Checklist (and a short version for the impatient)
Make commits of logical units.
Check for unnecessary whitespace with "git diff --check" before committing.
Commit using Unix line endings (check the settings around "crlf" in git-config(1)).
Do not check in commented out code or unneeded files.
The first line of the commit message should be a short description (50 characters is the soft limit, excluding ticket number(s)), and should skip the full stop.
If there is an associated Redmine ticket then the first line should include the ticket number in the form "(#XXXX) Rest of message".
The body should provide a meaningful commit message, which:
uses the imperative, present tense: "change", not "changed" or "changes".
includes motivation for the change, and contrasts its implementation with the previous behavior.
Make sure that you have tests for the bug you are fixing, or feature you are adding.
Make sure the test suite passes after your commit (rake spec; see below for more).
Fork the repository on GitHub.
Push your changes to a topic branch in your fork of the repository.
Submit a pull request to the repository in the puppetlabs organization.
Mail patches to puppet-dev mailing list using
rake mail_patches, or
Attach patches to Redmine ticket.
The long version
Decide what to base your work on.
In general, you should always base your work on the oldest branch that your change is relevant to.
A bug fix should be based on the current stable series. If the bug is not present in the current stable release, then base it on
A new feature should be based on
Security fixes should be based on the current maintenance series (that is, the previous stable series). If the security issue was not present in the maintenance series, then it should be based on the current stable series if it was introduced there, or on
masterif it is not yet present in a stable release.
The current stable series is 2.7.x, and the current maintenance series is 2.6.x.
Make separate commits for logically separate changes.
Please break your commits down into logically consistent units which include new or changed tests relevent to the rest of the change. The goal of doing this is to make the diff easier to read for whoever is reviewing your code. In general, the easier your diff is to read, the more likely someone will be happy to review it and get it into the code base.
If you're going to refactor a piece of code, please do so as a separate commit from your feature or bug fix changes.
We also really appreciate changes that include tests to make sure the bug isn't re-introduced, and that the feature isn't accidentally broken.
Describe the technical detail of the change(s). If your description starts to get too long, that's a good sign that you probably need to split up your commit into more finely grained pieces.
Commits which plainly describe the the things which help reviewers check the patch and future developers understand the code are much more likely to be merged in with a minimum of bike-shedding or requested changes. Ideally, the commit message would include information, and be in a form suitable for inclusion in the release notes for the version of Puppet that includes them.
Please also check that you are not introducing any trailing whitespaces or other "whitespace errors". You can do this by running "git diff --check" on your changes before you commit.
Sign the Contributor License Agreement
Before we can accept your changes, we do need a signed Puppet Labs Contributor License Agreement (CLA).
You can access the CLA via the Contributor License Agreement link in the top menu bar of our Redmine instance. Once you've signed the CLA, a badge will show up next to your name on the Puppet Project Overview Page, and your name will be listed under "Contributor License Signers" section.
If you have any questions about the CLA, please feel free to contact Puppet Labs via email at firstname.lastname@example.org.
Sending your patches
We accept multiple ways of submitting your changes for inclusion. They are listed below in order of preference.
Please keep in mind that any method that involves sending email to the mailing list directly requires you to be subscribed to the mailing list, and that your first post to the list will be held in a moderation queue.
GitHub Pull Requests
To submit your changes via a GitHub pull request, we highly recommend that you have them on a topic branch, instead of directly on "master" or one of the release, or RC branches. It makes things much easier to keep track of, especially if you decide to work on another thing before your first change is merged in.
In general, after pushing your topic branch up to your repository on GitHub, you'll switch to the branch in the GitHub UI and click "Pull Request" towards the top of the page in order to open a pull request.
You'll want to make sure that you have the appropriate destination branch in the repository under the puppetlabs organization. This should be the same branch that you based your changes off of.
