Developer’s Guide

The goal of this document is to walk you through the concepts and specifics that should be understood while contributing to projects hosted in the OpenDev infrastructure.

Accounts creation

The steps necessary to create a Gerrit account are described in Getting Started. Only extra recommended steps will be described here.

IRC Account

While development on OpenDev only requires an account on the OpenDev Gerrit Code Review System, effective development in OpenDev hosted projects often requires interacting with other developers in IRC channels on Freenode. It is recommended to start by getting set up on IRC so that one can ask questions if one encounters issues with other phases of account setup.

If you do not know how to connect to Freenode, the Connecting to Freenode document will help.

It is highly recommended to Register your IRC Nick.

There are times when the OpenDev team needs to configure channels to only allow people to join who are using a Registered Nick. Registering a Nick also prevents someone else from taking a Nick that people come to know you by.

You can find the OpenDev community in the #opendev IRC channel on Freenode.

For further information about the use of IRC in OpenStack, see IRC Guide.

Extra Gerrit account setup steps

OpenStack Individual Contributor License Agreement

Projects that are part of OpenStack project require signing the Individual Contributor License Agreement, see these detailed instructions.

Accessing Gerrit over HTTPS

Git-review normally communicates with Gerrit using SSH over port 29418 with no further configuration needed. However, if you suspect that ssh over non-standards ports might be blocked (or you need to access the web using https) then you can configure git-review to use an https endpoint instead of ssh. Keep in mind that you will need to generate an HTTP password in Gerrit to use this connection. You should run the following command before “git review -s”:

git remote add gerrit https://<username>@review.opendev.org/<umbrella repository name>/<repository name>.git

In case you had already tried to setup git-review and it failed, it might be necessary to remove the Gerrit remote from git:

git remote rm gerrit

Development Workflow

Working on Bugs

Bug reports for a project are generally tracked on Launchpad at https://bugs.launchpad.net/<projectname>, or on StoryBoard ( https://storyboard.openstack.org). Contributors may review these reports regularly when looking for work to complete.

There are 4 key tasks with regards to bugs that anyone can do:

  1. Confirm new bugs: When a bug is filed, it is set to the “New” status. A “New” bug can be marked “Confirmed” once it has been reproduced and is thus confirmed as genuine.

  2. Solve inconsistencies: Make sure bugs are Confirmed, and if assigned that they are marked “In Progress”

  3. Review incomplete bugs: See if information that caused them to be marked “Incomplete” has been provided, determine if more information is required and provide reminders to the bug reporter if they haven’t responded after 2-4 weeks.

  4. Review stale In Progress bugs: Work with assignee of bugs to determine if the bug is still being worked on, if not, unassign them and mark them back to Confirmed or Triaged.

Learn more about working with bugs for various projects at:

https://wiki.openstack.org/wiki/BugTriage

Bug statuses are documented here:

https://docs.openstack.org/project-team-guide/bugs.html

If you find a bug that you wish to work on, you may assign it to yourself. When you upload a review, include the bug in the commit message for automatic updates back to Launchpad or StoryBoard. The following options are available for Launchpad:

Closes-Bug: #######
Partial-Bug: #######
Related-Bug: #######

and for StoryBoard:

Task: ######
Story: ######

Mentioning the story will create a handy link to the story from gerrit, and link to the gerrit patch in StoryBoard. Mentioning the task will change the task status in StoryBoard to ‘review’ while the patch is in review, and then ‘merged’ once the patch is merged. When all tasks in a story are marked ‘merged’, the story will automatically change status from ‘active’ to ‘merged’. If the patch is abandoned, the task status will change back to ‘todo’. It’s currently best to note both story and task so that the task status will update and people will be able to find the related story.

Also see the Including external references section of the OpenStack Git Commit Good Practices wiki page.

Working on Specifications and Blueprints

Many OpenStack project teams have a <projectteam>-specs repository which is used to hold approved design specifications for additions and changes to the project team’s code repositories.

The layout of the repository will typically be something like:

specs/<release>/

It may also have subdirectories to make clear which specifications are approved and which have already been implemented:

specs/<release>/approved
specs/<release>/implemented

You can typically find an example spec in specs/template.rst.

