This is a Gerrit guide that is dedicated to project owners. It explains the many possibilities that Gerrit provides to customize the workflows for a project.

What is a project owner?

Being project owner means that you own a project in Gerrit. Technically this is expressed by having the Owner access right on refs/* on that project. As project owner you have the permission to edit the access control list and the project settings of the project. It also means that you should get familiar with these settings so that you can adapt them to the needs of your project.

Being project owner means being responsible for the administration of a project. This requires having a deeper knowledge of Gerrit than the average user. Normally per team there should be 2 to 3 persons, who have a certain level of Git/Gerrit knowledge, assigned as project owners. It normally doesn’t make sense that everyone in a team is project owner. For normal team members it is sufficient to be committer or contributor.

Access Rights

As a project owner you can edit the access control list of your project. This allows you to grant permissions on the project to different groups.

Gerrit comes with a rich set of permissions which allow a very fine-grained control over who can do what on a project. Access control is one of the most powerful Gerrit features but it is also a rather complex topic. This guide will only highlight the most important aspects of access control, but the Access Control chapter explains all the details.

Editing Access Rights

To see the access rights of your project

  • go to the Gerrit Web UI

  • click on the BROWSE > Repositories menu entry

  • find your project in the project list and click on it

  • click on the Access menu entry

By clicking on the Edit button the access rights become editable and you may save any changes by clicking on the Save Changes button. Optionally you can provide a Commit Message to explain the reasons for changing the access rights.

The access rights are stored in the project’s Git repository in a special branch called refs/meta/config. On this branch there is a project.config file which contains the access rights. More information about this storage format can be found in the Project Configuration File Format chapter. What is important to know is that by looking at the history of the project.config file on the refs/meta/config branch you can always see how the access rights were changed and by whom. If a good commit message is provided you can also see from the history why the access rights were modified.

If a Git browser such as gitweb is configured for the Gerrit server you can find a link to the history of the project.config file in the Web UI. Otherwise you may inspect the history locally. If you have cloned the repository you can do this by executing the following commands:

  $ git fetch ssh://localhost:29418/project refs/meta/config
  $ git checkout FETCH_HEAD
  $ git log project.config

Non project owners may still edit the access rights and propose the modifications to the project owners by clicking on the Save for Review button. This creates a new change with the access rights modifications that can be approved by a project owner. The project owners are automatically added as reviewer on this change so that they get informed about it by email.

Inheritance

Normally when a new project is created in Gerrit it already has some access rights which are inherited from the parent projects. Projects in Gerrit are organized hierarchically as a tree with the All-Projects project as root from which all projects inherit. Each project can have only a single parent project, multi-inheritance is not supported.

Looking at the access rights of your project in the Gerrit Web UI, you only see the access rights which are defined on that project. To see the inherited access rights you must follow the link to the parent project under Rights Inherit From.

Inherited access rights can be overwritten unless they are defined as BLOCK rule. BLOCK rules are used to limit the possibilities of the project owners on the inheriting projects. With this, global policies can be enforced on all projects. Please note that Gerrit doesn’t prevent you from assigning access rights that contradict an inherited BLOCK rule, but these access rights will simply have no effect.

If you are responsible for several projects which require the same permissions, it makes sense to have a common parent for them and to maintain the access rights on that common parent. Changing the parent of a project is only allowed for Gerrit administrators. This means you need to contact the administrator of your Gerrit server if you want to reparent your project. One way to do this is to change the parent project in the Web UI, save the modifications for review and get the change approved and merged by a Gerrit administrator.

References

Access rights in Gerrit are assigned on references (aka refs). Refs in Git exist in different namespaces, e.g. all branches normally exist under refs/heads/ and all tags under refs/tags/. In addition there are a number of special refs and magic refs.

Gerrit only supports tags that are reachable by any ref not owned by Gerrit. This includes branches (refs/heads/*) or custom ref namespaces (refs/my-company/*). Tagging a change ref is not supported. When filtering tags by visibility, Gerrit performs a reachability check and will present the user ony with tags that are reachable by any ref they can see.

Access rights can be assigned on a concrete ref, e.g. refs/heads/master but also on ref patterns and regular expressions for ref names.

