[ConstraintSystem] Fix a couple of issues related to generic specialization #66971
Conversation
…ric typealiases If type alias declaration doesn't add new generic parameters refer to its underlying type to find them.
…ete type
Diagnose attempts to specialize a concrete type or its alias:
```swift
struct Test {}
typealias X = Test
_ = X<Int>() // error
```
The situations where number of parameters and arguments didn't match.
…f generic arguments
```swift
struct Test<T, U> {}
_ = Test<Int>() // error
```
…valid number of arguments
|
@swift-ci please test |
|
@swift-ci please clean test macOS platform |
lib/Sema/CSSimplify.cpp
Outdated
| return nullptr; | ||
|
|
||
| auto genericParams = genericContext->getGenericParams(); | ||
| if (!genericParams || genericParams->size() == 0) { |
There was a problem hiding this comment.
I don't think you can have a non-null generic parameter list with zero parameters in it
There was a problem hiding this comment.
This was the original check but I can simplify it down to just !genericParams
lib/Sema/CSSimplify.cpp
Outdated
| if (auto *TA = dyn_cast<TypeAliasDecl>(decl)) { | ||
| if (TA->isGeneric()) | ||
| return nullptr; | ||
| if (auto underlying = TA->getUnderlyingType()->getAnyNominal()) |
There was a problem hiding this comment.
How about checking if TA->getUnderlyingType()-is<UnboundGenericType>()?
There was a problem hiding this comment.
Other places I saw check for isGeneric but I can change if that’s better
There was a problem hiding this comment.
@slavapestov I made it AnyGenericType to make it possible to properly diagnose cases like typealias Print = Printer<Any>, using UnboundGenericType won't work because we'd just return nullptr and following logic would just return SolutionKind::Solved and I don't want to mess with that in this PR.
|
@swift-ci please test |
Resolves: rdar://111239949