This is a proposal to fix KT-49364.
Operand stack type verification happens before `checkcast`
is executed. When we implicitly cast an inline class
to a non-inline type, if type of stack value is not subtype
of the target, then coercing is necessary.
It helps to:
- keep declarations even if they are not reachable and not exported
- not minify names of not exported declarations
Compiler argument: -Xir-keep=A,B
Can be used for top-level declarations or for member
^KT-54118 fixed
For non-suspend lambdas annotations are carried over to the
invoke method so that tooling can find the annotation there.
It seems reasonable that annotations are carried over to
the invokeSuspend method on suspend lambdas as well so that
similar tooling can be built and work for suspend lambdas.
Disable it if we do not have required `noCompiledSerializer` function in
runtime. Leave it enabled in tests.
Rollback some changes for old backend as it is unsupported now.
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.
#KT-53947 fixed
without the fix, in the presence of the implicit receiver of the same
type as the processed receive, the lowering was incorrectly substituting
the correct receiver with the accessor to the implicit one
ProcessCanceledException indicates that the compilation process was
terminated by user. This kind of exception should not be wrapped anyhow
in order to be handled on toplevel in CLICompiler.
#KT-38483 Fixed
After an accidental change in 846537b367, we started generating metadata
into anonymous classes for callable references. This metadata contained
the Kotlin representation of the referenced function. In KT-53794, this
led to a problem where Java's protected visibility could not be
represented in Kotlin terms, which crashed the backend.
But also, this metadata is useless because there's no real use case for
interpreting it (since you already have a function reference object at
runtime with all the needed information), so it would take some extra
space in the bytecode.
#KT-53794 Fixed
This intrinsic was always calling the supplied suspend function
non-virtually, and this was hardcoded in the intrinsic generator.
Now this can be configured with `superQualifierSymbol` in the respective
IrCall, just the way this works with any other calls.
As a side effect, global optimizations now process this intrinsic more
correctly (before that, they didn't take the intrinsic generator into
account, and treated the call as virtual).
Igor Yakovlev