Initial problem is started in `capturedBoundReferenceReceiver` method
where we assume that bound receiver is captured for usual call.
Note that if method is inline then we don't pass actual name reference
receiver, but pass special CAPTURED_RECEIVER_FIELD, which is then
is used to find special instructions during inline and fold several
instructions in `foldFieldAccessChainIfNeeded`.
As a result, we got unboxed reference receiver for inline call, which
caused CCE and to fix it we should box receiver one more time during
inline
#KT-28188 Fixed
Getter of a primary value of an inline class belongs to the box class.
Its arguments should not be unboxed when the method is called.
However, its result might require boxing if it's an inline class value.
When we have an internal primary value, there's no getter method.
In fact, we can use box/unbox methods for inline class directly
(don't forget to box the result, it may be an inline class type value).
#KT-26748
Note that this commit doesn't fix case when some inline class over
`Any` is returned from a lambda, it'll be fixed further as part of the
#KT-27586
#KT-27737 Fixed
Use only getDeclaredMethod/getDeclaredConstructor instead. The reason is
that getMethod/getConstructor only finds public-API (public or protected
on JVM) declarations, and to determine if a declaration is public-API in
the class file we used isPublicInBytecode, which was trying to load
annotations on the declaration to see if it was InlineOnly, and that
required lots of time-consuming actions and worsened the stack trace (as
can be seen e.g. in KT-27878). In fact, the implementation of
Class.getMethod is not supposed to do anything complicated except
loading annotations from each superclass and superinterface of the given
class. Doing it in our codebase simplifies implementation and probably
improves performance
Unsubstituted type should be used for coercion to box value of
inline class type if it's needed. For the substituted value it's not
known if it was a generic parameter or not.
#KT-27502 Fixed
This change reverts the AssignmentTranslator logic to a previous state
of "if we assign to a val, tranlate to backing field". Previously a
check whether or not we are inside of a constructor was added. The
check didn't detect secondary constructors, hence initializing of
val's with backing field started to work incorrectly.
The check itself was added in an attempt to prevent augmented assignment
operators to reference the backing field. The check seems to have been
wrong, because an augmented assignment could happen inside a construcotr.
A more correct fix was added later. It seems that it is safe now to
revert the logic back and rely on the frontend to only allow assignment
to a val property during initilization.
Alexander Udalov