Check the repository for the project team you’re working on for specifics about repository organization.

Specifications are proposed for a given release by adding them to the specs/<release> directory and posting it for review. The implementation status of a blueprint for a given release can be found by looking at the blueprint in Launchpad. Not all approved blueprints will get fully implemented.

Specifications have to be re-proposed for every release. The review may be quick, but even if something was previously approved, it should be re-reviewed to make sure it still makes sense as written.

Historically, Launchpad blueprints were used to track the implementation of these significant features and changes in OpenStack. For many project teams, these Launchpad blueprints are still used for tracking the current status of a specification. For more information, see the Blueprints wiki page.

View all approved project team’s specifications at https://specs.openstack.org/.

Starting a change

The Getting Started page explains how to originally clone and prepare a git repository. This only has to be done once, as you can reuse the cloned repository for multiple changes.

Before creating your topic branch, just make sure you have the latest upstream changes:

git remote update
git checkout master
git pull --ff-only origin master

Git branch names

You may pick any name for your git branch names. By default, it will be reused as the topic for your change in Gerrit:

git checkout -b TOPIC-BRANCH

Best practices recommend, if you are working on a specific blueprint, to name your topic branch bp/BLUEPRINT where BLUEPRINT is the name of a blueprint in Launchpad (for example, bp/authentication). The general convention when working on bugs is to name the branch bug/BUG-NUMBER (for example, bug/1234567).

If you want to use a different gerrit topic name from the git branch name, you can use the following command to submit your change:

git review -t TOPIC

Committing changes

Git commit messages should start with a short 50 character or less summary in a single paragraph. The following paragraph(s) should explain the change in more detail.

If your changes addresses a blueprint or a bug, be sure to mention them in the commit message using the following syntax:

Implements: blueprint BLUEPRINT
Closes-Bug: ####### (Partial-Bug or Related-Bug are options)

For example:

Adds keystone support

...Long multiline description of the change...

Implements: blueprint authentication
Closes-Bug: #123456
Change-Id: I4946a16d27f712ae2adf8441ce78e6c0bb0bb657

Note that in most cases the Change-Id line should be automatically added by a Gerrit commit hook installed by git-review. If you already made the commit and the Change-Id was not added, do the Gerrit setup step and run: git commit --amend. The commit hook will automatically add the Change-Id when you finish amending the commit message, even if you don’t actually make any changes. Do not change the Change-Id when amending a change as that will confuse Gerrit.

Make your changes, commit them, and submit them for review:

git commit -a

Note

Do not check in changes on your master branch. Doing so will cause merge commits when you pull new upstream changes, and merge commits will not be accepted by Gerrit.

Using Signed-off-by

Projects may require the use of a Signed-off-by, and even if they do not, you are welcome to include Signed-off-by in your commits. By doing so, you are certifying that the following is true:

Developer's Certificate of Origin 1.1

By making a contribution to this project, I certify that:

(a) The contribution was created in whole or in part by me and I
    have the right to submit it under the open source license
    indicated in the file; or

(b) The contribution is based upon previous work that, to the best
    of my knowledge, is covered under an appropriate open source
    license and I have the right under that license to submit that
    work with modifications, whether created in whole or in part
    by me, under the same open source license (unless I am
    permitted to submit under a different license), as indicated
    in the file; or

(c) The contribution was provided directly to me by some other
    person who certified (a), (b) or (c) and I have not modified
    it.

(d) I understand and agree that this project and the contribution
    are public and that a record of the contribution (including all
    personal information I submit with it, including my sign-off) is
    maintained indefinitely and may be redistributed consistent with
    this project or the open source license(s) involved.

A Signed-off-by header takes the following form in a commit message:

Signed-off-by: Full Name <email@example.com>

If you add the -s option to git commit, this header will be added automatically:

git commit -s

Running Unit Tests

Before submitting your change, you should test it. To learn how to run python based unit tests in OpenStack projects see Running Python Unit Tests

Previewing a Change

Before submitting your change, you should make sure that your change does not contain the files or lines you do not explicitly change:

git show

Submitting a Change for Review

Once you have committed a change to your local repository, all you need to do to send it to Gerrit for code review is run:

git review

When that completes, automated tests will run on your change and other developers will peer review it.

