You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
IT: verify annotation processor in the same reactor can be used by a sibling moduleΒ #13911
Add an integration test that verifies an annotation processor (AP) built in the same Maven reactor can be used by a sibling module during its compile phase β with both the sequential and concurrent builders.
Background
Andres Almiray reported that he couldn't get this to work with Maven 3.8.x. Unlike reactor plugins (the MNG-1088 class of problem), annotation processors are discovered via a normal <dependency> in the POM (usually provided scope, or via <annotationProcessorPaths> in the compiler plugin config). That POM dependency gives Maven the ordering information it needs to build the AP module first β no special reactor-plugin detection is required.
However, there is no IT that explicitly verifies this scenario today.
What to test
A multi-module reactor with:
processor/ β a jar module that contains a javax.annotation.processing.Processor implementation writing a marker file
client/ β a jar module that declares processor as a provided dependency and verifies the generated output exists after compile
I reviewed the integration test and the recent change removing the concurrent-builder scenario.
I'd like to contribute to the remaining concurrent-builder issue. I understand that annotationProcessorPaths resolution currently does not use the reactor WorkspaceReader, which prevents resolving an annotation processor built in the same reactor without installing it first.
I am interested in investigating the workspace resolution path and adding a regression test to restore concurrent-builder support.
Thanks for adding this integration test. I noticed that the PR description originally mentions verification with both sequential and concurrent builders, but the concurrent test was subsequently removed because annotationProcessorPaths resolution does not use the workspace reader. Could you clarify whether concurrent-builder support is intentionally out of scope for this issue, and whether the remaining limitation should be tracked separately under MCOMPILER-496?
Activity
Remove concurrent builder test β annotationProcessorPaths not resolveβ¦
IT: verify annotation processor in the same reactor can be used by a β¦
apacheGH-13911: address review β named constant, javadoc fix, and tesβ¦
codewithsinchana commentedon Nov 11, 2026
I reviewed the integration test and the recent change removing the concurrent-builder scenario.
I'd like to contribute to the remaining concurrent-builder issue. I understand that annotationProcessorPaths resolution currently does not use the reactor WorkspaceReader, which prevents resolving an annotation processor built in the same reactor without installing it first.
I am interested in investigating the workspace resolution path and adding a regression test to restore concurrent-builder support.
codewithsinchana commentedon Nov 11, 2026
Thanks for adding this integration test. I noticed that the PR description originally mentions verification with both sequential and concurrent builders, but the concurrent test was subsequently removed because
annotationProcessorPathsresolution does not use the workspace reader. Could you clarify whether concurrent-builder support is intentionally out of scope for this issue, and whether the remaining limitation should be tracked separately under MCOMPILER-496?