Add better image testing as a part of CI about rstudio-docker-products HOT 3 CLOSED

Some ideas for things to test:

• R and Python versions in all images (changing them should be considered a breaking change)
• Add licensing to the docker-compose files (from the parent environment, then provide env vars via Actions secrets)
• Start the products (or for RSP, run verify-installation) and ensure that they work
• Expected RStudio UID/GIDs exist and are unchanged (changing UID/GID in the container should be considered a breaking change)
• After startup, expected log files exist (or do not exist, in the case that we are properly echoing to stdout/stderr)
• Any expected sysdeps / R packages / drivers / etc. that we have installed. Changing such things could be considered a breaking change for some users. Some patterns available for testing R packages: https://github.com/rstudio/sol-eng-demo-server/blob/main/goss.yaml and https://github.com/rstudio/sol-eng-demo-server/blob/main/goss_vars_basic.yaml, along with how it's used: https://github.com/rstudio/sol-eng-demo-server/blob/65f0404b4cb9abc218941fd052922b58c6dcb153/Makefile#L16 (in dgoss, which is a bit of a different structure)

from rstudio-docker-products.

Thanks for all the thoughts! I'm a little leery of the licensing parts unless it's only done on the main branch, and I can explain why in Slack to you. In the original description I see "we should see if there is a way to simplify the CI pipeline away from docker Hub's hooks". Is that still the case? I don't remember seeing dockerhub hooks in how this was kicked off right now.

from rstudio-docker-products.

Woops - sorry for the delay getting back to this. What makes you leery of licensing parts on just the main branch? That someone could do a PR from a fork and get the license? Actions never run on/from forks with secrets 😄 I don't know that GitHub has made this very clear, but I am glad they implemented it that way!

Also, that's right - some of the hooks directories are still hanging around, but they are not in use anymore

from rstudio-docker-products.