Updating a Change

If the code review process suggests additional changes, make and amend the changes to the existing commit. Leave the existing Change-Id: footer in the commit message as-is. Gerrit knows that this is an updated patchset for an existing change:

git commit -a --amend
git review

Understanding Changes and Patch Sets

It’s important to understand how Gerrit handles changes and patch sets. Gerrit combines the Change-Id in the commit message, the project, and the target branch to uniquely identify a change.

A new patch set is determined by any modification in the commit hash. When a change is initially pushed up it only has one patch set. When an update is done for that change, git commit --amend will change the most current commit’s hash because it is essentially a new commit with the changes from the previous state combined with the new changes added. Since it has a new commit hash, once a git review is successfully processed, a new patch set appears in Gerrit.

Since a patch set is determined by a modification in the commit hash, many git commands will cause new patch sets. Three common ones that do this are:

  • git commit --amend

  • git rebase

  • git cherry-pick

As long as you leave the “Change-Id” line in the commit message alone and continue to propose the change to the same target branch, Gerrit will continue to associate the new commit with the already existing change, so that reviewers are able to see how the change evolves in response to comments.

Squashing Changes

If you have made many small commits, you should squash them so that they do not show up in the public repository. Remember: each commit becomes a change in Gerrit, and must be approved separately. If you are making one “change” to the project, squash your many “checkpoint” commits into one commit for public consumption. Here’s how:

git checkout master
git pull origin master
git checkout TOPIC-BRANCH
git rebase -i master

Use the editor to squash any commits that should not appear in the public history. If you want one change to be submitted to Gerrit, you should only have one “pick” line at the end of this process. After completing this, you can prepare your public commit message(s) in your editor. You start with the commit message from the commit that you picked, and it should have a Change-Id line in the message. Be sure to leave that Change-Id line in place when editing.

Once the commit history in your branch looks correct, run git review to submit your changes to Gerrit.

Adding a Dependency

When you want to start new work that is based on the commit under the review, you can add the commit as a dependency.

Fetch change under review and check out branch based on that change:

git review -d $PARENT_CHANGE_NUMBER
git checkout -b $DEV_TOPIC_BRANCH

Edit files, add files to git:

git commit -a
git review

Note

git review rebases the existing change (the dependency) and the new commit if there is a conflict against the branch they are being proposed to. Typically this is desired behavior as merging cannot happen until these conflicts are resolved. If you don’t want to deal with new patchsets in the existing change immediately you can pass the -R option to git review in the last step above to prevent rebasing. This requires future rebasing to resolve conflicts.

If the commit your work depends on is updated, and you need to get the latest patchset from the depended commit, you can do the following.

Fetch and checkout the parent change:

git review -d $PARENT_CHANGE_NUMBER

Cherry-pick your commit on top of it:

git review -x $CHILD_CHANGE_NUMBER

Submit rebased change for review:

git review

The note for the previous example applies here as well. Typically you want the rebase behavior in git review. If you would rather postpone resolving merge conflicts you can use git review -R as the last step above.

Rebasing a commit

Sometimes the target branch you are working on has changed, which can create a merge conflict with your patch. In this case, you need to rebase your commit on top of the current state of the branch. This rebase needs to be done manually:

  1. Checkout and update master:

    $ git checkout master
    $ git remote update
    
  2. Checkout the working branch and rebase on master:

    $ git review -d 180503
    $ git rebase origin/master
    
  3. If git indicates there are merge conflicts, view the affected files:

    $ git status
    
  4. Edit the listed files to fix conflicts, then add the modified files:

    $ git add <file1> <file2> <file3>
    
  5. Confirm that all conflicts are resolved, then continue the rebase:

    $ git status
    $ git rebase --continue
    

Cross-Repository Dependencies

If your change has a dependency on a change outside of that repository, like a change for another repository or some manual setup, you have to ensure that the change merge at the right time.

For a change depending on a manual setup, mark your change with the “Work in Progress” label until the manual setup is done. A core reviewer might also block an important change with a -2 so that it does not get merged accidentally before the manual setup is done.

