- Support partial linkage for declarations from inline functions that are lazily deserialized from a KLIB with static cache
- Linker API: Split IrModuleDeserializer.deserializeIrSymbol() into deserializeIrSymbolOrFail() and tryDeserializeIrSymbol()
- Linker API: Refactor KotlinIrLinker.handleSignatureIdNotFoundInModuleWithDependencies() to deserializeOrReturnUnboundIrSymbolIfPartialLinkageEnabled()
^KT-52478
Compiler tests use the old test infrastructure, but fir2ir tests use the
new one (tests-common-new). Removing this dependency means that now the
two big modules, `:compiler:compileTestKotlin` and
`:compiler:tests-common-new:compileTestKotlin` can be compiled in
parallel, which improves the total build time.
Add an analysis parameter that can be used to create a
package fragment provider for a common KLIB library
when composite analysis aka HMPP support in IDE is disabled.
Such KLIBs can appear since 1.6.20, when hierarchical
project structure was enabled by default. This means, that even
if HMPP is explicitly disabled in KGP via the Gradle property,
common code is still represented by KLIBs and not by
.kotlin_metadata as it used to be. IDE can't resolve
agains common KLIB dependencies without a corresponding
package fragment provider that can read KLIB libraries.
KT-53164
When incremental compilation fails, we currently log it at the `debug`
level (and fall back to non-incremental compilation). This commit will
change it to `warning` so that we can get more user reports, which will
allow us to fix the root cause.
Also make sure the warning includes a stack trace.
Additionally, let ReportSeverity.fromCode() return a non-null value
or throw an exception otherwise as that case is not expected.
^KT-52839 In Progress
- Fix test data of light class tests
(SymbolLightClassesForLibraryTestGenerated).
- Revert the changes in kt8666.kt because they fail on JS IR, and it
seems that this test is mostly needed for JS anyway.
- Use 'invoke(' as the pattern in bytecode text tests because plain
'invoke' is used when these tests are run on JDK 11, where indy string
concatenation is enabled by default and therefore we reference
`java/lang/invoke/StringConcatFactory`.
This also changes the transformation to inline the body of a directly
invoked lambda rather than producing a call to an anonymous local
function. The latter is unsupported in inline functions and problematic
from an ABI perspective, since it results in functions whose name
depends on the entire source code up to this point.
Dmitriy Dolovov