Details
-
Task
-
Status: Resolved (View Workflow)
-
Minor
-
Resolution: Fixed
-
None
-
-
2.14.4
Description
as per https://github.com/jenkinsci/subversion-plugin/pull/254#issuecomment-858288074 reported by basil
There looks like there is a regression (in waiting?) in the Subversion plugin against Pipeline and Jenkins core. (and potentially the other plugins in which Digester2 was removed.
Attachments
Issue Links
Activity
basil I was able to confirm this manually. I attempted to reproduce the failing test with subversion 2.14.0, 2.14.1, and then a SNAPSHOT with asm excluded from Subversion and I still got the exact same behavior in all three tries.
Basil Crow
added a comment - That's great news! So I think the main action item from my perspective is to get the tests for any plugin that depends on the Subversion plugin working again, especially if those tests are run as part of the BOM's PCT matrix. So far, I see that such tests are present in workflow-multibranch, workflow-cps-global-lib, and workflow-scm-step.
Basil Crow
added a comment - That's great news! So I think the main action item from my perspective is to get the tests for any plugin that depends on the Subversion plugin working again, especially if those tests are run as part of the BOM's PCT matrix. So far, I see that such tests are present in workflow-multibranch , workflow-cps-global-lib , and workflow-scm-step . 
Olivier Lamy
added a comment - for the record, the issue do not happen on real Jenkins instance but only for some unit tests. (for what I have been able to test subversion plugin)
Olivier Lamy
added a comment - for the record, the issue do not happen on real Jenkins instance but only for some unit tests. (for what I have been able to test subversion plugin) I've changed the type from Bug to Task as it seems there is no real regression in the production environment, but in test one.
digester has been removed in https://github.com/jenkinsci/subversion-plugin/pull/259
Technically not completed until workflow-scm-step takes in the new subversion
Since the problematic frames in the stack trace originate with Stapler, the impact of the regression should be limited to JenkinsRule-based tests only, not production (based on the reasoning James and I agreed on in the mailing list thread). Several of the Pipeline plugins depend on Subversion in their tests, though, and the BOM runs PCT on those Pipeline plugins, so we need to find some way to get these tests working again before we start a new BOM line for the LTS release after 2.289.