User Guide

The @PLUGIN@ plugin allows to define code owners for directories and files and requires their approval for changes that touch these files.

This user guide explains the functionality of the @PLUGIN@ plugin. For a walkthrough of the UI please refer to the intro page.

TIP: You may also want to check out the presentation about code owners from the Gerrit Contributor Summit 2020.

NOTE: How to setup the code owners functionality is explained in the setup guide.

What are code owners?

A code owner is a user that is configured as owner of a path (directory or file) and whose approval is required to modify the path or files under that path.

Who is a code owner of a path is controlled via code owner config files.

Code owner config files

Code owner config files are stored in the source tree of the repository and define the code owners for a path.

In which files code owners are defined and which syntax is used depends on the configured code owner backend. Example: if the find-owners backend is used, code owners are defined in OWNERS files.

To create/edit code owner config files, clone the repository, edit the code owner config files locally and then push the new commit to the remote repository in Gerrit. This the same as creating/editing any other source files.

On push, Gerrit validates any code owner config file that is touched by the new commits, unless the validation for received commits is disabled. If the validation is enabled, commits that make code owner config files invalid are rejected.

NOTE: There is no dedicated editor for code owner config files in the Gerrit UI.

NOTE: It is the responsibility of the project owners to maintain the code owner config files (e.g. take care to remove code owners that leave the team).

Code owner approval

For a change to be submittable Gerrit requires that all files that are touched in the change are approved by a code owner.

Code owners apply their approval by voting on the change. By default, voting with Code-Review+1 counts as code owner approval, but it's possible that the host administrators or the project owners have configured a different label/vote that is required as code owner approval.

The code owner check for a file is satisfied as soon as one of its code owners grants the code owner approval. Negative votes from other code owners do not block the submission (unless it's a veto vote which is configured independently of the @PLUGIN@ plugin).

It's possible to configure that for changes/patch-sets that are uploaded by a code owner an implicit code owner approval from the uploader is assumed. In this case, if a code owner only touches files that they own, no approval from other code owners is required. If this is configured, it is important that code owners are aware of their implicit approval when they upload new patch sets for other users (e.g. if a contributor pushes a change to a wrong branch and a code owner helps them to get it rebased onto the correct branch, the rebased change has implicit approvals from the code owner, since the code owner is the uploader).

For files that are renamed/moved Gerrit requires a code owner approval for the old and the new path of the files.

If code owner approvals are missing, it is possible to submit the change with a code owner override, but this should only be done in exceptional cases.

NOTE: It is possible that users are code owners, but miss permissions to vote on the required label. This is a configuration issue that should be reported to the project owners (who should either grant the permission or remove the code owner).

NOTE: It's possible that the change submission is still blocked after all necessary code owner approvals have been granted. This is the case, if futher non-code-owner approvals are required and missing, or if further non-code-owner submit requirements are not fulfilled yet.

NOTE: Whether code owner approvals are sticky across patch sets depends on the definition of the required label. If the label definition has copy rules enabled so that votes are sticky across patch sets, then also the code owner approvals which are based on these votes will be sticky.

NOTE: Whether code owners can approve their own changes depends of the definition of the required label. If the label definition has ignoreSelfApproval enabled, code owner approvals of the patch set uploader are ignored.

NOTE: Code owner approvals are always applied on the whole change / patch set and count for all files in the change / patch set. It is not possible to approve individual files only. This means code owners should always review all files in the change / patch set before applying their approval. E.g. it is discouraged to only review the owned files, since the set of owned files can change if OWNERS files in the destination branch are changed after the approval has been applied.

Code owner override

Usually some privileged users, such as sheriffs, are allowed to override the code owner approval check to make changes submittable, even if code owner approvals for some or all files are missing.

Overrides are intended to be used in emergency cases only. What qualifies as an emergency depends on the project. Usually overrides are used when changes need to be submitted quickly to address an urgent production issue and there is no time to await all required code owner approvals.

A code owner override is applied by voting on the change. Which label/vote counts as code owner override depends on the configuration.

NOTE: It's possible that overrides are disabled for a project.

Code owner Suggestion

As a change owner, you need to request code owner approvals for the files that are modified in the change, so that your change can become submittable. This is done by selecting a code owner for each of the files and adding them as reviewer to the change. To help you with this task, Gerrit suggests you suitable code owners for the files in the change and lets you pick which of them should be added as reviewers (how this looks in the UI is shown in the intro page).

