In the newly added test, prior to this change, JVM IR was generating
DefaultImpls classes with calls to things like
`kotlin/collections/MutableList$DefaultImpls.spliterator` and other
default methods present in JDK 8+. This obviously didn't make much
sense. Although these weren't explicitly mentioned anywhere in the
bytecode, they caused some validation tools to report errors (e.g.
animalsniffer used in arrow).
(cherry picked from commit 5647a935a2)
Instead of generating these annotation classes as package-private on
JVM, serialize their metadata to the .kotlin_module file, and load it
when compiling dependent multiplatform modules.
The problem with generating them as package-private was that
kotlin-stdlib for JVM would end up declaring symbols from other
platforms, which would include some annotations from package
kotlin.native. But using that package is discouraged by some tools
because it has a Java keyword in its name. In particular, jlink refused
to work with such artifact altogether (KT-21266).
#KT-38652 Fixed
This fixes the weird cases when a class gets overwritten by an imperfect
copy, reduces the number of classes in the output if an inline function
contains an inline call that causes it to have regenerated anonymous
objects, and makes inlining of same module functions a bit faster in
general. On the other hand, this may increase memory footprint a bit
because classes cannot be flushed to the output jar, as the inliner
would not be able to locate classes for anonymous objects if they have
already been unloaded from memory.
The fields containing crossinline lambdas should be package-private to
avoid generating synthetic accessors, which break object regeneration.
Note that the inline methods cannot actually be called, as call sites
will attempt to read the captured lambda from a field through a *copy*
of the local containing the object, so these reads will not be inlined,
causing an exception at runtime:
inline fun f(crossinline g: () -> Unit) = object : I {
inline fun h() = g()
// effectively `val tmp = this; return tmp.$g()`:
override fun run() = h()
}
f {}.run() // NoSuchFieldError: $g
This particular example can be fixed by reusing locals for receiver
parameters in IrInlineCodegen, but explicitly assigning `this` to
another variable and calling an inline method on it will break it again.
(This is only applicable to the JVM_IR backend, as the non-IR one fails
to generate `f` at all for some other reason.)
Instead of generating overrides for getOwner/getName/getSignature in
each anonymous class representing a callable reference, pass them to the
superclass' constructor and store as fields. This occupies some small
memory but helps to reduce the size of the generated class files, and
will be helpful for adding further runtime information to callable
references, such as information about implicit conversions this
reference has been subject to.
Represent owner as java.lang.Class + boolean instead of
KDeclarationContainer, so that the unnecessary wrapping Class->KClass
wouldn't happen before it's needed, and also to make sure all callable
references remain serializable.
Note that the argument type where the "is declaration container a class"
is passed is int instead of boolean. The plan is to pass the
aforementioned implicit conversion information as bits of this same
integer value.
#KT-27362 Fixed
They are inline-only.
Generate $$forInline versions of inline suspend functions as private.
This way, there is no nullability annotation on there parameters and return
values. Unfortunately, old BE does generate them.
#KT-37088 Fixed
coroutines intrinsic lambda.
The logic is if the lambda is crossinline we need to generate the
accessor. However, suspendCoroutine's and
suspendCoroutineUninterceptedOrReturn's parameter, despite being
crossinline, are effectively inline. Thus, we do not need to generate
the accessor.
#KT-27503 Fixed
In 1.3.31 I fixed Java interop for inline function with coroutines
(TL;DR: when we need a state machine, generate two methods: one with
normal name, and the other one with $$forInline suffix, for the inliner
to use, just like inline suspend functions), however, I forgot a case
with inline suspend function with inline suspend function parameter.
In this case, the compiler a generated two functions, as needed, but,
neither of them had a state-machine. This change adds the state-machine
for the method with normal name. Note, that suspend inline functions
with crossinline parameter, which are also supported by the change,
did not cause incorrect behaviour, since until now they were generated
as synthetic.
#KT-31354 Fixed
This fixes Java interop of inline functions, which use coroutines.
However, we cannot transform the state-machine. Thus, we generate
a $$forInline counterpart for suspend functions (similar to inline
suspend functions) and invokeSuspend$$forInline for lambdas if these
coroutines are going to transformed (i.e. are declared inside inline
functions).
During transformation we just skip method with state-machine and
transform the $$forInline counterpart. Of course, if inline site is
inline itself, we generate both state-machine version (which will be
dropped during the next transformation) and $$forInline version.
Consequently, the final version of the coroutines will not have
$$forInline counterpart.
Unfortunately, since CompileKotlinAgainstInlineKotlin tests do not allow
java sources, the tests for the interop are usual box tests.
#KT-30707 Fixed
This directive generates TailCallOptimizationChecker in package helpers.
The check for tail call optimization is based on coroutine stack traces
bug (feature?): when tail call optimization hits, the continuation
object is not generated. Thus, there is no debug metadata for this
suspend function. Consequently, the coroutines stack trace does not
contain stack trace element for that function.
This check is performed by TailCallOptimizationChecker.
Since this is runtime check, unlike bytecode tests, it does not require
test data adjustments on each codegen or inliner change.
Since the check is based on debug metadata, which is JVM specific, there
is not support for other backends yet.
Use `// !LANGUAGE: -ReleaseCoroutines` instead in tests which require
old (1.2) coroutines, and nothing in tests which require new coroutines
because master is already 1.3. Also remove superfluous API_VERSION and
other directives which have no effect anymore. Do not include runtime
automatically with `WITH_COROUTINES`/`COMMON_COROUTINES_TEST` in box
tests; require `WITH_RUNTIME` for that (majority of tests already had it
anyway), but remove it from bytecode text tests where runtime is always
added automatically. Fix the coroutine package selection code in
KotlinTestUtils and update the bunch files correspondingly.
Disable tests in `box/coroutines/noStdLib` on JVM: despite the name,
these tests were launched with stdlib because of the code in
CodegenTestCase, and they do not work without it because at least
CoroutineUtil.kt requires stdlib to compile correctly
Most of these tests used this directive as a way to opt in to a new
language feature, and most of those features are already stable for a
long time, so no opt-in is needed. Some other tests used the directive
to opt out from a language feature, replace those by the `LANGUAGE`
directive. One test used the directive to test behavior that actually
depended on the API version; use `API_VERSION` directive there instead.
Return type is not needed for checking overloads, but querying it may
involve resolving function bodies, which usually happens after overload
checking (see LazyTopDownAnalyzer.analyzeDeclarations) and at this point
can lead to incorrect BACKING_FIELD_REQUIRED value being computed for
some properties (see KT-27895)
#KT-27895 Fixed
