These declarations should not be visible to users (and therefore are not added to FIR),
but plugin itself can reference them in already compiled serializable classes,
and therefore they should be available in metadata:
- synthetic deserialization constructor
- static write$Self function
See also:
^KT-55885
- Add IrPluginContext to JvmBackendContext, so plugin intrinsics can
reference external functions properly.
- Do not use module.findClassAcrossModuleDependencies as Descriptor API does not work for FIR.
- Add asm listing tests in serialization plugin for K2
- Remove Delegated.kt asm listing test as we have similar test in boxIr group.
#KT-56553 Fixed
Because SerializationJvmIrIntrinsicSupport does not instantiate annotations yet,
this info could be lost. As a workaround, it is possible to call Companion.serializer()
functions instead of direct serializer instantiation, as they are plugin-generated
and correctly handle annotations.
Note that for some cases (enums & interfaces) this WA is not enough, so additional work
is needed later.
Fixes https://github.com/Kotlin/kotlinx.serialization/issues/2179
`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
as it may have no primary constructor (only actual class has it).
Refactor & rename Fir2IrGeneratorUtils.kt as having 'Fir' in stacktrace
when actual K2 compiler is not used is very confusing.
Fixes https://github.com/Kotlin/kotlinx.serialization/issues/2134
#KT-55683 Fixed
Name check was forgotten during refactoring, so any user function with
matching signature could be overwritten by the plugin, while correct
function would not have body.
#KT-55682 Fixed
of the class used in @UseSerializers: Use methods to receive
full list of supertypes.
K1 supertypes() call returned all supertypes, while
IrClass.supertypes and FirClassSymbol.resolvedSuperTypes return only immediate ones.
This lead to a difference in behavior between K1 and K2, and regression
after plugin backend was rewritten from descriptors to IR.
#KT-55340 Fixed
Now annotations are resolved with following algorithm:
1. On COMPILER_REQUIRED_ANNOTATIONS we resolve all annotations
and store results if this is compiler annotation, plugin annotation,
or annotation with meta-annotation (meta annotations are checked
recursively with designated resolution if needed)
2. On TYPES stage we resolve all those annotations once again and if
some annotation changes resolution then we keep type from p.1 and
report error on this annotation, so user should disambiguate it
Ambiguity may occur because of nested annotations with same name as
plugin annotations:
```
annotation class SomeAnnotation // (1) plugin annotation
open class Base {
annotation class SomeAnnotation // (2)
}
class Derived : Base() {
@SomeAnnotation // <-----------------
class Inner
}
```
At COMPILER_REQUIRED_ANNOTATIONS annotation call will be resolved to (1)
because at this stage supertypes are not resolved yet, and we consider
only importing scopes. At the TYPES stage we will find correct
annotation from supertype
K2 plugin API has a limitation that meta annotations from plugin predicates
can't be used on nested annotations because their resolve process includes supertypes
resolve, that can be affected by the plugin itself. Therefore, @MetaSerializable
can't be applied to nested annotation classes in K2, which is reflected by
this diagnostic.
For old FE, diagnostic is lowered to WARNING with deprecation message.
- IR plugin does not use it anymore,
regular annotation instantiation feature is used
(insertion of IrConstructorCall(annotationCtorSymbol)).
- IR plugin still generates `SerialInfo$Impl` class to be compatible
with previously compiled code & libraries.
- Custom implementation over descriptors in IR plugin is removed;
plugin now delegates to the existing lowering with some tweaks.
- $Impl descriptor is removed from FE 1.0 because it is no longer needed
for IR plugin; it shouldn't be exposed to users so FIR doesn't need it either.
- Since language lowering is now used, it is possible to correctly instantiate
SerialInfo annotations with default values even if they are from other modules
and even if they were not processed by the plugin.
#KT-48733 Fixed
Fixes https://github.com/Kotlin/kotlinx.serialization/issues/1574
#KT-54297 Fixed
Properly substitute surrogate UnitSerializer in the Companion.serializer()
function generated on classes that use polymorphic or sealed serializer
by default. (Fixes https://github.com/Kotlin/kotlinx.serialization/issues/1692)
instead, plugins should emit the code similar to the TYPE_OF one with
a special call to MagicApiIntrinsics.voidMagicApiCall directly afterwards.
This is required because old compiler need to correctly inline code
rewritten by plugin.
Add intrinsic for kotlinx.serialization.serializer<T>() function.
Plugin intrinsic for old backend is removed because it is too hard
and unjustifiable to unify them.
(Old is created first because all intrinsics emit bytecode anyway)
Provide intrinsic for serializer<T>() function so it won't
invoke typeOf() construction and KType->KSerializer conversion
making it fast and truly reflectionless
Add support for recalculating stack size in plugin-defined intrinsics
since it is needed for correct work:
Unify method for recalculating stack size with existing typeOf intrinsic
Add testdata for IR for future intrinsic in IR
Dmitriy Novozhilov