When suggesting code owners for a file, Gerrit filters out code owners that:

  • are inactive
  • are not visible to you (according to accounts.visibility setting),
  • are referenced by non-visible secondary emails
  • are not resolvable (emails for which no Gerrit account exists)
  • are ambiguous (the same email is assigned to multiple accounts)
  • are referenced by an email with a disallowed domain (see allowedEmailDomain configuration)
  • do not have read access to the destination branch of the change
  • are service users (members of the Service Users group)

The suggested code owners are sorted by score, so that the best suitable code owners appear first. The following criteria are taken into account for computing the score:

  • The distance of the code owner config file that defines the code owner from the owned path.
    The smaller the distance the better we consider the code owner as reviewer/approver for the path.

How to submit changes with files that have no code owners?

If the code owners functionality is enabled, all touched files require an approval from a code owner. If files are touched for which no code owners are defined, the change can only be submitted with an approval of a fallback code owner (if configured) or with a code owner override. Please note that fallback code owners are not included in the code owner suggestion.

Renames

A rename is treated as a deletion at the old path and a creation at the new path. This is why for files that are renamed, Gerrit requires a code owner approval for the old and the new path of the files (also see code owner approval section).

When files/folders get renamed, their code owner configuration should stay intact. Renaming a file/folder should normally not result in a situation where the code owner configuration for this file/folder no longer applies, because it was renamed.

Mostly this is not a problem because code owner config files are stored inside the folders to which they apply. This means, if a folder gets renamed, the code owner config files in it still apply.

However if a file/folder is renamed for which specific code owners are defined via path expressions, it is possible that the code ownership changes. For example, this can happen if the old name is matched by a path expression that makes user A a code owner, but the new name is only matched by another path expression that makes user B a code owner. E.g. ‘*.md’ is owned by user A, ‘*.txt’ is owned by user B and ‘config.md’ is renamed to ‘config.txt’. In this case it is the responsibility of the author doing the rename and the current code owners to ensure that the file/folder has the proper code owners at the new path. This is also the reason why matching subfolders via path expressions is discouraged.

Merge commits

By default, changes for merge commits require code owner approvals for all files that differ between the merge commit that is being reviewed and the tip of the destination branch (the first parent commit). This includes all files that have been touched in other branches and that are now being integrated into the destination branch (regardless of whether there was a conflict resolution or whether the auto-merge succeeded without conflicts). To see these files in the change screen, Parent 1 needs to be selected as base for the comparison (instead of the Auto Merge that is selected as base by default).

By configuration it is possible, that changes for merge commits only require code owner approvals for files that differ between the merge commit that is being reviewed and the Auto Merge. In this case, code owners only need to approve the files for which a conflict resolution was done. These are the files that are shown in the change screen when Auto Merge is selected as base. Using this configuration makes sense if all branches require code owner approvals and the code owners of all branches are trusted, as it prevents that code owners need to approve the same changes multiple times, but for different branches.

Code Owners Submit Rule

The logic that checks whether a change has sufficient code owner approvals to be submitted is implemented in the code owners submit rule. If the code owners submit rule finds that code owner approvals are missing the submission of the change is blocked. In this case it's possible to use a code owner override to unblock the change submission.

NOTE: Besides the code owners submit rule there may be further submit rules that block the change submission for other reasons that are not related to code owners. E.g. configured label functions are completely orthogonal to code owner approvals. If, for example, Code-Review+1 votes are required as code owner approval, but the Code-Review label has the function MaxWithBlock the change submission is still blocked if a max approval (aka Code-Review+2) is missing or if a veto vote (aka Code-Review-2) is present.

NOTE: Gerrit submit rules are executed on submit and when change details are loaded, e.g. when loading the change screen (to know whether the submit button should be enabled). In addition, submit rules are executed on every change update because the result of running submit rules is stored as submit records in the change index. This makes the submit records available when querying changes (without needing to run the submit rules for every change in the result which would be too expensive). For code owners the submit records that are stored in the index can become stale for 2 reasons: 1. code owner config files are changed after the change has been indexed (e.g. new code owners are added), 2. if the code owners plugin configuration was changed in a way that affected the result of the code owners submit rule. Callers of change queries should be aware of this.


Back to @PLUGIN@ documentation index

Part of Gerrit Code Review