Commit Graph

9 Commits

Author SHA1 Message Date
Arya Irani
40ec9404b5 fix fix1696 and fix2297 transcripts
these had broken in the past when we made the modifier mandatory
and nobody noticed until now
2024-01-05 14:58:47 -10:00
andrii
4456c95c7a Fixing test for type modifiers, updating trnscripts 2024-01-04 09:42:12 +01:00
andrii
8b857b152c Adding a message for when ucm started reloading changes 2023-12-22 12:55:24 +01:00
Rebecca Mark
e0a8a832bf updates transcripts 2022-03-07 14:11:10 -08:00
rlmark
79ad0fa8df updates transcripts 2021-08-23 14:05:37 -07:00
Dan Doel
4c07479f0c Keep track of sources for wanted abilities
- Used for more precise type errors
2021-06-15 18:20:56 -04:00
Dan Doel
5d80a8ddb2 Minimal change for improved ability inference
- This commit delays making up inference variables until the
  outermost universal quantifiers of a type annotation have
  been unwrapped and added to the context. However, failures
  can still be generated by putting quantifiers inside other
  connectives.
2020-12-03 10:23:33 -05:00
Dan Doel
2699b6c470 Pre-calculate abilities involved in pattern matching
- Previously, the type checker was not using the patterns to inform
  which abilities are matched on in cases at all.
- Instead, we walk through all patterns and compile a list of all
  abilities involved in matching, then perform a subtype check on
  the scrutinee type with a type consisting of all the abilities.
- This strategy seems to be more robust than trying to handle this
  case-by-case. In such scenarios, recursion can lead to additional
  effects being pushed into the `Request` type that are not informed by
  the pattern matching.
- This causes the 1696 transcript to fail appropriately (I believe).
2020-11-17 17:03:28 -05:00
Paul Chiusano
a07011f1de Adding a test case that fails in the wrong way 2020-11-11 12:33:46 -05:00