It's necessary to have the sane logic that FirMemberDeclaration is a subtype of FirDeclaration
The only kind of controversial change here is making FirAnonymousObject
also be FirMemberDeclaration, thus having its own declaration status
These two fields references the parent of the import rather than the
classes that are imported. For example
```
import java.util.Map // resolvedClassId -> null
import java.util.Map.Entry // resolvedClassId -> java.util.Map
import java.util.Map.* // resolvedClassId -> java.util.Map
import java.util.Map.someStaticMethod // resolvedClassId -> java.util.Map
import kotlin.package.someTopLevelFuntion // resolvedClassId -> null
import kotlin.package.MyObject.someObjectFuntion // resolvedClassId -> kotlin.package.MyObject
```
FirAnnotation represents simple annotation placed on some element.
This element contains only annotation type ref and simple argument
mapping
FirAnnotationCall is a element for call of annotation (`@Ann(args)`)
which was written in user code. FirAnnotationCall is a resolvable call
Also, add `classId` default value to the `FirErrorResolvedQualifierImpl`
just as in `FirResolvedQualifierImpl` - it seemed inconsistent that
we can have a class symbol in error-resolved qualifier, but the
`classId` field will be `null`
In order to make resolution still work for members not available from
`Nothing`, we track the type without `Nothing?` and use that for
resolution instead.
After removal of all diamonds in FirDeclaration hierarchy
FirCallableMemberDeclaration was only one inheritor of
FirCallableDeclaration, so there is no need to keep them both
The package directive is needed for KtElement -> FirElement mapping in IDE.
There should be some FirElement which will be result of mapping of elements
inside KtPackageDirective.
The fix is a bit hacky, but it's very simple. In addition, it still does
not handle the case where the receiver is a lambda function. But such
case seems to be fairly rare in practice.
As part of this change, we also extend the usage of RealVariable in more
places during DFA. Now mutable properties, property with custom getters,
delegated properties, etc are also treatd as a `RealVariable`. In
general this is needed in order to carry out smartcast computation in
order to report `SMARTCAST_IMPOSSIBLE`. It seems to also have side
effects that improves behavior of some test files.
Victor Petukhov