A ref pattern ends with /* and describes a complete ref name namespace, e.g. access rights assigned on refs/heads/* apply to all branches.

Regular expressions must start with ^, e.g. access rights assigned on ^refs/heads/rel-.* would apply to all rel-* branches.

Groups

Access rights are granted to groups. It is useful to know that Gerrit maintains its own groups internally but also supports different external group backends.

The Gerrit internal groups can be seen in the Gerrit Web UI by clicking on the BROWSE > Groups menu entry. By clicking on a group you can edit the group members (Members tab) and the group options (General tab).

Gerrit internal groups contain users as members, but can also include other groups, even external groups.

Every group is owned by an owner group. Only members of the owner group can administrate the owned group (assign members, edit the group options). A group can own itself; in this case members of the group can, for example, add further members to the group. When you create new groups for your project to assign access rights to committer or other roles, make sure that they are owned by the project owner group.

An important setting on a group is the option Make group visible to all registered users., which defines whether non-members can see who is member of the group.

New internal Gerrit groups can be created under BROWSE > Groups and then clicking on the CREATE NEW button in the upper right corner. The CREATE NEW button is only available if you have the global capability Create Group assigned.

Gerrit also has a set of special system groups that you might find useful.

External groups need to be prefixed when assigning access rights to them, e.g. LDAP group names need to be prefixed with ldap/.

If the singleusergroup plugin is installed you can also directly assign access rights to users, by prefixing the username with user/ or the user’s account ID by userid/.

Common Access Rights

Different roles in a project, such as developer (committer) or contributor, need different access rights. Examples for which access rights are typically assigned for which role are described in the Access Control chapter.

Code Review

Gerrit’s main functionality is code review, however using code review is optional and you may decide to only use Gerrit as a Git server with access control. Whether you allow only pushes for review or also direct pushes depends on the project’s access rights.

To push a commit for review it must be pushed to refs/for/<branch-name>. This means the Push access right must be assigned on refs/for/<branch-name>.

To allow direct pushes and bypass code review, the Push access right is required on refs/heads/<branch-name>.

By pushing for review you are not only enabling the review workflow, but you can also get automatic verifications from a build server before changes are merged. In addition you can benefit from Gerrit’s merge strategies that can automatically merge/rebase commits on server side if necessary. You can control the merge strategy by configuring the submit type on the project. If you bypass code review you always need to merge/rebase manually if the tip of the destination branch has moved. Please keep this in mind if you choose to not work with code review because you think it’s easier to avoid the additional complexity of the review workflow; it might actually not be easier.

You may also enable auto-merge on push to benefit from the automatic merge/rebase on server side while pushing directly into the repository.

Project Options

As project owner you can control several options on your project. The different options are described in the Project Options section.

To see the options of your project:

  1. Go to the Gerrit Web UI.

  2. Click on the BROWSE > Repositories menu entry.

  3. Find your project in the project list and click it.

  4. Click the General menu entry.

Submit Type

An important decision for a project is the choice of the submit type and the content merge setting (see the Allow content merges option). The submit type is the method Gerrit uses to submit a change to the project. The submit type defines what Gerrit should do on submit of a change if the destination branch has moved while the change was in review. The content merge setting applies if the same files have been modified concurrently and tells Gerrit whether it should attempt a content merge for these files.

When choosing the submit type and the content merge setting one must weigh development comfort against the safety of not breaking the destination branch.

The most restrictive submit type is Fast Forward Only. Using this submit type means that after submitting one change all other open changes for the same destination branch must be rebased manually. This is quite burdensome and in practice only feasible for branches with very few changes. On the other hand, if changes are verified before submit, e.g. automatically by a CI integration, with this submit type, you can be sure that the destination branch never gets broken.

Choosing Merge If Necessary as submit type makes the life for developers more comfortable, especially if content merge is enabled. If this submit strategy is used developers only need to rebase manually if the same files have been modified concurrently or if the content merge on such a file fails. The drawback with this submit type is that there is a risk of breaking the destination branch, e.g. if one change moves a class into another package and another change imports this class from the old location. Experience shows that in practice Merge If Necessary with content merge enabled works pretty well and breaking the destination branch happens rarely. This is why this setting is recommended at least for development branches. You likely want to start with Merge If Necessary with content merge enabled and only switch to a more restrictive policy if you are facing issues with the build and test stability of the destination branches.

