To convert smart cast expression, now we just convert
original expression in standard way and wrap it with TYPE_OP.
Before this commit original expression was converted in different way,
that led to errors e.g. for this expression casting.
Before this commit, extension receivers were declared only for
properties with container source, which is strange & inconsistent.
Now we declare accessor extension receiver iff corresponding property
has extension receiver.
This commit addresses the following issues:
* accessors didn't take into account their property's receiver type,
which caused NoSuchMethod due to signature mismatch. Now the property's
receiver type is passed to Fir2Ir translation of accessors.
* property's parent was not class, e.g., kotlin.collections.indices.
Now the symbol table collects WrappedPropertyDescriptorWithContainerSource
besides WrappedFunctionDescriptorWithContainerSource, so that
facade classes for such properties can be generated before codegen.
* accessor's parent was not class. Now the containerSource of
the property descriptor is passed to accessor descriptor.
Otherwise:
* should the dispatch receiver of a call be another call to a `suspend
fun` wrapped in something that is optimized away later, the owner of
the method will be incorrect;
* references to functions returning non-Unit but casted to `() ->
Unit` (allowed by new inference) might in fact not return Unit after
tail call optimization.
Usually FIR enum entry is initialized by anonymous object,
which is the container for all enum entry' declarations.
However, for simple enum entries there is no need of initializer at all.
The only case we may need fake override for private members is
a situation when class refers itself with different type arguments.
So in this commit we forbid such fake overrides when we can prove
that class does not refers itself here.
Library methods such as 'listOf' are resolved
to have the package fragments as their parents,
but JVM expects their containing file classes as parents.
This fix generates those file classes and
uses them as parent replacements for such library methods.
* TailCallOptimizationLowering should go into local classes in order to
transform their suspend methods;
* the check for invokes of noinline lambda arguments in codegen was
incorrect, as it also returned true for calls of lambdas stored in
local variables;
* IrInlineCodegen should mark non-inlinable arguments used as inline
suspend parameters;
* detection of suspend/inline call sites was incorrect (or maybe it's
the `compilationContextDescriptor` that was incorrect?..)
`invoke` in suspend lambdas overrides FunctionN.invoke, so the
refactored BridgeLowering already generates correct bridges there.
All the hack does is break overrides of interface suspend methods.
We always set the $assertionsDisabled field based on the top-level
enclosing class. This means that for anonymous objects we have to
rewrite the call to Class.desiredAssertionStatus.
values within initializer blocks.
The issue occurs in code like this:
```
class C {
var b = true
init {
b = false // Missing PUTFIELD for this statement
}
}
```
Added a new statement origin for field initialization (at declaration)
instead of relying on `origin == null` in ExpressionCodegen to determine
whether to generate the initializations.
This was unintentionally broken in
d68a1898d0.
Before this commit we had two statements blocks for 'for' loops:
range variable declaration + iterator variable declaration + while loop.
However, BE requires a bit different loop structure to make lowerings
properly so in this commit iterator declaration & while loop were
extracted to separate block.
Dmitry Petrov