This report has been sent

Apache Gump is a cross-project continuous integration server. It is different from "usual" CI servers in that it expects the individual project builds to succeed; its purpose is to check the integration of a project with the latest code rather than a fixed version of the project's dependencies. If you want a more traditional nightly build server, Gump is not for you. Use Gump if you want to know when a change in your dependencies breaks your project or when your changes break other projects.

Gump's intention isn't so much to be a CI server but rather a vehicle that makes people look beyond their project's boundaries and helps the projects to collaborate.

Gump is written in Python and supports several build tools and version control systems. The Apache installation of Gump builds many ASF projects and their dependencies. It started in the Java part of the foundation but also builds projects like APR, HTTPd and log4net.


Big infrastructure changes, light development, no issues.


There are no Board level issues.


The Gump project really consists of two parts, the code base for the project and the ASF installations[1] running this code base to build many ASF projects as well as some related projects.

The code base mostly does what its current users need so there isn't much development going on at all. No new committers have been added.

All ASF committers have write access to the metadata that configure the ASF installations. There are a few people contributing across all projects and a few additional people maintaining the metadata of the projects they are interested in the most.

No changes to the PMC.


While migrating to the new servers a few issues with Gump's database access have been identified and fixed. A new "builder" has been added that removes the boilerplate code previously required when installing a file to the local Maven repository.

We've managed to build a few projects that have been failing for a long time in Gump - among them the ASF projects Forrest, Lucene, Cactus and big parts of Cocoon. We've also added builds for Solr, Tika and PDFBox.

Some projects that have been failing for a long time and will likely never become buildable again have been removed. Also we've disabled a few builds (mostly running some sort of tests) that caused Gump to hang for an hour (Gump's timeout for build processes).


The ASF installations of Gump work on the latest code base almost all of the time. The project is in a state of a perpetual beta. There have been no releases.


Our main machine vmgump has been replaced by a brand new virtual Ubuntu machine. The old database has been migrated to keep history. Since the new machine now runs OpenJDK6 rather than an "official" Java environment a few dependencies on Sun VMs have shown up in some project's builds.

We used to have a Solaris zone which has now been replaced by a FreeBSD jail. The installation is working very well and we see almost the same build failures and successes on FreeBSD as on Linux.

Many thanks to the infrastructure team for the support during the migration - and for all the other stuff you do.


As of Sun, 12 Sep 2010 the ASF installations check out a bit less than 200 source trees (113 from the ASF repository) and try to build a bit less than 600 "projects". A complete Gump run takes more than nine hours on vmgump and eight on the FreeBSD jail.

[1] the main instance at and a FreeBSD jail at

Drafts/BoardReports/20100922 (last edited 2010-09-12 04:48:47 by StefanBodewig)