JDK types first reached by a source scan stay uninspected when any package is preloaded: `isFunctionalInterface()` is false for `PrivilegedAction`
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 48/100
Research direction
Start by tracing ClassSymbolScanner.loadType(..., LAZILY) and the JavaInspectorImpl copy and commit loops described in the issue, then compare the preload and no-preload paths in the reproduction. Check how CompiledTypesManagerImpl.preload and on-demand inspection affect the type's inspection state. Done means the reproduced PrivilegedAction reports run() as its single abstract method and isFunctionalInterface() returns true after parsing with any preload.
Written by the indexing model from the issue text.
Description
Reproduction
SourceSet javaBase = SourceSetImpl.javaBase();
SourceSet sources = new SourceSetImpl.Builder().setName(JavaInspector.TEST_PROTOCOL)
.setUri(URI.create("file:/")).setDependencies(List.of(javaBase)).build();
InputConfiguration config = new InputConfigurationImpl.Builder().addSourceSets(sources)
.addClassPath("jmod:java.base").build();
JavaInspector ji = new io.codelaser.maddi.inspection.openjdk.JavaInspectorImpl();
ji.preload("java.base::java.util."); // any preload; see the table below
ji.initialize(config);
TypeInfo x = ji.parse(Map.of("a.X", """
package a;
import java.security.PrivilegedAction;
class X {
static String m() {
PrivilegedAction<String> action = () -> "x";
return action.run();
}
}
"""), new JavaInspector.ParseOptions.Builder().build())
.parseResult().primaryTypes().stream().findFirst().orElseThrow();
// visit m()'s body, take the Lambda, and inspect l.abstractFunctionalTypeInfo()
| preloads | PrivilegedAction after parse |
|---|---|
| none | hasBeenInspected()=true, singleAbstractMethod()=run(), isFunctionalInterface()=true |
java.base::java.util. |
hasBeenInspected()=false, singleAbstractMethod()=null, isFunctionalInterface()=false |
java.base::java.util. + java.base::java.security |
hasBeenInspected()=true, run(), true |
Seen on 0.9.5-alpha.11.
Expected
After parse returns, every type the parse result refers to answers member queries correctly, whatever was preloaded. Failing that, it should answer by completing the type on demand. A TypeInfo that silently reports "no single abstract method" while the parser has just resolved a lambda against it is the worst case: callers can't tell an uninspected type from a non-functional one.
Where it seems to come from
This is from reading the code; I haven't stepped through it.
ClassSymbolScanner.loadType(..., LAZILY)skips the members andsetSingleAbstractMethod, and doesn't commit.- The type is completed later, by
JavaInspectorImpl's "copy into CTM" loop. That loop commitsscanned.primaryTypes(),classSymbolScanner().typesLoaded()(=infoByFqn.typesLoadedForThisSourceSet()), and the preloads. - With a preload, a shared-JDK task exists (
preloadPass/wireSharedJdk). A platform type first reached during the source scan is then loaded on that shared task and registered for the JDK source set. It is not registered for the scanned set, so the loop never sees it. The preload pass's own commit loop has already run by then. - Without any preload there is no shared task, so the scanner's own
typesLoaded()includes the type and it gets committed. - Also,
CompiledTypesManagerImpl.preloadis a TODO, and LAZILY-loaded types from the openjdk scanner don't registersetOnDemandInspection. SoTypeInfoImpl.inspection.get()returns the uncommitted builder instead of completing the type.
Impact
Code that branches on isFunctionalInterface() or singleAbstractMethod() for a lambda's or method reference's target type gets the wrong answer whenever that JDK interface wasn't covered by a preload. The same probably holds for any other member query on such a type (methods(), fields(), ...).
- Dominant language
- Java
- Stars
- 1
- Forks
- 1
- Avg merge
- 2h 51m
- Merged PRs (30d)
- 3
Getting set up
- No Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from CodeLaser/maddi
-
build/ci good first issue
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
-
Difficulty 4/5 3-5 days Newbie friendliness 45/100
-
Difficulty 3/5 Half a day Newbie friendliness 62/100
-
enhancement
Difficulty 3/5 1-2 days Newbie friendliness 45/100
-
Java → C# translation: an idiomatic C# printer driven by the modification and nullability analysesOpenenhancement extension front-end:csharp
Difficulty 5/5 Over a week Newbie friendliness 8/100
Similar issues
-
needs-area-label
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
quelea-projection/Quelea#711 ·
-
bug pdd
Difficulty 2/5 Under an hour Newbie friendliness 67/100
objectionary/eo#9450 ·
Maintainers usually reply within 1 day
-
[EC2] DescribeAvailabilityZones should return real AWS error for unknown zone namePossibly taken @omatheusmesmo claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 75/100
Maintainers usually reply within 1 day
-
area: core good first issue type: enhancement
Difficulty 2/5 1-3 hours Newbie friendliness 68/100