Other pull requests
If you already have a publicly accessible version of the repository hosted elsewhere, and don't wish to or cannot use GitHub, you can submit your change by requesting that we pull the changes from your repository by sending an email to the puppet-dev Google Groups mailing list.
git-request-pull(1)provides a handy way to generate the text for the email requesting that we pull your changes (and does some helpful sanity checks in the process).
Mailing patches to the mailing list
If neither of the previous methods works for you, then you can also mail the patches inline to the puppet-dev Google Group using either
rake mail_patches, or by using
rake mail_patcheshandles setting the appropriate flags to
git-send-email(1)for you, but doesn't allow adding any commentary between the '---', and the diffstat in the resulting email. It also requires that you have created your topic branch in the form
If you decide to use
git-send-email(1)directly, please be sure to use the following flags for
git-format-patch(1): -C -M -s -n --subject-prefix='PATCH/puppet'
Attaching patches to Redmine
As a method of last resort you can also directly attach the output of
git-diff(1)to a Redmine ticket.
If you are generating the diff outside of Git, please be sure to generate a unified diff.
Update the related Redmine ticket.
If there's a Redmine ticket associated with the change you submitted, then you should update the ticket to include the location of your branch, and change the status to "In Topic Branch Pending Merge", along with any other commentary you may wish to make.
Running the Tests
You will need a fresh VM with only the gems described here installed. Ruby does not play nicely with dependencies. Consider using vagrant or an EC2 instance. These instructions are for a CentOS 6.3 VM, but similar-vintage distros should work.
Install rubygems and the following gems (with
* rake * mysql (this can also be installed as a system package)
Install a MySQL server, and create a test database and user account.
config/database.yml and edit its test
section to include your new MySQL credentials.
spec/spec.opts. You can edit this
file, but the defaults should do.
Next, initialize your test database:
RAILS_ENV=test rake db:migrate.
./script/spec spec to actually run the tests.
How to track the status of your change after it's been submitted
Shortly after opening a pull request on GitHub, there should be an automatic message sent to the puppet-dev Google Groups mailing list notifying people of this. This notification is used to let the Puppet development community know about your requested change to give them a chance to review, test, and comment on the change(s).
If you submitted your change via manually sending a pull request or mailing the patches, then we keep track of these using patchwork. When code is merged into the project it is automatically removed from patchwork, and the Redmine ticket is manually updated with the commit SHA1. In addition, the ticket status must be updated by the person who merges the topic branch to a status of "Merged - Pending Release"
We do our best to comment on or merge submitted changes within a week. However, if there hasn't been any commentary on the pull request or mailed patches, and it hasn't been merged in after a week, then feel free to ask for an update by replying on the mailing list to the automatic notification or mailed patches. It probably wasn't intentional, and probably just slipped through the cracks.
If you have commit access to the repository
Even if you have commit access to the repository, you'll still need to go through the process above, and have someone else review and merge in your changes. The rule is that all changes must be reviewed by a developer on the project (that didn't write the code) to ensure that all changes go through a code review process.
Having someone other than the author of the topic branch recorded as performing the merge is the record that they performed the code review.
Merging topic branches
When merging code from a topic branch into the integration branch (Ex: master, 2.7.x, 1.6.x, etc.), there should always be a merge commit. You can accomplish this by always providing the
git merge --no-ff --log tickets/master/1234-fix-something-broken
The reason for always forcing this merge commit is that it provides a consistent way to look up what changes & commits were in a topic branch, whether that topic branch had one, or 500 commits. For example, if the merge commit had an abbreviated SHA-1 of
coffeebad, then you could use the following
git loginvocation to show you which commits it brought in:
git log coffeebad^1..coffeebad^2
The following would show you which changes were made on the topic branch:
git diff coffeebad^1...coffeebad^2
Because we always merge the topic branch into the integration branch the first parent (
^1) of a merge commit will be the most recent commit on the integration branch from just before we merged in the topic, and the second parent (
^2) will always be the most recent commit that was made in the topic branch. This also serves as the record of who performed the code review, as mentioned above.