The coroutine transformation would leave locals in the local
variable table across the code that reloads local variables from
the continuation on reentry. However, when reentering the function
the local has no value until after the reloads from the
continuation.
This change splits the locals in the local variable table to
avoid such uninitialized locals. A local alive across a
suspension point has its range split in two. One that goes
from the original start to the state label for the restart
after the suspension. The other goes from after the local
has been reloaded from the continuation until the previous
end of the local.
Also, do not try to use invokedynamic on SAM calls with intersection
types, because intersection type is not allowed as an immediate type
projection of a supertype, and constructing a fake override in
LambdaMetafactoryArgumentsBuilder led to an exception. This fixes the
problem which was worked around earlier in e6c089ef, effectively
reverting that commit.
The main motivation for this change is that LambdaMetafactory also
doesn't generate generic signature for SAM wrapper classes at runtime.
Since these classes are synthetic, nobody should rely on the fact that
they have generic supertypes, which was observable only via Java
reflection.
#KT-46149 Fixed
#KT-46238 Fixed
Coroutine transform would occasionally *extend* the range of a
local instead of shrinking it (if the next suspension point
is after the end point for the local). That leads to the local
variable table having a local that covers code where it is not
defined. That is invalid and leads to D8 removing the locals
table from the code.
Commonize (in terms of TypeSystemCommonBackendContext implementations
for KotlinType/IrType) code that computes optimal TypeMappingMode to
apply to different positions where inline class types can be present.
KotlinTypeMapper.mapInlineClassTypeAsDeclaration and
mapUnderlyingTypeOfInlineClassType invoked mapType which is defined in
descriptorBasedTypeSignatureMapping.kt and works on KotlinType.
It didn't lead to any problems, other than the fact that we were
constructing IrBasedClassDescriptor in JVM IR, and then KotlinType to
pass it to mapType, on each call to StackValue.boxInlineClass or
unboxInlineClass, which seems wasteful.
Instead of this, refactor these utilities to use type markers instead,
pass IrType and IrTypeMapper directly from JVM IR, and move the "static
type mapper" logic (which is used only in the old backend) out of
KotlinTypeMapper.
When a suspension point is inlined, the inlining local spans
unspilling code where the local slot has not been initialized.
The transformer already inserted initialization code for the local,
however, it did not split the local variable table. Therefore,
the inlining local is defined on instructions where the local
slot has no value (namely the instructions that initialize
the local slot on the unspilling path).
This change adds splitting of the local variable table as well.
When updating to a new version of D8, the uninitialized local
slot showed up. D8 will repair it in this case, but it is
better to not generate such code.
Pass parentContext to SamWrapperCodegen from the outside instead of
using the one from parentCodegen. The difference is that in case of an
inline lambda, we're creating an InlineLambdaContext whose parent is a
ClosureContext, but the codegen for that lambda has that latter
ClosureContext as its context. So the getNonInlineOuterContext call in
SamWrapperCodegen.generateInnerClassInformation wasn't able to identify
that this is a context of a lambda that needs to be skipped, and
generated it as EnclosingMethod, which caused Java reflection to fail
because once that lambda was inlined, it was removed and thus didn't
make it to runtime.
#KT-44827 Fixed
Alexander Udalov