- The exceptions in KT-63130 occur because PSI tree change events may be
published when parts of the PSI under modification are inconsistent.
Such inconsistent elements are then analyzed by
`LLFirDeclarationModificationService` to check if the modification
occurs in a contract statement.
- The solution adds `*OrNull` functions to `KtQualifiedExpression` which
return `null` instead of failing with an exception when the receiver
or selector cannot be found.
^KT-63130 fixed
- The cause of KT-63130 is that `LLFirDeclarationModificationService`
tries to analyze inconsistent PSI. These new tests replicate the
inconsistent PSI by deleting different children and then call the
declaration modification service with it, making sure that the service
doesn't throw any exceptions. Both exceptions mentioned in the issue
are reproduced by the added tests, specifically, the receiver and
selector resilience tests for dot-qualified expressions.
- A few of the tests need to be marked with `IGNORE_FIR` until the error
has been fixed in the subsequent commit.
^KT-63130
- To support PSI modification, the test environment requires some
additional configuration, which is encapsulated in
`AnalysisApiPsiModificationTestServiceRegistrar`.
- The test also has to allow write access explicitly, which is
configured in `AnalysisApiFirModifiablePsiSourceTestConfigurator`.
- In general, Analysis API tests forbid write access because the
Analysis API should not be used from write actions. However, in some
cases we might want tests to e.g. modify PSI, which requires write
access. This change allows `AnalysisApiTestConfigurator`s to enable
write access for the specific test.
The commit removes all the KLib resolution logic,
now Analysis API Standalone clients need to provide all the KLib
list directly.
The resolution logic was removed as too error-prone and requiring compiler configurations.
Kotlin Gradle plugin can provide a full set of required KLibs,
so if a client is a Gradle plugin, this should not be an issue.
Probably, some fancy API which will explicitly perform all KLib dependency searches
should be introduced in the future (KT-63395)
^KT-63126 fixed
Intercept decompiler calls to process files that belong to classpath
builtins from K2 IDE differently. Their content won't be filtered w.r.t.
JVM logic during decompilation.
Split stub versions for .kotlin_builtins with K1 and K2 IDE.
K2 currently uses a single shared symbol provider for builtins of all
platforms. KotlinBuiltInDecompiler filters out duplicated declarations
contained in JVM classfiles as required for K1. For K2 this logic
doesn't fit: non-JVM modules (common, in particular) rely on
.kotlin_builtins, because kotlin-stdlib-common doesn't contain
.kotlin_builtins or .knm files for some of the built-in declarations
(e.g., kotlin.Unit). IDE has to restore the missing declarations using
built-ins from classloader. Filtering declarations from there using K1
logic breaks resolution of these declarations inside common modules in
IDE. The change shouldn't affect JVM modules as are able to resolve
all builtin symbols from the standard library before falling back to
built-ins.
KT-61757
In certain situations like constructor calls, it's possible
to have different WasExperimental/SinceKotlin pairs associated
with one use-site, e.g. a constructor call is associated both
with a constructor declaration and a class declaration.
However, looks like it make no sense to handle this situation
in FirSinceKotlinHelpers, because FirOptInUsageBaseChecker already
handles them.
#KT-59825 Fixed
Completion in the IDE relies on correct scopes for KDoc positions.
So we add testData for `KtScopeProvider` and `ContextCollector`
to ensure that there are no regressions
- In addition to file declarations and class members, we also need to
commit pending diagnostics on declarations in scripts, to honor
suppressions before the diagnostics are committed indiscriminately.
^KT-62899
- `scriptUninitializedTopLevelProperty` is currently disabled in FIR
because it fails with a mismatch between diagnostics from the file and
the individual elements. It will be enabled in the next commit.
^KT-62899
- In KT-62899, diagnostics produced by file checkers weren't suppressed
because the top-level declarations on which the diagnostics are
reported were never visited by the same reporter.
`LLFirDiagnosticVisitor` skips nested declarations by design, as
diagnostics are collected for each structure element separately.
However, suppression takes effect when diagnostics are committed on
some element. If the element is not visited, suppression is skipped
and the diagnostic is committed at the end without a check by
`FileStructureElementDiagnosticsCollector.collectForStructureElement`.
- The fix ensures that diagnostics are committed on top-level
declarations, similar to how this was already done for class members.
^KT-62899 fixed
- `testConflictingOverloadsAtTopLevel` is currently disabled in FIR
because it fails with a mismatch between diagnostics from the file and
the individual elements. It will be enabled in the next commit.
^KT-62899
for `ACTUAL_ANNOTATIONS_NOT_MATCH_EXPECT` checker near to
other tests of this diagnostic. It is fine to move it out of
`multiplatformCompositeAnalysis` dir, because, actually, it tests
the case with comparison of FirAnnotationCall and FirAnnotationImpl,
which doesn't require composite analysis mode.
^KT-62559
...`ACTUAL_ANNOTATIONS_NOT_MATCH_EXPECT` checker near to
other tests of this diagnostic. This is because now such tests
are run in lazy resolve test suite, so no need to store them
separately.
^KT-62559
New tests are red in current commit!
All `.ll.kt` files are just copies of `.fir.kt` or `.kt`. This will be
fixed in the next commit to clearly show difference between LL and FIR.
The original purpose of this change was to test
`ACTUAL_ANNOTATIONS_NOT_MATCH_EXPECT` diagnostic, for which FIR checker
will be run only in IDE (to be changed in next MR), so its reports
a lost from usual KMP tests.
But this looks like a useful change for all KMP tests, not only this
diagnostic.
^KT-62559
Mikhail Glukhikh