If your change has a dependency on a change in another repository, you can use cross-repo dependencies (CRD) in Zuul:

  • To use them, include “Depends-On: <gerrit-change-url>” in the footer of your commit message. Use the permalink of the change. This is output by Gerrit when running git-review on the change, or you can find it in the top-left corner of the Gerrit web interface. Where it says “Change ###### - Needs …” the number is the link to the change; you can copy and paste that URL. A patch can also depend on multiple changes as explained in Multiple Changes.

  • These are one-way dependencies only – do not create a cycle.

Gate Pipeline

When Zuul sees CRD changes, it serializes them in the usual manner when enqueuing them into a pipeline. This means that if change A depends on B, then when they are added to the gate pipeline, B will appear first and A will follow. If tests for B fail, both B and A will be removed from the pipeline, and it will not be possible for A to merge until B does.

Note that if changes with CRD do not share a change queue (such as the “integrated gate”), then Zuul is unable to enqueue them together, and the first will be required to merge before the second is enqueued.

Check Pipeline

When changes are enqueued into the check pipeline, all of the related dependencies (both normal git-dependencies that come from parent commits as well as CRD changes) appear in a dependency graph, as in the gate pipeline. This means that even in the check pipeline, your change will be tested with its dependency. So changes that were previously unable to be fully tested until a related change landed in a different repo may now be tested together from the start.

All of the changes are still independent (so you will note that the whole pipeline does not share a graph as in the gate pipeline), but for each change tested, all of its dependencies are visually connected to it, and they are used to construct the git references that Zuul uses when testing. When looking at this graph on the Zuul status page, you will note that the dependencies show up as grey dots, while the actual change tested shows up as red or green. This is to indicate that the grey changes are only there to establish dependencies. Even if one of the dependencies is also being tested, it will show up as a grey dot when used as a dependency, but separately and additionally will appear as its own red or green dot for its test.

Multiple Changes

A Gerrit URL refers to a single change on a single branch, so if your change depends on multiple changes, or the same change on multiple branches of a project, you will need to explicitly list each URL. Simply add another “Depends-On:” line to the footer for each additional change.

Cycles

If a cycle is created by use of CRD, Zuul will abort its work very early. There will be no message in Gerrit and no changes that are part of the cycle will be enqueued into any pipeline. This is to protect Zuul from infinite loops. The developers hope that they can improve this to at least leave a message in Gerrit in the future. But in the meantime, please be cognizant of this and do not create dependency cycles with Depends-On lines.

Limitations and Caveats

Keep in mind that these dependencies are dependencies on changes in other repositories. Thus, a Depends-on only enforces an ordering but is not visible otherwise especially in these cases:

  • Changes for the CI infrastructure like changes openstack/project-config are never tested in a production simulated environment. So, if one of the changes adjusts the job definitions or creates a new job, a Depends-On will not test the new definition, the CI infrastructure change needs to merge to master and be in production to be fully evaluated.

  • If a test job installs packages from PyPI and not via source, be aware that the package from PyPI will always be used, a Depends-On will not cause a modified package to be used instead of installing from PyPI.

    As an example, if you are testing a change in python-novaclient that needs a change in python-keystoneclient, you add a Depends-On in the python-novaclient change. If a python-novaclient job installs python-keystoneclient from PyPI, the Depends-On will not have any effect since the PyPI version is used. If a python-novaclient job installs python-keystoneclient from source, the checked out source will have the change applied.

Do not add a Depends-On an abandoned change, your change will never merge.

If you backport a change to another branch, it will recieve a new URL. If you need to additionally depend on the backported change, you will need to amend the commit message to add another Depends-On footer.

A change that is dependent on another can be approved before the dependent change merges. If the repositories share the gate queue, it will merge automatically after the dependent change merged. But if the repositories do not share the gate queue, it will not merge automatically when the dependent change has merged, even a recheck will not help. Zuul waits for a status change and does not see it. The change needs another approval or a toggle of the approval, toggle means removing the approval and readding it again.

Code Review

Log in to https://review.opendev.org/ to see proposed changes, and review them.

To provide a review for a proposed change in the Gerrit UI, click on the Reply... button. In the code review, you can add a message, as well as a vote (+1,0,-1).

