[QoI, Crasher] Don't include type variables in missing generic requirement match info #31869
Conversation
…t an unhelpful note for it. Fixes SR-12759
|
@swift-ci Please smoke test. |
lib/Sema/TypeCheckProtocol.cpp
Outdated
| @@ -784,6 +784,9 @@ static Optional<RequirementMatch> findMissingGenericRequirementForSolutionFix( | |||
| default: | |||
| return Optional<RequirementMatch>(); | |||
| } | |||
|
|
|||
| if (missingType->hasTypeVariable()) | |||
There was a problem hiding this comment.
Would it be possible to actually provide a solution here and use solution.simplifyType on missingType? That should make sure that there are no type variables.
There was a problem hiding this comment.
Solution associated with the fix considered that is.
There was a problem hiding this comment.
Will a damaged solution always substitute out every type variable?
There was a problem hiding this comment.
Yes, if there was a fixed solution produced it would not leave any type variables after substitution.
There was a problem hiding this comment.
Fantastic. Then it seems like we just have to substitute in each one of these little closures that act as shared return points and assert that we're not leaking type variables.
|
Note that, as written, this cuts off emitting these diagnostics altogether for a lot of common generic cases. These diagnostics are really valuable, and I don't think we should consider this solved until we can emit them again. |
…e taus to generic param names for emissions.
|
Makes sense! @swift-ci Please smoke test. |
|
@swift-ci ASAN test |
|
⛵ |
If the missing generic requirement includes type variables, don't emit an unhelpful note for it.
Resolves SR-12759