summaryrefslogblamecommitdiff
path: root/_posts/2010-09-27-jitterbug.textile
blob: f8d315b7f98a42d78f7c38c5d80763400f5b29c1 (plain) (tree)


























                                                                                                                                                                                                                  

                                                        




















                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      
---
layout: post
category: perl
title: jitterbug - a first step to continuous integration for Perl modules on GitHub
---

bq. If you'd like to be a jitter bug,
First thing you must do is get a jug,
Put whiskey, wine and gin within,
And shake it all up and then begin.


Earlier this month sukria set up a "shell script":http://github.com/sukria/capsule to run the Dancer's tests suit. The script is simple, but does what we want:

 * every 15 minutes, it pulls from the github repository
 * for every perl installed via perlbrew, it executes the tests suite
 * it stores the results in text file
 * finally, it creates a code coverage report for the master branch

The only problem I had with this is that it wasn't really easy to find the appropriate test report you might want to check.

That's when I decided to write an interface to this: "jitterbug":https://github.com/franckcuny/jitterbug.

h3. demo

"You can check Dancer's version":http://jitterbug.perldancer.org/. The interface is really simple: there's a list of repositories, the status of the last build, and a link to a list of all the project's builds.

<center>!/static/imgs/jitterbug.png(jitterbug)!</center>

h3. how it works

For each project you want to use with jitterbug, you set the url of the HTTP hook in the administration page of your project. Each time a push is detected, GH will send you a notification (you can see details of the content). If the project doesn't already exist in your setup, it will be created, so you don't need to maintain or update a configuration file each time you create a new repository. The notification creates a task in a queue, and a script pulls the task and executes it.

bq. If you don't want anyone to use your setup for building their modules, you can protect the */hook/* path with a .htaccess, and use the following kind of url : http://user:pass@.../

The script (*builder.pl*, provided with JB) will clone the repository, switch to the given commit and run the tests. It works both with classic distribution having a Makefile.PL and Dist::Zilla setups. The output is stored on the disk and some information about the project will be updated. "You can see the result for Dancer":http://jitterbug.perldancer.org/project/Dancer.

This script relies on two important things: "perlbrew":http://github.com/gugod/App-perlbrew and "cpanminus":http://github.com/miyagawa/cpanminus. So your tests will be executed against every Perl version you have installed and thanks to cpanminus, all the deps will be installed too.

If the test fails, a mail is sent to the author.

h3. what's next

A list of some features I want to have:

 * IRC notifications (something like: [author's name]: link to the commit - commit message - link to the build)
 * customizable builder script (so you can add your own hook to building a package, having private notifications, etc.)
 * simple administration (restart a build, etc.)

(thanks to "sawyer":http://github.com/xsawyerx and "sukria":http://github.com/sukria/).