It’s also possible to add comments to specific lines in the file, for giving context to the comment. For that look at the diff of changes done in the file (click the file name), and click on the line number for which you want to add the inline comment. After you add one or more inline comments, you still have to send the Review message (see above, with or without text and vote). Prior to sending the inline comments in a review comment the inline comments are stored as Drafts in your browser. Other reviewers can only see them after you have submitted them as a comment on the patchset.

Any developer may propose or comment on a change (including voting +1/0/-1 on it). A vote of +2 is allowed from core reviewers, and should be used to indicate that they are a core reviewer and are leaving a vote that should be counted as such.

Some OpenDev hosted projects, like many OpenStack project teams, have a policy requiring two positive reviews from core reviewers.

When a review has enough +2 reviews and one of the core team believes it is ready to be merged, he or she should leave a +1 vote in the “Workflow” category. You may do so by clicking the “Review” button again, with or without changing your code review vote and optionally leaving a comment. When a +1 Approved review is received, Zuul will run tests on the change, and if they pass, it will be merged.

A green checkmark indicates that the review has met the requirement for that category. Under “Code-Review”, only one +2 gets the green check.

For more details on reviews in Gerrit, check the Gerrit documentation.

Automated Testing

When a new patchset is uploaded to Gerrit, that project’s “check” tests are run on the patchset by Zuul. Once completed the test results are reported to Gerrit by Zuul in the form of a Verified: +/-1 vote. After code reviews have been completed and a change receives an Approved: +1 vote that project’s “gate” tests are run on the change by Zuul. Zuul reports the results of these tests back to Gerrit in the form of a Verified: +/-2 vote. Code merging will only occur after the gate tests have passed successfully and received a Verified: +2. You can view the state of tests currently being run on the Zuul Status page.

If a change fails tests in Zuul, please follow the steps below:

  1. Zuul leaves a comment in the review with links to the log files for the test run. Follow those links and examine the output from the test. It will include a console log, and in the case of unit tests, HTML output from the test runner, or in the case of a devstack-gate test, it may contain quite a large number of system logs.

  2. Examine the console log or other relevant log files to determine the cause of the error. If it is related to your change, you should fix the problem and upload a new patchset. Do not use “recheck”.

  3. It is possible that the CI infrastructure may be having some issues which are causing your tests to fail. You can verify the status of the OpenDev infrastructure by doing the following:

    Note

    If a job fails in the automated testing system with the status of POST_FAILURE rather than a normal FAILURE, it could either be that your tests resulted with the system under test losing network connectivity or an issue with the automated testing system.

    If you are seeing repeated POST_FAILURE reports with no indication of problems in the CI system, make sure that your tests are not impacting the network of the system.

  4. It may be the case that the problem is due to non-deterministic behavior unrelated to your change that has already merged. In this situation, you can help other developers and focus the attention of QA, CI, and developers working on a fix by performing the following steps:

    1. For OpenStack projects, check elastic-recheck to see whether the bug is already identified and if not, add it.

      If your error is not there, then:

    2. Identify which project or projects are affected, and search for a related bug on Launchpad. You can search for bugs affecting all OpenStack Projects here: https://bugs.launchpad.net/openstack/ If you do not find an existing bug, file a new one (be sure to include the error message and a link to the logs for the failure). If the problem is due to an infrastructure problem (such as Zuul or Gerrit), file (or search for) the bug against the openstack-gate project.

  5. It may also happen that the CI infrastructure somehow cannot finish a job and restarts it. If this happens several times, the job is marked as failed with a message of RETRY_LIMIT. Usually this means that network connectivity for the job was lost and the change itself causes the job node to become unreachable consistently.

  6. To re-run check or gate jobs, leave a comment on the review with the form “recheck”.

A patchset has to be approved to run tests in the gate pipeline. If the patchset has failed in the gate pipeline (it will have been approved to get into the gate pipeline) a recheck will first run the check jobs and if those pass, it will again run the gate jobs. There is no way to only run the gate jobs, the check jobs will first be run again.

More information on debugging automated testing failures can be found in the following recordings:

Post Processing

After patches land, jobs can be run in the post queue. Finding build logs for these jobs works a bit differently to the results of the pre-merge check and gate queues.

