Code Quality

Introduced in GitLab Starter 9.3.

Made available in all tiers in 13.2.

Ensuring your project's code stays simple, readable and easy to contribute to can be problematic. With the help of GitLab CI/CD, you can analyze your source code quality using GitLab Code Quality.

Code Quality:

Uses Code Climate Engines, which are free and open source. Code Quality does not require a Code Climate subscription.
Runs in pipelines using a Docker image built in the GitLab Code Quality project using default Code Climate configurations.
Can make use of a template.
Is available with Auto DevOps.
Can be extended through Analysis Plugins or a custom tool.

Code Quality Widget

Introduced in GitLab Starter 9.3.

Made available in all tiers in 13.2.

Going a step further, GitLab can show the Code Quality report right in the merge request widget area if a report from the target branch is available to compare to:

Watch a quick walkthrough of Code Quality in action:

See the video: Code Quality: Speed Run.

NOTE: For one customer, the auditor found that having Code Quality, SAST, and Container Scanning all automated in GitLab CI/CD was almost better than a manual review! Read more.

See also the Code Climate list of Supported Languages for Maintainability.

Use cases

For instance, consider the following workflow:

Your backend team member starts a new implementation for making a certain feature in your app faster.
With Code Quality reports, they analyze how their implementation is impacting the code quality.
The metrics show that their code degrades the quality by 10 points.
You ask a co-worker to help them with this modification.
They both work on the changes until Code Quality report displays no degradations, only improvements.
You approve the merge request and authorize its deployment to staging.
Once verified, their changes are deployed to production.

Example configuration

This example shows how to run Code Quality on your code by using GitLab CI/CD and Docker. It requires GitLab 11.11 or later, and GitLab Runner 11.5 or later. If you are using GitLab 11.4 or earlier, you can view the deprecated job definitions in the documentation archive.

Using shared runners, the job should be configured For the Docker-in-Docker workflow.
Using private runners, there is an alternative configuration recommended for running CodeQuality analysis more efficiently.

In either configuration, the runner mmust have enough disk space to handle generated Code Quality files. For example on the GitLab project the files are approximately 7 GB.

Once you set up GitLab Runner, include the Code Quality template in your CI configuration:

include:
  - template: Code-Quality.gitlab-ci.yml

The above example creates a code_quality job in your CI/CD pipeline which scans your source code for code quality issues. The report is saved as a Code Quality report artifact that you can later download and analyze.

It's also possible to override the URL to the Code Quality image by setting the CODE_QUALITY_IMAGE variable. This is particularly useful if you want to lock in a specific version of Code Quality, or use a fork of it:

include:
  - template: Code-Quality.gitlab-ci.yml

code_quality:
  variables:
    CODE_QUALITY_IMAGE: "registry.example.com/codequality-fork:latest"

In GitLab 13.4 and later, you can override the Code Quality environment variables:

variables:
  TIMEOUT_SECONDS: 1

include:
  - template: Code-Quality.gitlab-ci.yml

By default, report artifacts are not downloadable. If you need them downloadable on the job details page, you can add gl-code-quality-report.json to the artifact paths like so:

include:
  - template: Code-Quality.gitlab-ci.yml

code_quality:
  artifacts:
    paths: [gl-code-quality-report.json]

The included code_quality job is running in the test stage, so it needs to be included in your CI configuration, like so:

stages:
  - test

NOTE: This information is automatically extracted and shown right in the merge request widget.

WARNING: On self-managed instances, if a malicious actor compromises the Code Quality job definition they could execute privileged Docker commands on the runner host. Having proper access control policies mitigates this attack vector by allowing access only to trusted actors.

Set up a private runner for code quality without Docker-in-Docker

It's possible to configure your own runners and avoid Docker-in-Docker. You can use a configuration that may greatly speed up job execution without requiring your runners to operate in privileged mode.

This alternative configuration uses socket binding to share the Runner's Docker daemon with the job environment. Be aware that this configuration has significant considerations to be consider, but may be preferable depending on your use case.

