Configuration as code round trip test fails with parent pom 4.35

40 views
Skip to first unread message

Mark Waite

unread,
Feb 21, 2022, 9:43:47 AM2/21/22
to Jenkins Developers
The git plugin includes configuration as code round trip tests like this one.  I like those tests because they confirm that configuration as code is well behaved with plugin use cases.

One of the round trip tests fails on Windows with the upgrade from parent pom 4.34 to parent pom 4.35.  I've confirmed that the failure is consistent.  I can see it on my Windows desktop computer and in the CI job.

The failure message reports

java.lang.NoClassDefFoundError: jenkins/plugins/git/traits/package-info (wrong name: jenkins\plugins\git\traits\package-info)

That looks suspiciously like a Windows specific issue with directory separators and might be a mistake attempting to call the constructor of the file "package-info".

Any recommendations for a preferred solution?
Mark Waite

Jesse Glick

unread,
Feb 21, 2022, 1:14:36 PM2/21/22
to jenkin...@googlegroups.com
Following the link trail, this is https://issues.apache.org/jira/browse/MCOMPILER-485 so I think https://github.com/jenkinsci/plugin-pom/pull/504 should either be reverted until https://github.com/apache/maven-compiler-plugin/pull/95 is released, or the new feature explicitly disabled.

Mark Waite

unread,
Feb 21, 2022, 2:53:50 PM2/21/22
to Jenkins Developers
On Monday, February 21, 2022 at 11:14:36 AM UTC-7 Jesse Glick wrote:
Following the link trail, this is https://issues.apache.org/jira/browse/MCOMPILER-485 so I think https://github.com/jenkinsci/plugin-pom/pull/504 should either be reverted until https://github.com/apache/maven-compiler-plugin/pull/95 is released, or the new feature explicitly disabled.

Thanks very much.  Based on that, I added a property to the pom file

  <maven.compiler.createMissingPackageInfoClass>false</maven.compiler.createMissingPackageInfoClass>

Test that was failing on my Windows computer is now passing.

I'll remove that property once the bug is fixed in the maven compiler plugin.  Thanks also to Olivier Lamy for his work on the issue in the maven compiler plugin.

Jesse Glick

unread,
Feb 21, 2022, 4:49:17 PM2/21/22
to jenkin...@googlegroups.com
On Mon, Feb 21, 2022 at 2:53 PM Mark Waite <mark.ea...@gmail.com> wrote:
Based on that, I added a property to the pom file

  <maven.compiler.createMissingPackageInfoClass>false</maven.compiler.createMissingPackageInfoClass>

Test that was failing on my Windows computer is now passing.

Good. Would you mind filing a PR to `plugin-pom` accordingly, so we can push out a release for the regression?

Mark Waite

unread,
Feb 21, 2022, 7:18:39 PM2/21/22
to Jenkins Developers


On Monday, February 21, 2022 at 2:49:17 PM UTC-7 Jesse Glick wrote:
On Mon, Feb 21, 2022 at 2:53 PM Mark Waite wrote:
Based on that, I added a property to the pom file

  <maven.compiler.createMissingPackageInfoClass>false</maven.compiler.createMissingPackageInfoClass>

Test that was failing on my Windows computer is now passing.

Good. Would you mind filing a PR to `plugin-pom` accordingly, so we can push out a release for the regression?

https://github.com/jenkinsci/plugin-pom/pull/506 submitted.  Let me know if that's not the right solution for this temporary issue. 

Jesse Glick

unread,
Feb 22, 2022, 11:47:15 AM2/22/22
to jenkin...@googlegroups.com
To follow up, 4.36 should fix this.
Reply all
Reply to author
Forward
0 new messages