This cache wasn't working too well, because ConeTypes have
equals/hashCode implementations which do not account for types'
attributes (like annotations, for example).
Because of that, `String` and `@Annotated String` were considered the
same type, and the cache was remembering whichever came first.
This is a temporary fix to avoid crashes in FIR IDE
Of all stub types, only `ConeStubTypeForChainInference` is handled now,
so we can catch the other possible errors with stub types (in the
delegates' inference, for example)
The fix introduced in this commit should be reworked when KT-50916 is
fixed
^KTIJ-20826 Fixed
This commit fixes DFA problem (see test) when we accidentally try
to merge incoming flows from member property (val ... by cached ...) and
the following function. While completing lambda of buildList { ... }
inside by cached we accidentally assume that delegate is already left
and add the lambda into exitsFromCompletedPostponedAnonymousFunction.
This commit fixes the problem by exiting delegate later, after completion.
This big refactoring is needed to cleanup building of overrides
mappings and prevent creating redundant intersection overrides in
cases when there is no need in them:
```kotlin
interface A {
fun foo()
}
interface B {
fun foo()
}
interface C : A, B {
override fun foo()
}
```
Before this refactoring there was next override tree:
C.foo
intersection override (A.foo, B.foo)
A.foo
B.foo
Also this commit fixes special mapping of overrides in jvm scopes
for declarations which have kotlin builtins in supertypes with
special java mapping rules (collections, for example)
If there is a `coneType` call immediately after the `fir.bounds` call,
it means that the fully resolved type is expected, hence
`resolvedBounds` should be used
When stdlib is represented as a source dependency (which seems to be the
case in the `kotlin` project), it will be resolved lazily for the IDE.
Same thing can happen if someone decides to "extend" the stdlib
by declaring their declaration in the `kotlin` package
In both of those cases, we need to make sure that receiver type is fully
resolved before trying to get a `coneType` from it
N.B. To make resolve tests work, I've added a separate folder
`withAllowedKotlinPackage` to the `testData`, because in the default
test setup it is not allowed to extend the `kotlin` package by user's
definitions
^KT-50728 Fixed
Mostly, it only affects FIR
It partially allows to consider several variance of constraints like
A<Int> & A<T> <: A<X_var> that are mostly brought by smart casts
^KT-49542 Fixed
^KT-50489 Relates
Mikhail Glukhikh