I'm currently trying to run cockpit's integration tests on the dist-git pipeline. My current approach is to run the test itself in a cockpit/tests podman container, the most important ingredient is chromium-browser. This avoids having to install chromium and a few others onto the actual test subject, to avoid influencing the tested cockpit OS. This nicely keeps the test and tested subject apart.

Right now I'm pulling directly from hub.docker.com. While this is historically rather stable, it does introduce a point of instability, and also takes quite some time/bandwidth when doing it for every test. It would be nice if the pipeline infrastructure (which runs on OpenShift anyway) would offer a local container registry to tests, which they can use to cache images. This could be exported through an environment or ansible variable, so that tests can do something like

if ! podman pull ${REGISTRY_CACHE}/project/container; then
    podman pull docker.io/project/container
    # cache  for next time
    podman tag docker.io/project/container ${REGISTRY_CACHE}/project/container
    podman push ${REGISTRY_CACHE}/project/container  
fi