Support invocation inference across narrowing reference conversions#8
Support invocation inference across narrowing reference conversions#8biboudis wants to merge 1 commit into
Conversation
| !targetType.hasTag(NONE) && | ||
| !targetType.hasTag(ERROR) && | ||
| targetType.isReference() && | ||
| !targetInferenceContext.free(targetType) && |
There was a problem hiding this comment.
Maybe this can be replaced by targetType.hasTag(CLASS)
There was a problem hiding this comment.
(e.g. the issue here is that a type variable might erase to a class type, which might lead you to lift more cases than needed. E.g. if we lift T to C<...> where T was an inference variable, we kind of lose T).
There was a problem hiding this comment.
not sure I follow, erasure during inference?
There was a problem hiding this comment.
It's the check in the line below -- which checks if |T| <: |R| where T is the target type and R is the return type. Basically, we're checking here if the relationship is "the wrong way around", and, if so, lift T to the same height as R. That's because with the new narrowing sealing conversion you can have legit cases of |T| <: |R| that are valid assignments.
| private Type resultCompatibilityTarget(Type returnType, Type targetType, | ||
| InferenceContext inferenceContext, InferenceContext targetInferenceContext) { | ||
| if (returnType.hasTag(CLASS) && | ||
| returnType.getTypeArguments().stream() |
There was a problem hiding this comment.
inferenceContext.free(returnType)?
| } | ||
|
|
||
| private Type lift(Type targetType, Type returnType) { | ||
| return types.asSuper(targetType, returnType.tsym); |
There was a problem hiding this comment.
what if we are dealing with arrays? better to use Infer::asSuper
| } | ||
| Type qtype = inferenceContext.asUndetVar(from); | ||
| Type to = resultInfo.pt; | ||
| Type to = resultCompatibilityTarget(mt.getReturnType(), resultInfo.pt, |
There was a problem hiding this comment.
probably the highest risk of semantic change is when qtype is an UndetVar, we need to test that case and look for any possible semantic change
3 participants
No description provided.