For jobs in the post queue, logs are found via the builds tab of https://zuul.opendev.org/, for example to search for post jobs of the openstack tenant, go to the Zuul openstack build tab .

Peer Review

Anyone can be a reviewer: participating in the review process is a great way to learn about social norms and the development processes.

General review flow:

  1. Review is a conversation that works best when it flows back and forth. Submitters need to be responsive to questions asked in comments, even if the score is +0 from the reviewer. Likewise, reviewers should not use a negative score to elicit a response if they are not sure the patch should be changed before merging.

    For example, if there is a patch submitted which a reviewer cannot fully understand because there are changes that aren’t documented in the commit message or code documentation, this is a good time to issue a negative score. Patches need to be clear in their commit message and documentation.

    As a counter-example, a patch which is making use of a new library, which the reviewer has never used before, should not elicit a negative score from the reviewer with a question like “Is this library using standard python sockets for communication?” That is a question the reviewer can answer themselves, and which should not hold up the review process while the submitter explains things. Either the author or a reviewer should try to add a review comment answering such questions, unless they indicate a need to better extend the commit message, code comments, docstrings or accompanying documentation files.

  2. In almost all cases, a negative review should be accompanied by clear instructions for the submitter how they might fix the patch.

There may be more specific items to be aware of inside the projects’ documentation for contributors.

Contributors may notice a review that has several +1’s from other reviewers, passes the functional tests, etc. but the code still has not been merged. As only core contributors can approve code for merging, you can help things along by getting a core reviewer’s attention in IRC (never on the mailing lists) and letting them know there is a changeset with lots of positive reviews and needs final approval.

Work in Progress

To get early feedback on a change which is not fully finished yet, you can submit a change to Gerrit and mark it as “Work in Progress” (WIP).

Note

The OpenDev Gerrit system does not support drafts, use “Work in Progress” instead. Draft changes have been disabled because people assume they are private when they are not. They also create confusion if child changes are not drafts. Additionally it is difficult to run CI on them. It is better to assume changes are public and mark the not yet ready state.

To do so, after submitting a change to Gerrit in usual way (git review), You should go to Gerrit, and do Code Review of your own change while setting “Workflow” vote to “-1”, which marks the change as WIP.

This allows others to review the change, while at the same time blocking it from being merged, as you already plan to continue working on it.

Note

After uploading a new patchset, this -1 (WIP) vote disappears. So if you still plan to do additional changes, do not forget to set Workflow to -1 on the new patchset.

Merging

Once a change has been approved and passed the gate jobs, Gerrit automatically merges the latest patchset.

Each patchset gets merged to the head of the branch before testing it. If Gerrit cannot merge a patchset, it will give a -1 review and add a comment notifying of merge failure.

Each time a change merges, the “merge-check” pipeline verifies that all open changes on the same project are still mergeable. If any job is not mergeable, Zuul will give a -1 review and add a comment notifying of merge failure.

After a change is merged, project-specific post jobs are run. Most often the post jobs publish documentation, run coverage, or send strings to the translation server.

Project Gating

Project gating refers to the process of running regression tests before a developer’s patchset is merged. The intent of running regression tests is to validate that new changes submitted against the source code repository will not introduce new bugs. Gating prevents regressions by ensuring that a series of tests pass successfully before allowing a patchset to be merged into the mainline of development.

The system used for gating is Zuul, which listens to the Gerrit event stream and is configured with YAML files to define a series of tests to be run in response to an event.

The jobs in the gate queue are executed once a core reviewer approves a change (using a +1 Workflow vote) and a verified +1 vote exist. When approving, at least one +2 Code-Review vote needs to exist (can be given by core reviewer when approving). The convention is that two +2 Code-Reviews are needed for approving.

Once all of the jobs report success on an approved patchset in the configured gate pipeline, then Gerrit will merge the code into trunk.

Besides running the gate tests, the gate pipeline determines the order of changes to merge across multiple projects. The changes are tested and merged in this order, so that for each change the state of all other repositories can be identified.

Additional information about project gating and Zuul can be found in the Zuul documentation, located at: https://zuul-ci.org/docs/zuul/user/gating.html