Three modes:
- 'disable' (default): normalize constructor calls in coroutines only
(required because uninitialized objects can't be stored in fields),
don't insert additional code for forced class initialization;
- 'enable': normalize constructor calls,
don't insert additional code for forced class initialization;
- 'preserve-class-initialization': normalize constructor calls,
insert additional code for forced class initialization.
Jump out from expression (e.g., break or continue expression in call
arguments) requires stack normalization, which inserts POP instructions.
POPping an uninitialized value is similar to ASTORE, except that it
doesn't store a value to a local variable. Such POP instructions should
be removed during postprocessing of the uninitialized stores.
Use it for char boxing/unboxing and unit materialization.
Possible to use for other purposes, for example, to add type checks
to dynamics.
See KT-18793, KT-17915, KT-19081, KT-18216, KT-12970, KT-17014,
KT-13932, KT-13930
It's necessary for generic inline suspend as a codegen
for it uses binding slice SUSPEND_FUNCTION_TO_JVM_VIEW
to generate fake continuation parameter, so all the
descriptors that are used for body generation must be
obtained from the SUSPEND_FUNCTION_TO_JVM_VIEW
#KT-19528 Fixed
FixStack transformation divides on phases:
- Fixing stack before break/continue
- Fixing stack for inline markers/try-catch blocks
After the first stage all ALWAYS_TRUE markers are replaced
with simple GOTO's and if we're skipping break/continue edges
we won't reach the code after while (true) statement.
At the same time it's fine to not to skip them in the second phase
as the stack for them is already corrected in the first phase
#KT-19475 Fixed
Dex ignores subsequent line numbers for same instructions and interprets
instruction after inline as if they were inlined. This makes debugger
behaves as if there's nowhere to stop on line with breakpoint.
This also makes stepping through inline function consistent with
non-inline analog. In both context debugger now stops on '}'.
#KT-18949 Fixed
#KT-17120 Fixed
The main reason is avoiding complicated operations like
dup + dup_x2 + pop2 for instances obtained by NEW instruction.
Otherwise it leads to problems in performRefinedTypeAnalysis
because code there has a sensible assumption that NEW instances
can be only dupped or stored into a local (rare cases)
#KT-17457 Fixed
These changes has been introduced in c3a032ea0b
to support new scheme for named functions
At the same time they can be avoided by generating some additional
code/accessors in the anonymous classes for coroutine state
Now we cannot simply remove introduced API because of bootstrap problems
#KT-17898 Fixed
The problem was that the resume call (from doResume) for open members
was based on common INVOKEVIRTUAL to the original function
that lead to the invocation of the override when it was expected
to be the overridden (after super-call being suspended)
The solution is to generate method bodies for open members into
the special $suspendImpl synthetic function that may be called
from the doResume implementation
#KT-17587 Fixed
The main idea is to leave all the inline functions as is, without
state machines (but keeping suspend-calls markers) and
determine whether we need a state machine from the bytecode
after inlining into a non-inline function
#KT-17585 In Progress
#KT-16603 In Progress
#KT-16448 Fixed
Don't use coroutine facade in following cases:
* When calling coroutine functions in non-suspend mode (i.e.
suspend flag is either false or skipped).
* When passing continuation from suspend function to suspend function.
Use facade only for corresponding intrinsics.
Reason: interceptor should not be called on each suspend function
invocation, it should be called after resume from innermost
suspend function.
Fix KT-17067. The example provided with the issue is very similar to
dispatchResume.kt which passes after these changes.
Add test to prove that KT-17446 is no more reproducible.
Stop making aliasing suspend function descriptor with reference to
instance of state machine. This may cause problems in some cases,
for example, when compiling recursive suspend function. See KT-17281.
Instead, make alias for synthetic continuation parameter. This
additionally required some refactoring, e.g. *always* generating
continuation parameter during codegen.
This patch mutes the following test categories:
* Tests with java dependencies (System class,
java stdlib, jvm-oriented annotations etc).
* Coroutines tests.
* Reflection tests.
* Tests with an inheritance from the standard
collections.
The problem was that he number of mask parameters for defaults when
generating methods declaration was being calculated upon resulting signature
(with additional parameters: extension receivers, enum name/ordinal),
while on call-sites the masks number was calculated by the arguments number
in resolved call, i.e. by the number of real value parameters.
And because of the additional synthetic parameters (like enum.ordinal) these
two numbers could be different.
The solution is just to use value parameters number in both places.
Note, that we only count value parameters from the original sourse
declaration, ignoring synthetic ones generated by backend (e.g.
Continuation for suspend functions)
#KT-14565 Fixed
Ilmir Usmanov