$ gitlab-runner register --executor "docker" \
  --docker-image="docker:stable" \
  --url "https://gitlab.com/" \
  --description "cq-sans-dind" \
  --tag-list "cq-sans-dind" \
  --locked="false" \
  --access-level="not_protected" \
  --docker-volumes "/cache"\
  --docker-volumes "/var/run/docker.sock:/var/run/docker.sock" \
  --registration-token="<project_token>" \
  --non-interactive

Optional, but recommended: Set the builds directory to /tmp/builds, so job artifacts are periodically purged from the runner host. If you skip this step, you must clean up the default builds directory (/builds) yourself. You can do this by adding the following two flags to gitlab-runner register in the previous step.

--builds-dir /tmp/builds
--docker-volumes /tmp/builds:/tmp/builds

The resulting configuration:

[[runners]]
  name = "cq-sans-dind"
  url = "https://gitlab.com/"
  token = "<project_token>"
  executor = "docker"
  builds_dir = "/tmp/builds"
  [runners.docker]
    tls_verify = false
    image = "docker:stable"
    privileged = false
    disable_entrypoint_overwrite = false
    oom_kill_disable = false
    disable_cache = false
    volumes = ["/cache", "/var/run/docker.sock:/var/run/docker.sock", "/tmp/builds:/tmp/builds"]
    shm_size = 0
  [runners.cache]
    [runners.cache.s3]
    [runners.cache.gcs]

Apply two overrides to the code_quality job created by the template:

include:
  - template: Code-Quality.gitlab-ci.yml

code_quality:
  services:            # Shut off Docker-in-Docker
  tags:
    - cq-sans-dind     # Set this job to only run on our new specialized runner

The end result is that:

Privileged mode is not used.
Docker-in-Docker is not used.
Docker images, including all CodeClimate images, are cached, and not re-fetched for subsequent jobs.

With this configuration, the run time for a second pipeline is much shorter. For example this small change to an open merge request running Code Quality analysis ran significantly faster the second time:

This configuration is not possible on gitlab.com shared runners. Shared runners are configured with privileged=true, and they do not expose docker.sock into the job container. As a result, socket binding cannot be used to make docker available in the context of the job script.

Docker-in-Docker was chosen as an operational decision by the runner team, instead of exposing docker.sock.

Disabling the code quality job

The code_quality job doesn't run if the $CODE_QUALITY_DISABLED environment variable is present. Please refer to the environment variables documentation to learn more about how to define one.

To disable the code_quality job, add CODE_QUALITY_DISABLED as a custom environment variable. This can be done:

For the whole project, in the project settings or CI/CD configuration.
For a single pipeline run:
1. Go to CI/CD > Pipelines
2. Click Run Pipeline
3. Add CODE_QUALITY_DISABLED as the variable key, with any value.

Using with merge request pipelines

The configuration provided by the Code Quality template does not let the code_quality job run on pipelines for merge requests.

If pipelines for merge requests is enabled, the code_quality:rules must be redefined.

The template has these rules for the code quality job:

code_quality:
  rules:
    - if: '$CODE_QUALITY_DISABLED'
      when: never
    - if: '$CI_COMMIT_TAG || $CI_COMMIT_BRANCH'

If you are using merge request pipelines, your rules (or workflow: rules) might look like this example:

job1:
  rules:
    - if: '$CI_PIPELINE_SOURCE == "merge_request_event"' # Run job1 in merge request pipelines
    - if: '$CI_COMMIT_BRANCH == "master"'                # Run job1 in pipelines on the master branch (but not in other branch pipelines)
    - if: '$CI_COMMIT_TAG'                               # Run job1 in pipelines for tags

To make these work together, you need to overwrite the code quality rules so that they match your current rules. From the example above, it could look like:

include:
  - template: Code-Quality.gitlab-ci.yml

code_quality:
  rules:
    - if: '$CODE_QUALITY_DISABLED'
      when: never
    - if: '$CI_PIPELINE_SOURCE == "merge_request_event"' # Run code quality job in merge request pipelines
    - if: '$CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH'      # Run code quality job in pipelines on the master branch (but not in other branch pipelines)
    - if: '$CI_COMMIT_TAG'                               # Run code quality job in pipelines for tags

Configuring jobs using variables

The Code Quality job supports environment variables that users can set to configure job execution at runtime.

For a list of available environment variables, see Environment variables.

Implementing a custom tool

It's possible to have a custom tool provide Code Quality reports in GitLab. To do this:

Define a job in your .gitlab-ci.yml file that generates the Code Quality report artifact.
Configure your tool to generate the Code Quality report artifact as a JSON file that implements a subset of the Code Climate spec.

