.

I kept the other alias branch because it still needs to survive until the root, where the outlives assumption can prove it. Both OR orders are still covered. The case without that assumption still errors, along with the AND that requires ambiguity and the OR where every branch is ambiguous.