* Use type_layout to declaratively express heap object headers in both
custom allocator and ObjectFactory.
* Invoke constructor (w/o invoking Kotin constructors) for created
objects and arrays from both custom allocator and ObjectFactory.
Previously:
- custom allocator only checked body for nullability (now this is
performed in body constructor)
- ObjectFactory only constructed ObjectData
* In each GC have a AllocatorImpl.hpp and ObjectData.hpp headers
the first encapsulating allocator-specific types, the second
containing specific ObjectData implementation.
* In each GC have a separate ObjectFactoryTraits that does not
actually depend on the specific GC anymore.
* Each GC now expose ObjectData (as undefined type) and its descriptor,
the latter being used by the custom allocator and ObjectFactory.
* Descriptors for ObjectBody and ArrayBody now live in Memory.h and the
code calculating size is now shared. Their constructors check that the
memory is zeroed (Kotlin constructors will expect this).
This is needed because after changing default language version to 2.0,
we still need to test many scenarios from this test in K1.
Move K1-specific tests to OldCompileKotlinAgainstCustomBinariesTest, and
K2-specific tests to FirCompileKotlinAgainstCustomBinariesTest.
Mute failing K2 tests via `muteForK2`. It will throw exception if the
muted test will suddenly start passing.
This test uses multiple features which are obsolete and will be
discontinued soon:
- Language version 1.4
- Expect/actual declarations in the same module
- Declaring inline classes via "inline class"
Check that the client code compiles against the "broken" classpath
instead. Descriptor text dump is a part of the obsolete test
infrastructure and should not be used anymore.
`DelegatingConstructorCall` -
because in this case there is
either `UNRESOLVED_REFERENCE` or
`SYNTAX` already.
`ErrorTypeRef` - because in this
case there is `SYNTAX` already.
Note that there is only 1 test
that shows this behavior, and
it's LT-specific. It's
`FirLightTreeOldFrontendDiagnosticsTestGenerated.testKt847`, and
in this test the string template
is parsed and processed
differently in LT and PSI:
- in LT there are 6 nodes: "",
"$this", " must be", "$as", "$t",
"".
- in PSI there 4 notes: "$this",
" must be", "$as", "$t".
Because of KT-50547, this test doesn't work on Windows with HMPP
enabled.
Previously, because of that HMPP was disabled for all samples.
But now disabling HMPP is no longer possible, so ce9a45b enables HMPP
for samples.
Disable echoServer sample on Windows to workaround ^KT-50547
Exclude kotlin-stdlib-js project for JPS build. Related to KT-56106
#KT-61059 Fixed
Merge-request: KT-MR-11527
Merged-by: Aleksei Cherepanov <aleksei.cherepanov@jetbrains.com>
We should copy `-Xcontext-receivers` flag from the original arguments
during the setup of compiler arguments in the FP test
Otherwise, UNSUPPORTED_CONTEXTUAL_DECLARATION_CALL diagnostic gets
triggered where it shouldn't be present
This helps to completely get rid of Visitors API during reading and writing
of the JVM metadata.
Note: Klib writers still use .accept and left for future refactorings.
See also #KT-59442
Rewrite Readers.kt and Writers.kt to avoid using
Visitors API
Align KotlinCommonMetadata/KotlinModuleMetadata API with
KotlinClassMetadata one (it is necessary for restructured reading/writing)
#KT-59442 In Progress
This change covers the case where some f/o was generated in common module
and it is referenced in platform code. But signature of this f/o may be
different in different modules because of e.g. actualization of value
parameters with actual typealias
^KT-60850 Fixed
...other variants where it can be expected. Such as:
* commonMainMetadataElements (legacy metadata)
* metadataElements (klib metadata)
* nativeApiElements
This would prevent from leaking kotlin-stdlib-common dependencies of
older Kotlin versions. So every KGP configuration where
stdlib dependencies is configured automatically will enforce dependency
constraints on kotlin-stdlib-common.
^KT-61053 Verification Pending
This commit includes the basic Java stubs generation infrastructure and
the corresponding tests. The main entry point is called
Kapt4StubGenerator. Calls to it from production code will be added in a
separate commit.
#KT-51982
Without this change, the Unit type does not have Any members declared
in its scope which breaks the expect/actual mapping
(aka we cannot find actual `toString` for expect one).
As 'FirCodeFragments' are converted to IR independently of its context,
in some cases duplicate (and not quite correct) symbols for local
classes and functions are created.
Until properly fixed in fir2ir, here we replace such duplicates with
original symbols.
In the old JVM backend, local functions were compiled as JVM classes,
so there were an instance to pass around. Today, they are compiled to
static functions in the containing class/facade, and calls to them
compile natively.
Alexander Shabalin