You are viewing the latest unreleased documentation v3.3.dev0. You may prefer a stable version.

Continuous Integration (CI) Testing#


Iris is currently supported and tested against Python 3.8 running on Linux. We do not currently actively test on other platforms such as Windows or macOS.

The Iris GitHub repository is configured to run checks against all its branches automatically whenever a pull-request is created, updated or merged. The checks performed are:


Iris unit and integration tests are an essential mechanism to ensure that the Iris code base is working as expected. Running the Tests may be performed manually by a developer locally. However Iris is configured to use the cirrus-ci service for automated Continuous Integration (CI) testing.

The cirrus-ci configuration file .cirrus.yml in the root of the Iris repository defines the tasks to be performed by cirrus-ci. For further details refer to the Cirrus-CI Documentation. The tasks performed during CI include:

  • linting the code base and ensuring it adheres to the black format

  • running the system, integration and unit tests for Iris

  • ensuring the documentation gallery builds successfully

  • performing all doc-tests within the code base

  • checking all URL references within the code base and documentation are valid

The above cirrus-ci tasks are run automatically against all Iris branches on GitHub whenever a pull-request is submitted, updated or merged. See the Cirrus-CI Dashboard for details of recent past and active Iris jobs.

Cirrus CI Test environment#

The test environment on the Cirrus-CI service is determined from the requirement files in requirements/ci/py**.yml. These are conda environment files that list the entire set of build, test and run requirements for Iris.

For reproducible test results, these environments are resolved for all their dependencies and stored as lock files in requirements/ci/nox.lock. The test environments will not resolve the dependencies each time, instead they will use the lock file to reproduce the same exact environment each time.

If you have updated the requirement yaml files with new dependencies, you will need to generate new lock files. To do this, run the command:

python tools/ -o requirements/ci/nox.lock requirements/ci/py*.yml

or simply:

make lockfiles

and add the changed lockfiles to your pull request.


If your installation of conda runs through Artifactory or another similar proxy then you will need to amend that lockfile to use URLs that Github Actions can access. A utility to strip out Artifactory exists in the ssstack tool.

New lockfiles are generated automatically each week to ensure that Iris continues to be tested against the latest available version of its dependencies. Each week the yaml files in requirements/ci are resolved by a GitHub Action. If the resolved environment has changed, a pull request is created with the new lock files. The CI test suite will run on this pull request. If the tests fail, a developer will need to create a new branch based off the auto-update-lockfiles branch and add the required fixes to this new branch. If the fixes are made to the auto-update-lockfiles branch these will be overwritten the next time the Github Action is run.

Skipping Cirrus-CI Tasks#

As a developer you may wish to not run all the CI tasks when you are actively developing e.g., you are writing documentation and there is no need for linting, or long running compute intensive testing tasks to be executed.

As a convenience, it is possible to easily skip one or more tasks by setting the appropriate environment variable within the .cirrus.yml file to a non-empty string:

  • SKIP_LINT_TASK to skip flake8 linting and black formatting

  • SKIP_TEST_MINIMAL_TASK to skip restricted unit and integration testing

  • SKIP_TEST_FULL_TASK to skip full unit and integration testing

  • SKIP_GALLERY_TASK to skip building the documentation gallery

  • SKIP_DOCTEST_TASK to skip running the documentation doc-tests

  • SKIP_LINKCHECK_TASK to skip checking for broken documentation URL references



e.g., to skip the linting task, the following are all equivalent:

SKIP_LINT_TASK: "unicorn"

GitHub Checklist#

An example snapshot from a successful GitHub pull-request shows all tests passing:


If any CI tasks fail, then the pull-request is unlikely to be merged to the Iris target branch by a core developer.

SciTools CLA Checker#

A bot which checks that the GitHub author of the pull-request has signed the SciTools Contributor’s License Agreement (CLA). For more information on this please see

pre-commit CI#

A CI service for the pre-commit framework that checks and auto fixes all pull-requests given the Iris GitHub repository .pre-commit-config.yaml.

See the dashboard for details of recent past and active Iris jobs.