fix(ci): sitemap generation

[vados-cosmonic]

No description provided.

[vados-cosmonic]

Hey @itowlson if we want to revert I think we have to go back to here 24d7eb9, which is a bit difficult to do with the merge commits in the history (and a long way back).

I'm trying to fix the python step instead -- it relies on relative paths so the output directory change broke that too unfortunately. This PR fixes that, but again is unfortunately another ~4min wait to ensure it works properly (in CI -- I already checked locally and it works).

[vados-cosmonic]

If you wouldn't mind just one more time 🤞 , I think this will finally fix all the steps for deploy.

[itowlson]

Thanks and belatedly merged!

[vados-cosmonic]

Thanks for the patience here, glad to note that the docs are back 😅

It's unfortunate it took this many steps to iron out CI -- hopefully I can make this process even more error-proof as part of #226 . I think it might be useful to:

• Change (and maybe factor out) the build process to produce a zip file artifact
• Introduce a top level script file (I am partial to just as an ergonomic make replacement) so that we stop cding around in CI
• Update the python script to only deal with absolute/script-relative paths (running via just obviates this but this is a nice to have)
• Refactor publish to only take the artifact and then deploy to pages
• Enable publish workflow to work on any commit (this way, we can "roll back" without actually rolling back code), so we could have just rolled back to a working commit the first time CI went wrong 😮‍💨

[itowlson]

Thank you for working through the series of fixes! Your suggestions for making this more reliable and understandable sound good - in particular I'm 100% up for running a lot more of this stuff during the PR process. I also like the idea of a manual publish workflow for rapid rollback - that speaks to my risk-averse soul!