@tacaswell @QuLogic If this is still of interest to be able to build the docs in this Docker image I can have the CI system that I showed in the SciPy sprints Slack chat do this and then deploy the built docs to wherever.

Cloning of the MPL repo may take a long time

The actions/checkout option makes this very fast. c.f. screenshot of working built in CI:

Thanks for the enhancement suggestions, @QuLogic! I accepted them. Would be great to have a proper Docker Entrypoint. Something like https://github.com/conda-forge/docker-images/blob/fe1a5326926fd75bb90bfa7bfe9e39e6509ab9f1/linux-anvil-comp7/Dockerfile#L49-L58?

Would be great to have a proper Docker Entrypoint. Something like https://github.com/conda-forge/docker-images/blob/fe1a5326926fd75bb90bfa7bfe9e39e6509ab9f1/linux-anvil-comp7/Dockerfile#L49-L58?

It is still a bit unclear to me what the intended use case of the Docker image is, but even if the purpose is only "build the docs" it seems like using the entrypoint.sh as ENTRYPOINT would be a bit heavy, as then if you wanted to use the image to build the docs locally you would have to override the ENTRYPOINT. Additionally, there are times when having the image clone down matplotlib might not be desirable. For example, if building the docs in CI, something like GitHub Actions's checkout action can be faster.

Indeed, valid points. Maybe the originally suggested flexibility with the bash command is sufficient.