The Ansible development cycle happens on two levels. At a macro level, the team plans releases and tracks progress with roadmaps and projects. At a micro level, each PR has its own lifecycle.
If you want to follow the conversation about what features will be added to Ansible for upcoming releases and what bugs are being fixed, you can watch these resources:
Ansible accepts code via **pull requests** ("PRs" for short). GitHub provides a great overview of `how the pull request process works <https://help.github.com/articles/about-pull-requests/>`_ in general. The ultimate goal of any pull request is to get merged and become part of Ansible Core.
Because Ansible receives many pull requests, and because we love automating things, we've automated several steps of the process of reviewing and merging pull requests with a tool called Ansibullbot, or Ansibot for short.
Ansibullbot runs continuously. You can generally expect to see changes to your issue or pull request within thirty minutes. Ansibullbot examines every open pull request in the repositories, and enforces state roughly according to the following workflow:
- If a pull request has no workflow labels, it's considered **new**. Files in the pull request are identified, and the maintainers of those files are pinged by the bot, along with instructions on how to review the pull request. (Note: sometimes we strip labels from a pull request to "reboot" this process.)
- If the module maintainer is not ``$team_ansible``, the pull request then goes into the **community_review** state.
- If the module maintainer is ``$team_ansible``, the pull request then goes into the **core_review** state (and probably sits for a while).
- If the pull request is in **community_review** and has received comments from the maintainer:
- If the maintainer says ``shipit``, the pull request is labeled **shipit**, whereupon the Core team assesses it for final merge.
- If the maintainer says ``needs_info``, the pull request is labeled **needs_info** and the submitter is asked for more info.
- If the maintainer says **needs_revision**, the pull request is labeled **needs_revision** and the submitter is asked to fix some things.
- If the submitter says ``ready_for_review``, the pull request is put back into **community_review** or **core_review** and the maintainer is notified that the pull request is ready to be reviewed again.
- If the pull request is labeled **needs_revision** or **needs_info** and the submitter has not responded lately:
- The submitter is first politely pinged after two weeks, pinged again after two more weeks and labeled **pending action**, and the issue or pull request will be closed two weeks after that.
- If the submitter responds at all, the clock is reset.
- If the pull request is labeled **community_review** and the reviewer has not responded lately:
- The reviewer is first politely pinged after two weeks, pinged again after two more weeks and labeled **pending_action**, and then may be reassigned to ``$team_ansible`` or labeled **core_review**, or often the submitter of the pull request is asked to step up as a maintainer.
- If Shippable tests fail, or if the code is not able to be merged, the pull request is automatically put into **needs_revision** along with a message to the submitter explaining why.
There are corner cases and frequent refinements, but this is the workflow in general.
-**backport**: this is applied automatically if the PR is requested against any branch that is not devel. The bot immediately assigns the labels backport and ``core_review``.
-**bugfix_pull_request**: applied by the bot based on the templatized description of the PR.
-**cloud**: applied by the bot based on the paths of the modified files.
-**docs_pull_request**: applied by the bot based on the templatized description of the PR.
-**easyfix**: applied manually, inconsistently used but sometimes useful.
-**feature_pull_request**: applied by the bot based on the templatized description of the PR.
-**networking**: applied by the bot based on the paths of the modified files.
-**owner_pr**: largely deprecated. Formerly workflow, now informational. Originally, PRs submitted by the maintainer would automatically go to **shipit** based on this label. If the submitter is also a maintainer, we notify the other maintainers and still require one of the maintainers (including the submitter) to give a **shipit**.
-**pending_action**: applied by the bot to PRs that are not moving. Reviewed every couple of weeks by the community team, who tries to figure out the appropriate action (closure, asking for new maintainers, etc).
**Note:**`new_plugin` kicks off a completely separate process, and frankly it doesn't work very well at present. We're working our best to improve this process.
Human PR review
---------------
After Ansibot reviews the PR and applies labels, the PR is ready for human review. The most likely reviewers for any PR are the maintainers for the module that PR modifies.
Each module has at least one assigned :ref:`maintainer <maintainers>`, listed in the `BOTMETA.yml <https://github.com/ansible/ansible/blob/devel/.github/BOTMETA.yml>`_ file.
