Descriptors are already supposed to be sorted in scopes. The problem is
that rendering descriptors for sorting takes a lot of time (~1.5% of
total compilation time of intellij with JVM IR), and simple heuristics,
like comparing by names first, don't fully help with it.
#KT-48233
It is not correct to assume that arg0 has been generated to have the
same IrType as the whole expression.
The reason it only backfired in throwing exceptions probably has to do
with the fact that visitThrow might be the only place in
ExpressionCodegen right now which uses the IrType from PromisedValue to
make a decision on whether to generate checkcast.
It seems suspicious that ExpressionCodegen.visitFieldAccess/visitCall
return PromisedValue whose IrType mentions type parameters which are
declared outside of the call site, but that should probably be
investigated separately.
#KT-48440 Fixed
Generate $delegate method as instance method in
PropertyReferenceDelegationLowering, and remove dispatch receiver later
in MakePropertyDelegateMethodsStatic. The method needs to be static to
be non-overridable (see delegateMethodIsNonOverridable.kt), and public
to be accessible in reflection.
Otherwise we generated incorrect IR where a static function accessed an
instance field of the containing class, which failed in multiple places
including LocalDeclarationsLowering.
#KT-48350 Fixed
Given inline class V(Any?), a coercion from (Object, V) to (Object, V?)
is boxing.
In theory, the same issue in the old backend can be fixed by making
`KotlinTypeMapper.mapUnderlyingTypeOfInlineClassType` use
`computeExpandedTypeForInlineClass`, but for some reason this breaks a
lot of stuff.
#KT-48430 Fixed
This is already the case for straightline code such as
```
inline fun <R> f(size: Int, block: () -> R): R {
var result: R
result = block()
return result
}
```
However, if the local variable introduction happens at a merge
point as in the following example, we allow the unboxing but
only do it halfway. The initialization of the local is still
done with a null value.
```
inline fun <R> f(size: Int, block: () -> R): R {
var result: R
while (true) {
result = block()
if (size == 0) break
}
return result
}
```
This change disallows unboxing for this move complicated
case as well by bailing out if a local use is with a
TaintedBoxedValue (merge of Object and Integer).
^KT-48394 Fixed.
Looking for the primary constructor manually doesn't work if it's
private in the other module on JVM, because private declarations are
skipped in IrLazyClass.
Otherwise, R8 does not transform kotlin-reflect, failing bootstrap.
Leaving end label the same is safe, since we do not remove labels during
transformation.
Do not check, that all Unit predecessors are POPs. This is safe for the
same reason, as it is safe to allow some of ARETURN sources not be
suspension point results.
To elaborate, before Unit, the stack is empty. This is because if there
are multiple paths to Unit and at least one of them comes from POP after
suspension point (we are interested in this case only - otherwise, the
call is not tail-call), in path from said POP the stack is empty, since
after suspension point the stack contains only one element. Thus, the
stack in other paths leading to Unit has to be empty, otherwise, merge
operation is not possible and ASM will report error during analysis.
Since the stack is empty in all paths, we can hoist Unit and following
ARETURN to predecessors, effectively turning path from suspension point
to tail-call.
Mikhail Glukhikh