regression: concrete type differs from previous defining opaque type use

[Mark-Simulacrum]

[INFO] [stdout] error: concrete type differs from previous defining opaque type use
[INFO] [stdout]    --> src/execution.rs:222:13
[INFO] [stdout]     |
[INFO] [stdout] 222 |             impl Iterator<Item = Node<'tree>> + 's,
[INFO] [stdout]     |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[INFO] [stdout]     |             |
[INFO] [stdout]     |             expected `impl Iterator<Item = tree_sitter::Node<'tree>> + 'tree`, got `impl Iterator<Item = tree_sitter::Node<'s>> + 's`
[INFO] [stdout]     |             this expression supplies two conflicting concrete types for the same opaque type

https://crater-reports.s3.amazonaws.com/beta-1.73-1.2/beta-2023-09-10/reg/tree-sitter-graph-0.11.0/log.txt

[apiraino]

this regressed in 0dd5730:

• Check for <&NotClone as Clone>::clone() calls and suggest to add Clone trait appropriately #112995 (Check for <&NotClone as Clone>::clone() calls and suggest to add Clone trait appropriately)
• Don't say that a type is uncallable if its fn signature has errors in it #113578 (Don't say that a type is uncallable if its fn signature has errors in it)
• Double check that hidden types match the expected hidden type #113661 (Double check that hidden types match the expected hidden type)
• factor out more stable impls #114044 (factor out more stable impls)
• CI: split nested GHA groups instead of panicking #114062 (CI: split nested GHA groups instead of panicking)

The exact commit is not available anymore (0ae3a8e2ad80d9e375cb8f391bfb0e5e57d10515) but by reading the tests of #113578 I think this is where it originated and (if I understand) is the intended behaviour i.e. introduce an error to fix an unsoundness (cc @oli-obk )

[apiraino]

WG-prioritization assigning priority (Zulip discussion).

@rustbot label -I-prioritize +P-medium

[oli-obk]

hmm... that PR was only supposed to fix a type_alias_impl_trait unsoundness, not have any effect on stable crates and RPIT. I'll investigate