Libvirt Continuous Integration

The libvirt project uses GitLab CI for automated testing. Here's our CI dashboard which shows the current status of our pipelines.

Builds and unit tests

Linux builds and cross-compiled Windows builds happen on GitLab CI's shared runners, while FreeBSD and macOS coverage is achieved by triggering Cirrus CI jobs behind the scenes.

Most of the tooling used to build CI pipelines is maintained as part of the libvirt-ci subproject.

Integration tests

Integration tests in our CI pipelines require dedicated HW which is not available to forks, see GitLab CI Custom Runners. Therefore, in order to execute the integration tests as part of your libvirt fork's GitLab CI you'll need to provide your own runner. You'll also need to set a few CI variables to run the integration tests as part of the CI pipeline, see below.

GitLab CI variables

  • LIBVIRT_CI_INTEGRATION - enables integration test suite execution as part

    of the pipeline (works in forks too if there's a registered shared runner)

  • LIBVIRT_CI_INTEGRATION_RUNNER_TAG - overrides the gitlab-runner tag for

    the pipeline; this is needed in forks because the default tag only matches upstream shared runners which may be (and will be) different from the tags defined on shared runners in forks

Retrieving test logs

In case the integration test suite fails in our CI pipelines, a job artifact is generated containing Avocado logs, libvirt debug logs, and the latest traceback (if one was produced during a daemon's execution).

Adding new OS platforms OR build pre-requisites

Since all of the Dockerfiles libvirt uses for CI have been generated by lcitool provided by the libvirt-ci project, most relevant changes will need to be introduced to lcitool first. Please follow the instructions outlined here

Running CI workloads locally

If you're interested in running the CI test workloads locally, please read our testing guide.