The maintainer's job is to review PRs that affect that module and decide whether they should be merged (``shipit``) or revised (``needs_revision``). We'd like to have at least one community maintainer for every module. If a module has no community maintainers assigned, the maintainer is listed as ``$team_ansible``.
Once a human applies the ``shipit`` label, the :ref:`committers <community_committer_guidelines>` decide whether the PR is ready to be merged. Not every PR that gets the ``shipit`` label is actually ready to be merged, but the better our reviewers are, and the better our guidelines are, the more likely it will be that a PR that reaches **shipit** will be mergeable.
Making your PR merge-worthy
===========================
We don't merge every PR. Here are some tips for making your PR useful, attractive, and merge-worthy.
.._community_changelogs:
Changelogs
----------
Changelogs help users and developers keep up with changes to Ansible.
Ansible builds a changelog for each release from fragments. You **must** add a changelog fragment to any PR that changes functionality or fixes a bug.
You don't have to add a changelog fragment for PRs that add new
modules and plugins, because our tooling does that for you automatically.
We build short summary changelogs for minor releases as well as for major releases. If you backport a bugfix, include a changelog fragment with the backport PR.
.._changelogs_how_to:
Creating a changelog fragment
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
A basic changelog fragment is a ``.yaml`` file placed in the
``changelogs/fragments/`` directory. Each file contains a yaml dict with
keys like ``bugfixes`` or ``major_changes`` followed by a list of
changelog entries of bugfixes or features. Each changelog entry is
rst embedded inside of the yaml file which means that certain
constructs would need to be escaped so they can be interpreted by rst
and not by yaml (or escaped for both yaml and rst if that's your
desire). Each PR **must** use a new fragment file rather than adding to
an existing one, so we can trace the change back to the PR that introduced it.
To create a changelog entry, create a new file with a unique name in the ``changelogs/fragments/`` directory. The file name should include the PR number and a description of the change. It must end with the file extension ``.yaml``. For example: ``40696-user-backup-shadow-file.yaml``
A single changelog fragment may contain multiple sections but most will only contain one section.
The toplevel keys (bugfixes, major_changes, etc) are defined in the
`config file <https://github.com/ansible/ansible/blob/devel/changelogs/config.yaml>` for our release note tool. Here are the valid sections and a description of each:
**major_changes**
Major changes to Ansible itself. Generally does not include module or plugin changes.
**minor_changes**
Minor changes to Ansible, modules, or plugins. This includes new features, new parameters added to modules, or behavior changes to existing parameters.
**deprecated_features**
Features that have been deprecated and are scheduled for removal in a future release.
**removed_features**
Features that were previously deprecated and are now removed.
**bugfixes**
Fixes that resolve issues. If there is a specific issue related to this bugfix, add a link in the changelog entry.
**known_issues**
Known issues that are currently not fixed or will not be fixed.
Most changelog entries will be ``bugfixes`` or ``minor_changes``. When writing a changelog entry that pertains to a particular module, start the entry with ``- [module name] -`` and include a link to the related issue if one exists.
Here are some examples:
..code-block:: yaml
bugfixes:
- win_updates - fixed issue where running win_updates on async fails without any error
..code-block:: yaml
minor_changes:
- lineinfile - add warning when using an empty regexp (https://github.com/ansible/ansible/issues/29443)
..code-block:: yaml
bugfixes:
- copy module - The copy module was attempting to change the mode of files for
remote_src=True even if mode was not set as a parameter. This failed on
filesystems which do not have permission bits.
You can find more example changelog fragments in the `changelog directory <https://github.com/ansible/ansible/tree/stable-2.6/changelogs/fragments>`_ for the 2.6 release. You can also find documentation of the format, including hints on embedding rst in the yaml, in the `reno documentation <https://docs.openstack.org/reno/latest/user/usage.html#editing-a-release-note>`_.
Once you've written the changelog fragment for your PR, commit the file and include it with the pull request.
Backporting merged PRs
======================
All Ansible PRs must be merged to the ``devel`` branch first.
After a pull request has been accepted and merged to the ``devel`` branch, the following instructions will help you create a
pull request to backport the change to a previous stable branch.
We do **not** backport features.
..note::
These instructions assume that:
*``stable-2.7`` is the targeted release branch for the backport
*``https://github.com/ansible/ansible.git`` is configured as a
``git remote`` named ``upstream``. If you do not use
a ``git remote`` named ``upstream``, adjust the instructions accordingly.
