Backport "Remove unnecessary and recursive Space decomposition" to LTS #20804
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
... makes it easier to disable one of them. [Cherry-picked dbb97ca]
Space decomposition recently learnt to decompose prefixes. Given a nested definition like in i19031, aggressively trying to decompose while intersecting can lead to recursive decompositions (building bigger and bigger nested prefixes). Turns out the decomposition isn't necessary. [Cherry-picked 5439e98]
If refineUsingParent/instantiateToSubType is passed a HK then it's not possible to instantiate a class to that type (at the moment and perhaps ever). So it's important we guard against that. This came up while trying to see if Mark[?] and Foo[Int] (from pos/i19031.ci-reg1.scala) are provably disjoint - which they should be reported not to be. Because they're not applied types of the same type constructor we end up trying to prove that HK type Mark is disjoint from HK type Foo. Because we don't know how to instantiate Foo's subclasses (e.g Bar2) such that it's a subtype of higher-kinded type "Mark", we end up discarding all of Foo's subclasses, which implies that Foo & Mark is uninhabited, thus they are provably disjoint - which is incorrect. We originally didn't encounter this because we eagerly decomposed in Space intersection, while now we've dispatched it to provablyDisjoint. (edit) We allow for some kindness in provablyDisjoint. [Cherry-picked 7e47294]
Minimising from the `case test.Generic =>` in ParallelTesting, the anonymous pattern match is expanded, wrapping the match with `applyOrElse`, which has a type parameter A1 as the scrutinee type, with an upper bound of the original element type (out.Foo for us). During reachability analysis the pattern type, e.g. out.Bar3.type, is intersected with the scrutinee type, A1 - giving out.Bar3.type & A1. Then that we attempt to decompose that type. Previously the abstract A1 in that type lead to 3 WildcardTypes, for the 3 subclasses, which are a subtype of previous cases. The fix that by generalising how we recognise the singleton types in the scrutinee type, so instead of the ownership chain we use the parameter type info, and we also match term parameters. For extra correctness we consider the failure to be a subtype of a mixin as a failure for instantiating. Also, make sure to handle and avoid recursion in traverseTp2. [Cherry-picked 0931c43]
|
No regressions detected in the community build up to lts-19356. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Backports #19216 to the LTS branch.
PR submitted by the release tooling.
[skip ci]