Fix #6314: match type unsoundness #6319
Merged
Conversation
This file contains hidden or 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
OlivierBlanvillain force-pushed the fix-6314 branch 2 times, most recently from 249af27 to ad3b0d0 Compare Seams to be a leftover from scalac old pattern matcher and has always been dead code in dotty.
Singleton is not part of the normal type hierarchy, special treatment in requiered in intersecting to avoid making incorrect assumptions.
It timed out on my machine...
This breaks tests/run/typeclass-derivation2c.scala There is no way to statically know that the type passed to constValue reduces.
OlivierBlanvillain force-pushed the fix-6314 branch from ad3b0d0 to 0cec095 Compare OlivierBlanvillain changed the title 
odersky requested changes May 7, 2019
SeklomType are already mapped over correctly, that is, as .info.
OlivierBlanvillain force-pushed the fix-6314 branch from a0819cd to 74ca923 Compare Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
2 participants
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 Outdated suggestions cannot be applied.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.
This PR fixes several match issues
The first issue was that the algorithm assumes types in an
AndTypealways came in the same order, so it wrongly concluded thatX & YandY & Typeare non intersecting from the fact thatXandYare non intersecting. The updated logic checks both permutations, and simply get stuck whenTypeis abstract, which fixes the unsoundness.The second issue is related to the
Singleton. This marker type needs to be taken it into account by the intersection algorithm, see the added test case.Finally the previous handling of abstract type was problematic a showed by several of the examples in #6314. The logic is updated to do a sub-type check where all the abstract types (members or parameters) are replace by
WildcardType. If the scrutinee is not a subtype of a pattern, but it is with wildcards, then the type shouldn't reduce.