- Access underlying fields directly, instead of using accessors. This is
safe since data classes are always final.
- Don't generate a temporary variable in hashCode to use as the
accumulator.
Both optimizations are effectively present in the current JVM backend
and with these changes the generated code is much closer to what the
current backend generates.
This is actually either a LAMBDA or an ANONYMOUS_FUNCTION.
Not quite sure if it's really required, but some tools such as IR-based
decompiler might require this information.
* Generate missing IR parameter declarations in external stubs
* Use origin = FAKE_OVERRIDE in external IR stubs for fake overrides
* Set .parent in external IR stubs
* Set .parent in IR generated by some utility methods
This causes a subtle issue with 'copy' function for data class,
which has parameters with source elements corresponding to
the data class primary constructor parameters: default value expression
for such parameters is generated second time (to be overwritten later),
which creates duplicate bindings if such expression included any
(possibly anonymous) declarations, such as lambdas or anonymous objects.
#KT-18208 Fixed
Dmitry Petrov