It is also possible to define the submit type dynamically via Prolog. This way you can use different submit types for different branches.

Please note that there are other submit types available; they are described in the Submit Type section.

Labels

The code review process includes that reviewers formally express their opinion about a change by voting on different labels. By default Gerrit comes with the Code-Review label and many Gerrit servers have the Verified label configured globally. However projects can also define their own custom labels to formalize project-specific workflows. For example if a project requires an IP approval from a special IP-team, it can define an IP-Review label and grant permissions to the IP-team to vote on this label.

The behavior of a label can be controlled by its function, e.g. it can be configured whether a max positive voting on the label is required for submit or if the voting on the label is optional.

By using a custom submit rule it can be controlled per change whether a label is required for submit or not.

A useful feature on labels is the possibility to automatically copy scores forward to new patch sets if it was a trivial rebase or if there was no code change (e.g. only the commit message was edited).

Submit Rules

A submit rule in Gerrit is logic that defines when a change is submittable. By default, a change is submittable when it gets at least one highest vote on each label and has no lowest vote (aka veto vote) on any label.

The submit rules in Gerrit are implemented in Prolog and projects that need more flexibility can define their own submit rules to decide when a change should be submittable. A good example from the Prolog cookbook shows how to allow submit only if a change has a Code-Review+2 vote from a person that is not the change author. This way a four-eyes principle for the reviews can be enforced.

A Prolog submit rule has access to information about the change for which it is testing the submittability. Among others the list of the modified files can be accessed, which allows special logic if certain files are touched. For example, a common practice is to require a vote on an additional label, like Library-Compliance, if the dependencies of the project are changed.

It is also possible to control the submit type from Prolog. For example this can be used to define a more restrictive submit type such as Fast Forward Only for stable branches while using a more liberal submit type, e.g. Merge If Necessary with content merge, for development branches. How this can be done can be seen from an example in the Prolog cookbook.

Submit rules are maintained in the rules.pl file in the refs/meta/config branch of the project. How to write submit rules is explained in the Prolog cookbook. There is also good support for testing submit rules while developing them.

Continuous Integration

With Gerrit you can have continuous integration builds not only for updates of central branches but also whenever a new change/patch set is uploaded for review. This way you get automatic verification of all changes before they are merged and any build and test issues are detected early. To indicate the build and test status the continuous integration system normally votes with the Verified label on the change.

There are several solutions for integrating continuous integration systems. The most commonly used are:

For the integration with the continuous integration system you must have a service user that is able to access Gerrit. To create a service user in Gerrit you can use the create-account SSH command if the Create Account global capability is granted. If not, you need to ask a Gerrit administrator to create the service user.

If the serviceuser plugin is installed you can also create new service users in the Gerrit Web UI under People > Create Service User. For this the Create Service User global capability must be assigned.

The service user must have read access to your project. In addition, if automatic change verification is enabled, the service user must be allowed to vote on the Verified label.

Continuous integration systems usually integrate with Gerrit by listening to the Gerrit stream events. For this the service user must have the Stream Events global capability assigned.

Commit Validation

Gerrit provides an extension point to do validation of new commits. A Gerrit plugin implementing this extension point can perform validation checks when new commits are pushed to Gerrit. The plugin can either provide a message to the client or reject the commit and cause the push to fail.

There are some plugins available that provide commit validation:

  • uploadvalidator:

    The uploadvalidator plugin allows project owners to configure blocked file extensions, required footers and a maximum allowed path length.

  • commit-message-length-validator

    The commit-message-length-validator core plugin validates that commit messages conform to line length limits.

Branch Administration

As project owner you can administrate the branches of your project in the Gerrit Web UI under BROWSE > Repositories > <your project> > Branches. In the Web UI branch creation is allowed if you have Create Reference access right and branch deletion is allowed if you have the Delete Reference or the Push access right with the force option.

By setting HEAD on the project you can define its default branch. For convenience reasons, when the repository is cloned Git creates a local branch for this default branch and checks it out.

