Codecov Report

Attention: Patch coverage is 69.23077% with 12 lines in your changes missing coverage. Please review.

Project coverage is 26.36%. Comparing base (33cb296) to head (c426adc).
Report is 4 commits behind head on main.

@@            Coverage Diff             @@
##             main     #169      +/-   ##
==========================================
+ Coverage   25.50%   26.36%   +0.86%     
==========================================
  Files          29       29              
  Lines        4027     4085      +58     
==========================================
+ Hits         1027     1077      +50     
- Misses       3000     3008       +8     

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

The failure on macos seems to be due to a change in newer version of numpy or scipy on macos (but not linux), the difference is in the ~10 significant decimal digit as far as I can tell, but I'm not sure what is causing this!

@pmav99 suggested using xfail to better track the current expected macos failure and to later know if a numpy change fixes the issue (to then remove xfail).

I like this idea, but I still feel that due to the nature of the failure, and the fact that pretty much ALL test fail on macos, this approach just masks the issue, maybe it makes more sense to keep the failure (to show up in the PR) but do not require pass for the macos tests. This also requires the coverage test to be based on linux test passing, and not both mac and linux tests success.

Maybe then it's better to do the equal check using values rather than lines of written files. This could also be it's own rabbit hole!