Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 86.18%. Comparing base (3eb16c0) to head (c8d93ee).

@@            Coverage Diff             @@
##             main    #1489      +/-   ##
==========================================
- Coverage   86.18%   86.18%   -0.01%     
==========================================
  Files         394      393       -1     
  Lines       50299    50277      -22     
  Branches     5309     5309              
==========================================
- Hits        43351    43331      -20     
+ Misses       5494     5493       -1     
+ Partials     1454     1453       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

What is the goal in not running them during “test”. From what I see, this is just moving when the docs are tested to after the other tests are completed.

From a conceptual perspective, I don't like including docs building as part of the test suite. From a practical point of view, since we already have a Docs stage, we can use it as the single place where we build the docs and upload the built docs as artifact.

Moreover, test_docs runs Python code that calls Sphinx through its API, but when we build the docs for publishing we use the targets in docs/Makefile. I think we can stick with a single way of checking that docs build properly.

I also noticed that test_docs don't fail if the linkcheck fails. Running make -C docs linkcheck-html does fail: https://dev.azure.com/simpeg/simpeg/_build/results?buildId=4433&view=logs&j=f5e0b957-a0bb-542d-63f3-48c18f3b89ce&t=6b3e234a-fb10-5df9-0d8b-35a3152b01c6.
So, test_docs is not currently ensuring that all links in the docs work properly.

Lastly, requiring tests to pass before building the docs would reduce the computation load we put in Azure. If test are not passing there's no much sense on building the docs.

I started this PR as a prototype for these ideas. We can continue the discussion.