Email Notifications

With Gerrit individual users control their own email subscriptions. By editing the watched projects in the Web UI under Settings > Notifications users can decide which events to be informed about by email. The change notifications can be filtered by change search expressions.

This means as a project owner you normally don’t need to do anything about email notifications, except maybe telling your project team where to configure the watches.

Gerrit also supports notifications on project level that allow project owners to set up email notifications for team mailing lists or groups of users. This configuration is done in the project.config file in the refs/meta/config branch as explained in the section about project level notifications.

Dashboards

Gerrit comes with a pre-defined user dashboard that provides a view of the changes that are relevant for a user. Users can also define their own custom dashboards where the dashboard sections can be freely configured. As a project owner you can configure such custom dashboards on project level. This way you can define a view of the changes that are relevant for your project and share this dashboard with all users. The project dashboards can be seen in the Web UI under BROWSE > Repositories > <your project> > Dashboards.

Issue Tracker Integration

There are several possibilities to integrate issue trackers with Gerrit.

  • Comment Links

    As described in the Comment Links section, comment links can be used to link IDs from commit message footers to issues in an issue tracker system.

  • Tracking IDs

    Gerrit can be configured to index IDs from commit message footers so that the tr/bug search operators can be used to query for changes with a certain ID. The configuration of tracking IDs can only be done globally by a Gerrit administrator.

  • Issue Tracker System Plugins

    There are Gerrit plugins for a tight integration with Jira, Bugzilla and IBM Rational Team Concert. If installed, these plugins can e.g. be used to automatically add links to Gerrit changes to the issues in the issue tracker system or to automatically close an issue if the corresponding change is merged. If installed, project owners may enable/disable the issue tracker integration from the Gerrit Web UI under BROWSE > Repositories > <your project> > General.

Gerrit can linkify strings in commit messages, summary comments and inline comments. A string that matches a defined regular expression is then displayed as hyperlink to a configured backend system.

So called comment links can be configured globally by a Gerrit administrator, but also per project by the project owner. Comment links on project level are defined in the project.config file in the refs/meta/config branch of the project as described in the documentation of the commentlink configuration parameter.

Often comment links are used to link an ID in a commit message footer to an issue in an issue tracker system. For example, to link the ID from the Bug footer to Jira the following configuration can be used:

  [commentlink "myjira"]
    match = ([Bb][Uu][Gg]:\\s+)(\\S+)
    link =  https://myjira/browse/$2

Reviewers

Normally it is not needed to explicitly assign reviewers to every change since the project members either watch the project and get notified by email or regularly check the list of open changes in the Gerrit Web UI. The project members then pick the changes themselves that are interesting to them for review.

If authors of changes want to have a review by a particular person (e.g. someone who is known to be expert in the touched code area, or a stakeholder for the implemented feature), they can request the review by adding this person in the Gerrit Web UI as a reviewer on the change. Gerrit will then notify this person by email about the review request.

With the reviewers plugin it is possible to configure default reviewers who will be automatically added to each change. The default reviewers can be configured in the Gerrit Web UI under BROWSE > Repositories > <your project> > General in the reviewers Plugin section.

The reviewers-by-blame plugin can automatically add reviewers to changes based on the git blame computation on the changed files. This means that the plugin will add those users as reviewer that authored most of the lines touched by the change, since these users should be familiar with the code and can most likely review the change. How many reviewers the plugin will add to a change at most can be configured in the Gerrit Web UI under BROWSE > Repositories > <your project> > General in the reviewers-by-blame Plugin section.

Download Commands

On the change screen in the Downloads drop-down panel Gerrit offers commands for downloading the currently viewed patch set.

The download commands are implemented by Gerrit plugins. This means that the available download commands depend on the installed Gerrit plugins:

  • download-commands plugin:

    The download-commands plugin provides the default download commands (Checkout, Cherry Pick, Format Patch and Pull).

    Gerrit administrators may configure which of the commands are shown on the change screen.

  • project-download-commands plugin:

    The project-download-commands plugin enables project owners to configure project-specific download commands. For example, a project-specific download command may update submodules, trigger a build, execute the tests or even do a deployment.