The Code Quality report artifact JSON file must contain an array of objects with the following properties:

Name	Description
`description`	A description of the code quality violation.
`fingerprint`	A unique fingerprint to identify the code quality violation. For example, an MD5 hash.
`severity`	A severity string (can be `info`, `minor`, `major`, `critical`, or `blocker`).
`location.path`	The relative path to the file containing the code quality violation.
`location.lines.begin`	The line on which the code quality violation occurred.

Example:

[
  {
    "description": "'unused' is assigned a value but never used.",
    "fingerprint": "7815696ecbf1c96e6894b779456d330e",
    "severity": "minor",
    "location": {
      "path": "lib/index.js",
      "lines": {
        "begin": 42
      }
    }
  }
]

NOTE: Although the Code Climate spec supports more properties, those are ignored by GitLab.

Code Quality reports

After the Code Quality job completes:

Potential changes to code quality are shown directly in the merge request. The Code Quality widget in the merge request compares the reports from the base and head of the branch, then lists any violations that are resolved or created when the branch is merged.
The full JSON report is available as a downloadable artifact for the code_quality job.
The full list of code quality violations generated by a pipeline is shown in the Code Quality tab of the Pipeline Details page. (STARTER)

Generating an HTML report

In GitLab 13.6 and later, it is possible to generate an HTML report file by setting the REPORT_FORMAT variable to html. This is useful if you just want to view the report in a more human-readable format or to publish this artifact on GitLab Pages for even easier reviewing.

include:
  - template: Code-Quality.gitlab-ci.yml

code_quality:
  variables:
    REPORT_FORMAT: html
  artifacts:
    paths: [gl-code-quality-report.html]

It's also possible to generate both JSON and HTML report files by defining another job and using extends: code_quality:

include:
  - template: Code-Quality.gitlab-ci.yml

code_quality_html:
  extends: code_quality
  variables:
    REPORT_FORMAT: html
  artifacts:
    paths: [gl-code-quality-report.html]

Extending functionality

Using Analysis Plugins

Should there be a need to extend the default functionality provided by Code Quality, as stated in Code Quality, Analysis Plugins are available.

For example, to use the SonarJava analyzer, add a file named .codeclimate.yml containing the enablement code for the plugin to the root of your repository:

version: "2"
plugins:
  sonar-java:
    enabled: true

This adds SonarJava to the plugins: section of the default .codeclimate.yml included in your project.

Changes to the plugins: section do not affect the exclude_patterns section of the default .codeclimate.yml. See the Code Climate documentation for excluding files and folders for more details.

Here's an example project that uses Code Quality with a .codeclimate.yml file.

Troubleshooting

Changing the default configuration has no effect

A common issue is that the terms Code Quality (GitLab specific) and Code Climate (Engine used by GitLab) are very similar. You must add a .codeclimate.yml file to change the default configuration, not a .codequality.yml file. If you use the wrong filename, the default .codeclimate.yml is still used.

No Code Quality report is displayed in a Merge Request

This can be due to multiple reasons:

You just added the Code Quality job in your .gitlab-ci.yml. The report does not have anything to compare to yet, so no information can be displayed. It only displays after future merge requests have something to compare to.
Your pipeline is not set to run the code quality job on your default branch. If there is no report generated from the default branch, your MR branch reports will not have anything to compare to.
If no degradation or error is detected, nothing is displayed.
The artifacts:expire_in CI/CD setting can cause the Code Quality artifact(s) to expire faster than desired.
If you use the REPORT_STDOUT environment variable, no report file is generated and nothing displays in the merge request.
Large codeclimate.json files (esp. >10 MB) are known to prevent the report from being displayed. As a work-around, try removing properties that are ignored by GitLab. You can:
- Configure the Code Quality tool to not output those types.
- Use sed, awk or similar commands in the .gitlab-ci.yml script to edit the codeclimate.json before the job completes.

Only a single Code Quality report is displayed, but more are defined

GitLab only uses the Code Quality artifact from the latest created job (with the largest job ID). If multiple jobs in a pipeline generate a code quality artifact, those of earlier jobs are ignored. To avoid confusion, configure only one job to generate a codeclimate.json.