What's done:
- Reduce the visibility of all entities inside the module to `internal`
or `private`. Only leave `fun main()` as `public` because it's the
single legal entry point.
- Add comments to few classes related to "contents" command that they
should be removed together with the command after 2.0.
- Drop useless `DeclarationHeaderRenderer` interface.
- Rename: KlibToolLinker -> KlibToolIrLinker
^KT-62340
When a declaration with an ObjC annotation is stored in a different
module, the annotation arguments are not resolved by default. This leads
to a bug in the checker. Before attempting to find the objCName, it is
necessary to resolve the annotation first.
#KT-64276
It is possible to add source files to task directly. Especially this is
a very popular case for Android integration. Android variants can
contain any number of extra-source files that should be compiled
together with other android-specific sources.
Android specific sources that come from Android source sets already
compiled together as part of "main" fragment (KT-62508).
With this change this logic is extended to any extra sources
that are added directly to the compile task.
^KT-62508 Verification Pending
Java sources can be produced by the task or can be configured with
task dependencies. Previously sources were included into task directly
now they are properly configured with Kotlin Android Source Set.
^KT-63753 Verification Pending
^KT-65426 Verification Pending
Android projects has a number of build type/flavor specific source set.
These source sets are included into compilation of multiple variants.
In Android compilation model they should be compiled together and they
should see each-others declarations. With Kotlin MPP plugin applied
these android source sets gets associated with Kotlin Source Sets
counter parts. However in Kotlin Multiplatform Source sets have
a semantics about "expect/actual" refinement and with K2 it also got
stricter and visibility is checked. This is not compatible with Android
compilation model thus the solution is to merge Android Source Sets
into one K2 Fragment.
^KT-62508 Verification Pending
Setting sources through compile task is not compatible with
Multiplatform K2 because for such sources no K2 Fragment relations
will be set and compilation will fail.
^KT-63753 Verification Pending
This is needed to reduce the size of generated test files, which started
to exceed default IDE limit
Also update some (mostly old) test utilities to remove exceptions from
java signatures
There's been a change in the error message for
INCOMPATIBLE_OBJC_NAME_OVERRIDE in K2 which makes it much more difficult
to diagnose the underlying issue. Previously in K1 the symbol that had
an incompatible objc name would be output, but in K2 only the class
declaration is output making it difficult to determine which member is
causing the error.
#KT-65572 Fixed
It is disabled by default, can be enabled by code change.
It's useful when changes in dumper is done with significant
amount of tests changed.
^KT-65460
Users quite often confuse that compilerOptions in KotlinCompilation is
actually the same object in the Kotlin compilation task. We want to make
our API more straightforward and remove this ambiguity by proposing to
use task compiler options DSL.
^KT-65568 Verification Pending
With shared test running test cases that may fail or timeout should
be excluded from shared run and executed separately.
This is done by getting all test cases for the current executable and
putting them to the map that is checked for all TestRuns
boxInline tests for suspend use StateMachineChecker and share the state
if compiled and run together. The single value accessed from all tests.
It is enough to reset state before the test is run.
Added the ability to execute multiple tests once and share the results.
With Test runner turned on tests are being compiled into the single
executable. Then they executed each one separately using --ktest_*
options. This commit makes test run executable once and the result goes
to the result handler.
`testIntersectionOfPrimitiveBasedFunctions`
is fine, because we deprecate this
already in ^KT-63243.
`CANNOT_INFER_VISIBILITY` is
positioned as `DECLARATION_NAME`,
just in case we decide for
some reason that we do want to
report it along with
`MANY_*_NOT_IMPLEMENTED`. In that
case, we'd get a problem that
their bounds overlap, but not
completely contain one another.
^KT-63741 Fixed
^KT-59400 Fixed
Even though SO may not be correct
overrides sometimes, it feels more
natural to treat fake overrides as...
well, "overrides". And without it
we'd need to make the code in
`FirOverrideChecker` less intuitive.
Make parcelize work consistently with
`FirOverrideChecker`.
Actually, there is a philosophical question about
what to do when a plugin generates a class
in a new package, and this class is red code
(for instance, doesn't implement an abstract
member from a supertype). There's no source
to report such an error, but we probably do
want to run checkers to avoid trying to
compile red code to binaries (because it may
crash in backend, or it may silently work).
The change in
`FirOptInUsageBaseChecker` is needed
to prevent `OPT_IN_OVERRIDE` on
intersection overrides where one
of the base functions requires opt-in.
We never designed this in such
a way that it's an error.
See:
`FirPsiOldFrontendDiagnosticsTestGenerated.testIncorrectTargetsForExperimentalAnnotation`.
According to KT-63741, the case like
`kt40510.fir.kt` should, in fact,
be an error. It will be addressed
separately.
Dmitriy Dolovov