All Downloads are FREE. Search and download functionalities are using the official Maven repository.

hudson.maven.package.html Maven / Gradle / Ivy

Go to download

This plug-in provides deep integration of Hudson and Maven. This functionality used to be part of the Hudson core. Now it is a plug-in that is installed by default, but can be disabled.

There is a newer version: 2.2.0
Show newest version



Maven support.

General Idea

One of the pain points of the freestyle project is that you have to configure a lot of things, such as where to look for test reports, what files to archive, where the findbugs report would go. But if we focus on Maven, we should be able to eliminate much of the configuration, since it introduces more uniform structures. So that's what this plugin does — at the expense of limiting the build tool to Maven, automate much of the configuration.

Implementation Approach

The core idea of the implementation is to monitor what Maven does, so that we can see which mojos are executed with what parameters. In this way, we can tell when/where javadoc is generated, if source code compilation had an error, and access other rich information about the project build process.

To make communication between Hudson JVM and Maven JVM easier, we use the remoting technology that Hudson uses between the master and the slave. We start a new JVM and bootstraps to the remoting, then use a socket to establish a connection to this process. This part of the code is in the "maven-agent" module. We then bootstrap Maven.

To intercept what's going on in Maven, we extend some key components in Maven, and configure Plexus in such a way that our components are used instead of default ones. Because injected components need to live in a different classloader, they are packaged in a separate "maven-interceptor" module. We also bring in objects (MavenReporters) from plugins via remoting, and distribute intercepted events to these guys. They can then digest information and send it back to Hudson JVM.

In addition to all this, we use embedded Maven to parse POMs, so that we can figure out the structure of the project before we even do a build (this information is used for example to set up dependencies among jobs.) This turns out to be rather fragile (in the presence of profiles that are activated by system property, platform, etc., which makes the effective POM different when in Hudson vs when built for real.)





© 2015 - 2024 Weber Informatics LLC | Privacy Policy