The 'Continuous Integration with Hudson' book build process
Continuous Integration with Hudson is a new open source book project in the works. In the spirit of 'eating our own dogfood', this book is produced using Hudson. In this article, I discuss the build framework used for the 'Continuous Integration with Hudson' book.
The 'Continuous Integration with Hudson' book is written in docbook, and is therefore XML source code which builds to PDF and HTML versions of the book. The source code is stored using git (in a github repository which will be eventually made public). The main branches are the following:
- master - the main development/integration branch
- review - review candidate versions go here
- release - the current latest release version
There is also a branch for each chapter, which gets merged into the master branch when it is ready to be integrated into the whole book. When they are ready for review, chapters are merged into the review branch. A team of volunteer technical reviews the review candidate, and fixes are integrated into the review branch, and then ported back into the master branch. When the review branch is ready, the new chapter is merged into the release branch.
Needless to say, git merging and branching makes it much easier to update additions and corrections across branches - it is trivially easy to update the review candidate branch with the latest chapter, and to move corrections made during the review process from the review branch back to the master branch, for example.
The docbook XML source code is built into PDF and HTML form using Maven. This process is based on the one developed by Tim O'Brian for the excellent Maven and Nexus books, and it works well.
The build process is automated using Hudson. There is a build job on each of the main branches (master, review and release, and the more long-lived chapter branches). The Hudson build job builds PDF and HTML versions of the source code. It also checks for docbook errors and warning such as missing figures or cross-references, and keeps track of TODO tags (using the 'Open Tasks' plugin).
The 'release' build job automatically deploys the latest version of the book to the 'Continuous Integration with Hudson' web page (using the SCP plugin), so the version found here is always guaranteed to be the latest and greatest.
You can see a high-level overview of this architecture here:
This process is still evolving. Things coming up soon on the roadmap include archiving and deploying the HTML version of the book as well as the PDF, and integrating an automatic spell checker in the Hudson build process.