As noted by @ianhbell via email: "It would be nice to have the chance to do some things pre-borging, for instance running doxygen and executing notebooks after borging is much more complicated with the fact that the doc folder has moved during borging and the hard-coded paths all need to be fixed up throughout my scripts and config files. It would be nice if this could run after the conda environment is loaded."

As noted by @ianhbell via email: "pulling the docker container every time makes the builds pretty slow from start to finish. There seem to be some means of caching the docker image: https://stackoverflow.com/questions/66421411/how-to-run-cached-docker-image-in-github-action"

As noted in email from @ianhbell, "the initial checkout of the code by default does not pull submodules"

When I try to run the action against a repo set to use rtd_sphinx_theme or another theme with custom CSS like piccolo_theme, the browser looks for /_static/css/rtd_sphinx_theme.css or /_static/css/piccolo_theme.css rather than what those packages actually call the file, iirc theme.css. Using alabaster works fine, likely because it doesn't have custom css to link in.

As requested by @ianhbell via email: "Do you think it would be possible to allow the user to specify the entire environment.yml, rather than interleaving the base one with the user-specified one? My current problem is a conflict between sphinx=7.1.2, which does not compile my pdf properly, and sphinx=7.2.6, which does compile, but which conflicts with some package, though I don’t know which one. I am able to build the docs with just my environment on my local machine with python=3.10 so it is something about the combination that is the problem. The conda file in the docker gives python=3.11 and I need python=3.10 to get everybody to play nicely together."

As noted by @ianhbell via email: "The upload of PDF output for some reason is incorrectly looking for files like latex/teqp-docs.pdf but the action seems to be looking correctly for latex/teqp.pdf: https://github.com/usnistgov/NISTtheDocs2Death/blob/a8ee6b552afc67d4894218e5577e5d3237cf98a6/action.yml#L177-L179 . As a result, no latex/PDF output gets uploaded. I have no idea why this is happening as the NTD2D action seems correct and I think the PDF is getting generated properly, but there is zero shell output from the PDF build step so ???"

I am not sure where this happens, but you have unicode in your generated rst, it will fail to compile to PDF, you need to use instead xelatex which handles unicode usually flawlessly.

Removing all numbered tags should remove stable, but it doesn't

As reported by @ianhbell via email: "There is something wrong in @main with pre-build. The pre-build seems to be interpreted in a different way than before, and it doesn’t find python to run my pre-build script. I thought maybe it was the multi-line pre-build script you helped me set up before, so I tried to pull it all into a single python script, but that doesn’t fix it either, it can’t find python. Perhaps it is how subprocess is being used to call the pre-build command. See: https://github.com/ianhbell/teqp-docs/actions/runs/6331159254/job/17195114598"

This happened after intentionally removing shell=True. Maybe interpretation of multi-line YAML strings is different?