[JackPGreen]

Partially reverts change in #1217, specifically:

Do not use --source-path in legacy mode as not to suck any of those module-info.java files back

This is required to ensure correct package resolution on projects that mix modular & non-module Java code.

Fixes: #1242

• I hereby declare this contribution to be licenced under the Apache License Version 2.0, January 2004

[JackPGreen]

Note for reviewers:

• I've not analysed in detail what why this issue occurred
• There is no test coverage for this case
• I've tested the fix both against the tests (e.g. -Prun-its verify) and the original reproducer

Ideally, the above should be addressed.

[elharo]

test really is mandatory for a bug like this

[JackPGreen]

test really is mandatory for a bug like this

I understand the value but adding one is non-trivial - would need to slim down the reproducer to suit, I don't think any existing test covers this scenario (specifically - the legacy + multi-module case I saw has no imports). I'll try to revisit.

Also - the fact that this fix doesn't affect any existing tests but does fix my issues suggests that (in hindsight) there was insufficient test coverage for the previous PR.

[JackPGreen]

test really is mandatory for a bug like this

@elharo I've added a test in e1a7473, based on a slimmed-down version of the reproducer.

This test fails on master:

[ERROR] Failed to execute goal org.apache.maven.plugins:maven-javadoc-plugin:3.11.4-SNAPSHOT:aggregate (default-cli) on project GITHUB-1242: An error has occurred in Javadoc report generation:

But passes on my branch:

[INFO] Building: GITHUB-1242/pom.xml
[INFO] GITHUB-1242/pom.xml .............................. SUCCESS (1.861 s)

[JackPGreen]

@olamy thanks for re-running the test, I've updated the reproducer to (hopefully) handle the failing legacy Maven cases.

292736e

[olamy]

LGTM. I have been tested this on a large project such https://github.com/jetty/jetty.project with mvn clean install -Pjavadoc-aggregate javadoc:aggregate -DskipTests
and all good
Thanks for your contribution

[github-actions]

@olamy Please assign appropriate label to PR according to the type of change.

[fridrich]

The problem of the sourcepath is that it sucks in the module-info.java files in aggregate goal. I am getting this for instance:
[ERROR] Failed to execute goal org.apache.maven.plugins:maven-javadoc-plugin:3.11.3:aggregate (default-cli) on project angus-activation-project: An error has occurred in Javadoc report generation: [ERROR] Exit code: 1 [ERROR] /home/abuild/rpmbuild/BUILD/angus-activation-2.0.2-build/angus-activation-2.0.2/activation-registry/src/main/java/module-info.java:17: error: module not found: jakarta.activation [ERROR] requires transitive jakarta.activation; [ERROR] ^ [ERROR] error: cannot access module-info [ERROR] cannot resolve modules [ERROR] 2 errors [ERROR] Command line was: /usr/lib64/jvm/java-21-openjdk-21/bin/javadoc -J-Duser.language= -J-Duser.country= @options @argfile

[fridrich]

I don't know how to do that differently. Excluding the module-info.java files from the source list was easy. Since the javadoc tool only can exclude packages or -- in the standard doclet -- the subdirectories, there is no easy way to prevent the module-info.java being sucked in. I am just wondering whether there was no other way to fix your bug then by reverting this one.

[fridrich]

But what I tried also to avoid in the original commit was not to refer to the sources by package. Maybe I missed a place. If one does not use the package names in legacy mode, one does not need the sourcepath