The problem was that we were using unenhanced constructors that still
might contain java type refs.
Basically, most of the `firClass.declarations` usages are unsafe in that
meaning.
The test is very complicated because to reproduce the exception there
should be some conditions met:
- MyClass should be compiled with K1, so it has `@EnhancedNullability`
annotation serialized (that is unnecessary, but don't do any harm)
- `@Nls` scope should be properly resolved, so enum entries are converted
to resolved type refs
- That leads to attempt to check if null-check might be needed to be
inserted and to failure at FirTypeUtilsKt.getCanBeNull
While this quite controversial to have this check being performed for
freshly created annotations calls, the problem with improper constructor
lookup anyway existed and fixed correctly.
^KT-57213 Fixed
In a lot of cases, when we want to process the
declaration in scope, it should be resolved to at least TYPES phase.
To avoid doing it manually in all our variety of scopes, we do it when the scope it created.
It was implicitly working manually before as lazy resolve did a lot of extra work
on resolving a declaration it was not supposed to resolve.
Now it's not the case, and we have to explicitly resolve all the declarations we need.
^KT-56543
Co-authored-by: Ilya Kirillov <ilya.kirillov@jetbrains.com>
1. Leaving no unbound symbols in the IR tree, KT-54491:
- All unbound symbols are bound to synthetic stub declarations
- Improved detection of the root cause for every partially linked classifier
- Improved error messages
2. Visibility valiation, KT-54469
3. Always check deserialized symbols:
If the deserialized symbol mismatches the symbol kind at the call site in the deserializer then generate and reference another symbol with the same signature. In case PL is off, just throw IrSymbolTypeMismatchException.
4. Handle class inheritance violation:
- Detect illegal inheritance (ex: inheriting from a final class)
- Detect invalid constructor delegation (ex: delegating to another class than the direct superclass)
- Simplification: Reduce the number of PartialLinkageCase subclasses
- Reworked error message generation to have shorter and clearer messages
5. Handle class transformations and all known side-effects, examples:
- nested <-> inner
- class <-> enum/object
- adding/removing subclasses of sealed class
- adding/removing enum entries
6. Check direct instantiation of abstract class.
Such instantiation could be possible if a class was non-abstract in the previous version of a library.
7. Handle unlinked annotations on declarations.
Such annotations are removed from the IR. The appropriate compiler error message is produced for every individual case.
8. Handle value argument count mismatch at call sites
9. Handle calling suspend function from non-suspend context.
This could happen if a suspen function was non-suspend in the previous version of a library.
10. Handle overriding inline callables.
Only the leaf final callable can be marked with `inline`.
11. Detect illegal non-local returns from noinline/crossinline lambdas.
With the partial linkage turned on it's hard to predict whether a newly created IrSymbol that is added to the SymbolTable via one of referenceXXX() calls will or won't be bound to some declaration. The latter may happen in certain cases, for example when the symbol refers from an IR expression to a non-top level declaration that was removed in newer version of Kotlin library (KLIB). Unless such symbol is registered as "probably unbound" it remains invisible for the linkage process.
The optimization that allows to reference symbols without registering them as "probably unbound" is fragile. It's better to avoid calling any referenceXXX(reg = false) functions. Instead, wherever it is s suitable it is recommended to use one of the appropriate declareXXX() calls.
For the future: Consider implementing the optimization once again for the new "flat" ID signatures.
Also: Temporarily mute assertion that checks symbols overwrites in SymbolTable ^KT-57049
- If the callee reference in a desugared compound assignment was an
error reference, the compound assignment wasn't cached by
`FirElementsRecorder`. This led to KTIJ-24895 (see the issue for more
details).
- The solution falls back on the compound assignment's source to decide
`isSourceForCompoundAccess`, as an operation name can be reconstructed
from the PSI.
^KTIJ-24895 fixed
Otherwise, when we come to `ClassCodegen.kt:173
for `GradleActionTest` and check
`FUN FAKE_OVERRIDE name:<get-project>`,
we then go to `JvmSignatureClashDetector.kt:37`,
and call `mapRawSignature(overriddenFunction)`
which ignores the original `@JvmName`.
It does so because it relies on the property copy
which forgets to copy the getter, but
`@get:JvmName` is stored in it.
Extra when-branches for `FAKE_OVERRIDE` are needed,
because otherwise the annotations would not
be copied in `Fir2IrDeclarationStorage.kt:723`.
Extra when-branches for `DELEGATED_MEMBER` are
needed, because otherwise the generated IR changes
in some tests. For example, see:
`FirLightTreeIrTextTestGenerated.Declarations.testKt35550`.
The `assumesBackingField`-related change is backed by
the `FirLightTreeIrTextTestGenerated.Stubs.testJavaEnum` test.
`this.body = null` ensures the resulting IR
matches K1.
The change in `FirImplicitBodyResolve.kt` is needed,
because otherwise the bootstrap compiler fails at
`:compiler:frontend:compileKotlin`,
but I didn't come up with a smaller test for it.
If we don't make an explicit accessor copy,
then when we later create a `Fir2IrLazyPropertyAccessor`
for the fake override getter, it's `fir` will
be a reference to the `FirProperty`, not `FirPropertyAccessor`.
That's why `DumpIrTreeVisitor` will render `IntrinsicConstEvaluation` as
a getter annotation as well.
`FirPsiIrTextTestGenerated.testDelegatedGenericImplementation`
renders type parameters from `<get-x>`,
because when assigning `extensionReceiverParameter`
of the setter `<set-x>` we come to
`Fir2IrClassifierStorage.kt:638`,
and in this cache there's the parameter with
the `<get-x>` parent.
Note that `typeContext.origin == DEFAULT` in
`getCachedIrTypeParameter`.
It's `DEFAULT` because at the line `Fir2IrDeclarationStorage.kt:335`
the `function` variable is `null` (because there's no setter).
The change in `declarationAttributes.kt` is backed
by the `FirPsiIrTextTestGenerated.testKt45853` test.
^KT-57104 Fixed
^KT-57432 Fixed
Merge-request: KT-MR-9210
Merged-by: Nikolay Lunyak <Nikolay.Lunyak@jetbrains.com>
We've muted some irText tests on JS in the previous commits because
test expectations in some tests are different when targeting JS.
AbstractKlibJsTextTestCase tests use a different logic:
they compare the dump of the deserialized IR with the frontend-generated
IR, not with the expectations in text files, thus muted tests weren't
actually failing.
Here we introduce a temporary fix, namely
a separate // IGNORE directive for klib tests.
When klib tests are moved to the new test infrastructure, there will be
no need to do this.
Fake overrides are needed at least for static fields inherited from Java
interfaces. Not static methods though, because they are not inherited.
#KT-57302 Fixed
Ivan Kochurkin