IDE:
Rewrite AnalyzerFacade and implementations for JS and JVM to support creating separate analyzers for each module
Introduce ModuleInfo which is an intermediate entity between configuration (tests or idea modules) and ModuleDescriptor
Implement IdeaModuleInfos which represent IDEA modules, sdks and libraries
Add (somewhat thin) test checking their behaviour
Implement getModuleInfo() - utility to obtain IdeaModuleInfo for PsiElement
Drop Project.getLazyResolveSession() - not possible to obtain resolve session for the whole project any more
Adjust JavaResolveExtension accordingly
KotlinSignature Intention/Marker - make sure that analyzed element is cls element (he's not in resolve scope otherwise)
LightClasses:
Create separate package light classes for each module
Java code can only reference light class from the first module among it's dependencies
Duplicate jvm signature is only reported on package declarations inside one module
Injectors:
Receive GlobalSearchScope as paramer for VirtualFileFinder and JavaClassFinder
which allows to narrow analyzer scope
JDR:
Introduce ModuleClassResolver resolves java classes in correct java descriptor resolver (corresponding ModuleDescriptor)
Add test checking that java classes belong to correct module
Debugger:
Provide context to analyze files created by debugger in
Converter:
Postprocessor now needs a context to analyze resulting code in
JetPsiFactory:
Add verification that files created by psi factory are not analyzed without context (that is almost never a good idea)
Other:
Use new API in various tests, utilities, run configuration producers and builtin serializers
Various "TODO: (module refactoring)" which mark the unfinished parts
Basically, this commit splits test data from the from java-txt-kt to two pairs java-txt and kt-txt.
This commit leads to some duplication in test data.
This is temporary: in the platform types branch the test data for LoadJava tests will be changed dramatically, so duplication will go away
Check if all of the syntax errors are at EOF, not that there's a single one.
In case of open parenthesis two syntax errors are reported for some reason
Following the rule of the least surprise: this option is named "-nowarn" in
other JVM language compilers. Besides, having an option with an argument which
can take exactly one predefined value is sort of confusing
x is a byte, x += 2 translated to x = x+2, but type of (x+2) currently
is Int, so it's not compiled because Int can't be stored in Byte.
Maybe we need to think about what should be a result of sum of to
Bytes, but currently it's still Int
Alexander Udalov