Webtools/leeroy

LeeroyBot runs tests pull requests before they're merged. Read more about it on the WebDev Blog.

Getting Leeroy set up for your project

• Create a new job on ci.mozilla.org
  • job name '[your job name]-github'
  • copy existing job
• Configure your new job
  • Job Notifications
    • Add Endpoint
      • HTTP
      • http://leeroybot.herokuapp.com/notification/jenkins
    • Check "This build is parameterized"
      • add String Parameter
        • name "GIT_BASE_REPO"
        • default value: mozilla/[project name]
      • add String Parameter
        • name "GIT_HEAD_REPO"
        • default value: mozilla/[project name]
      • add String Parameter
        • name "GIT_SHA1"
        • default value: "master"
      • add String Parameter
        • name "GITHUB_URL"
        • default value: *leave blank*
  • Source Code Management
    • Git
      • repository url: "git://github.com/$GIT_HEAD_REPO.git"
      • banches to build
        • Branch Specifier: "$GIT_SHA1"
  • remove any build artifact archiving and fingerprinting
  • (optional) remove any notifications to IRC or e-mail
  • (optional) remove any non-test steps such as coverage, leeroy only cares about pass / fail
• Configure Github
  • Add LeeroyBot (https://github.com/leeroybot ) to your team if he isn't already
    • necessary to set up hooks that cannot be set manually from the UI
• File a bug in your product
  • assign to :Lonnen
  • summary "hook up leeroy"?
  • include
    • github_repo: "mozilla/[your project]"
    • jenkins_job_name: "[your project]-github"
• After Leeroy has been updates
  • Lonnen will assign the bug back to you
  • open a pull request that should not break the build
    • make a trivial documentation change
    • close after it passes
  • open a pull request that will break the build
    • delete models.py
    • close after it fails
  • link both in the bug and mark it resolved -> fixed
  • It's important that these requests are opened or updated /after/ intergration is established.

Troubleshooting

Leeroy listens to events as they happen with no queue or storage. If a network event is dropped and something gets stuck in a strange state, the only thing you can do is change the sha or open a new PR. There is an issue open on leeroy to change this. Until then, I suggest adding a trivial commit (add some whitespace to your readme) and rerunning, then popping that commit off the PR with a rebase after it passes.