    The project-specific download commands must be configured in the project.config file in the refs/meta/config branch of the project:

      [plugin "project-download-commands"]
        Build = git fetch ${url} ${ref} && git checkout FETCH_HEAD && bazel build ${project}
        Update = git fetch ${url} ${ref} && git checkout FETCH_HEAD && git submodule update

    Project-specific download commands that are defined on a parent project are inherited by the child projects. A child project can overwrite an inherited download command, or remove it by assigning no value to it.

Integration with other tools

Gerrit provides many possibilities for the integration with other tools:

  • Stream Events:

    The stream-events SSH command allows to listen to Gerrit events. Other tools can use this to react on actions done in Gerrit.

    The Stream Events global capability is required for using the stream-events command.

  • REST API:

    Gerrit provides a rich REST API that other tools can use to query information from Gerrit and and to trigger actions in Gerrit.

  • Gerrit Plugins:

    The Gerrit functionality can be extended by plugins and there are many extension points, e.g. plugins can

    How to develop a Gerrit plugin is described in the Plugin Development section.

Project Lifecycle

Project Creation

New projects can be created in the Gerrit Web UI under BROWSE > Repositories and then clicking on the CREATE NEW button in the upper right corner. The CREATE NEW button is only available if you have the Create Project global capability assigned.

Projects can also be created via REST or SSH as described in the Project Setup section.

Creating the project with an initial empty commit is generally recommended because some tools have issues with cloning repositories that are completely empty. However, if you plan to import an existing history into the new project, it is better to create the project without an initial empty commit.

Import Existing History

If you have an existing history you can import it into a Gerrit project. To do this you need to have a local Git repository that contains this history. If your existing codebase is in another VCS you must migrate it to Git first. For Subversion you can use the git svn command as described in the Subversion migration guide. An importer for Perforce is available in the contrib section of the Git source code; how to use git p4 to do the import from Perforce is described in the Perforce migration guide.

To import an existing history into a Gerrit project you bypass code review and push it directly to refs/heads/<branch>. For this you must have the corresponding Push access right assigned. If the destination branch in the Gerrit repository already contains a history (e.g. an initial empty commit), you can overwrite it by doing a force push. In this case force push must be allowed in the access controls of the project.

Some Gerrit servers may disallow forging committers by blocking the Forge Committer access right globally. In this case you must use the git filter-branch command to rewrite the committer information for all commits (the author information that records who was writing the code stays intact; signed tags will lose their signature):

  $ git filter-branch --tag-name-filter cat --env-filter 'GIT_COMMITTER_NAME="John Doe"; GIT_COMMITTER_EMAIL="john.doe@example.com";' -- --all

If a max object size limit is configured on the server you may need to remove large objects from the history before you are able to push. To find large objects in the history of your project you can use the reposize.sh script which you can download from Gerrit:

  $ curl -Lo reposize.sh http://review.example.com:8080/tools/scripts/reposize.sh

or

  $ scp -p -P 29418 john.doe@review.example.com:scripts/reposize.sh .

You can then use the git filter-branch command to remove the large objects from the history of all branches:

  $ git filter-branch -f --index-filter 'git rm --cached --ignore-unmatch path/to/large-file.jar' -- --all

Since this command rewrites all commits in the repository it’s a good idea to create a fresh clone from this rewritten repository before pushing to Gerrit, this will ensure that the original objects which have been rewritten are removed.

Project Deletion

Gerrit core does not support the deletion of projects.

If the delete-project plugin is installed, projects can be deleted from the Gerrit Web UI under BROWSE > Repositories > <project> > General by clicking on the Delete command under Project Commands. The Delete command is only available if you have the Delete Projects global capability assigned, or if you own the project and you have the Delete Own Projects global capability assigned. If neither of these capabilities is granted, you need to contact a Gerrit administrator to request the deletion of your project.

Instead of deleting a project you may set the project state to ReadOnly or Hidden.

Project Rename

Gerrit core does not support the renaming of projects.

If the rename-project plugin is installed, projects can be renamed using the rename-project ssh command.

Find details about prerequisites in the plugin documentation.

If you don’t want to use the rename-project plugin you can perform the following steps as a workaround:

Please note that a drawback of this workaround is that the whole review history (changes, review comments) is lost.