Improving linkcheck

[eviau-artefactual]

Per sphinx-doc/sphinx#8791 (comment) , adding the -q or quiet option flag to the Sphinx options will allow us to only print out errors (broken URLs).

[eviau-artefactual]

Testing this: sphinx-doc/sphinx#10643

[eviau-artefactual]

It worked - with a few changes from the action set up in aiven/devportal#1181

The license is https://github.com/aiven/devportal?tab=CC-BY-4.0-1-ov-file#readme - where/how should we make sure we do the attribution ?

[eviau-artefactual]

Ok now it should work... but we will only know for sure when we make a change to a .rst file...

[replaceafill]

@eviau-artefactual After testing your branch a bit I realized that for my original pain point of linkcheck being too noisy I can instead use make linkcheck SPHINXOPTS='-q' which I think shows me only errors and it's supported by the current Makefile.

Also, don't you think that running the linkcheck target only on changed files could prevent us from detecting broken links early coming from unchanged files?

[eviau-artefactual]

About the -q option: note that we are still using it, it's on line 192 in the Makefile - https://github.com/artefactual/archivematica-docs/pull/463/files#diff-76ed074a9305c04054cdebb9e9aad2d818052b07091de1f20cad0bbac34ffb52R192 should be a direct link :)

If there were no external contributors to the documentation, e.g. if it was the work of a single person - it would make sense to linkhceck every link every time... The way I see this, is that right now, PRs can fail tests for things that are not related to the PR itself. If we can figure out a better way to do linkcheck, that is more granular, I think it would be best!

Maybe this conversation belongs in an issue ? If so, I can open one.

[replaceafill]

About the -q option: note that we are still using it, ...

I think my point is that we have always been able to set the -q flag for the existing linkcheck target. I just didn't know how to do it using SPHINXOPTS.

The way I see this, is that right now, PRs can fail tests for things that are not related to the PR itself.

Maybe that's what we need to fix.

Maybe this conversation belongs in an issue ? If so, I can open one.

Definitely and thank you! I think we've tried to address my pain points and yours in a single PR 😅

[eviau-artefactual]

Issue 1690 reference the verbose results and makes them quiet.

Issue 1689 will follow through on the implementation of linkcheck.

Closing this PR.