search

jenkinsci / winstone

Patched winstone used in Jenkins
https://jenkins.io
Other
88 stars 72 forks source link

Remove support for multiple WAR files #377

Closed basil closed 5 months ago

basil commented 5 months ago

Winstone contains 100 or so lines of code for a --webappsDir switch to run a directory full of WAR files, each with their own request logger which logs requests to a different file. There are no references to this in the jenkinsci, jenkins-infra, and cloudbees GitHub organizations. There is no test coverage for the feature. The single reference in Jira is JENKINS-5307, from 2010. While I am aware of people running Jenkins as a WAR file in Tomcat alongside other WAR files, I am not aware of anyone using Winstone to run multiple WAR files. I think it is quite likely there are zero people using this feature.

This all is changing in Jetty 12, which now does request logging at a per-Server level, not a per-webapp level. Thus the current code would require significant changes to support the existing functionality. Since it has no test coverage, and since we think there are zero people using it, this PR removes the feature. That way, we won't have to port it to Jetty 12, and we also won't have to manually test the result.

Proposed changelog and upgrade guide entries

The --webappsDir argument to run Winstone with a directory full of WAR files has been removed without replacement.

Testing done

mvn clean verify

basil commented 5 months ago

If we feel that removal without first deprecating this option is too aggressive, then I can file a separate PR to first deprecate it and then remove it in a few months. However, based on the research I provided in the PR description, I think it is highly likely there are zero people using this feature and that it can be treated as dead code and removed without a prior deprecation period.

MarkEWaite commented 5 months ago

If we feel that removal without first deprecating this option is too aggressive, then I can file a separate PR to first deprecate it and then remove it in a few months. However, based on the research I provided in the PR description, I think it is highly likely there are zero people using this feature and that it can be treated as dead code and removed without a prior deprecation period.

I think it should be removed without deprecation.