FunctionalTypeKind can be used in FE 1.0 too, so there is no need to
keep both classes. Also, removal of FunctionClassKind simplifies work
with FunctionalTypeKind in common code, like Analysis Api
This is needed because now set of available synthetic functional interfaces
depends on session, because each module can contain different plugins,
which provide different kinds of functional types
This is needed to provide an ability to extend different kinds of
functional types
Also, cleanup and rename utilities related to functional types to avoid
possible confusions
The main idea is pre-computing the sets of names that might be
met there, that helps to decrease the sizes of the backing maps
(by avoiding irrelevant keys)
Totally, this branch with previous commits speeds up MT Full Kotlin
approximately on 3 seconds (~5%)
This code may be called during status resolution, but
it can also be called during `COMPILER_REQUIRED_ANNOTATIONS`.
In the latter case we can't expand typealiases.
If we do try to expand typealiases if we can,
and leave them unexpanded otherwise, then we
still may have a problem. Consider the
FirBlackBoxCodegenTestGenerated.Reflection.Annotations.Repeatable
.testKotlinAnnotation test.
In this test we get the following cycle:
```
import kotlin.annotation.AnnotationTarget.* // <- requires TYPEALIAS
@SinceKotlin("1.1") // <- requires String
AnnotationTarget.TYPEALIAS
@IntrinsicConstEvaluation // <- fully expanding this requires
fun String.compareTo() // building it
@Target(AnnotationTarget.CONSTRUCTOR) // <- requires
annotation class IntrinsicConstEvaluation // AnnotationTarget
// in general
// The cycle repeats
```
Or in words if the picture isn't clear:
- Transform `import kotlin.annotation.AnnotationTarget.*`
- Find and deserialize class `kotlin/annotation/AnnotationTarget`
- Load the `TYPEALIAS` field
- Load its annotation `@SinceKotlin("1.1")`
- Find and deserialize `kotlin/String` for `"1.1"`
- Load `String.compareTo`
- Load its annotation `kotlin/internal/IntrinsicConstEvaluation`
>>> At this moment the `extractDeprecationAnnotationInfoPerUseSite`
function is called, and ideally we'd want to fullyExpand
`IntrinsicConstEvaluation`
- While calling `toAnnotationClassId` for it, we find and deserialize
`IntrinsicConstEvaluation`
- `IntrinsicConstEvaluation` has `@Target`
- At this moment we find and deserialize `AnnotationTarget` again.
But This time loading `@SinceKotlin("1.1")` succeeds,
because when loading `kotlin/String` we first create the symbol, then
cache it, then call `postCompute`.
It does not happen with `AnnotationTarget`, because it calls
`findAndDeserializeClass` during `createValue`,
not `postCompute`.
- Thus, we have duplicate `AnnotationClass`-es, and they fail in fir2ir
with "symbol is already bound"
We can't move `findAndDeserializeClass` into `postCompute`, because
during this process when we try to access the
already created and cached symbol for `AnnotationTarget`
(with yet empty fir) we are trying to build the declared member scope
by this `symbol.fir`.
`FirDeclarationGenerationExtension.generateClassLikeDeclaration` was split
into two functions: one for generating top level classes, and one for
nested classes. Such change reduces verbosity and error-proness of
this extension and also allows to smoothly run plugins on local classes
^KT-55248 Fixed
No clue where to put the test for that since diagnostic tests, even
multi-module ones, never touch the metadata serializer. So the test is
a bytecode text one pretending to be about nullability annotations even
though it also affects what resolution in another module will do.
Dmitriy Novozhilov