Happy to help! I use pytest all the time, so it's nice to give some code back too 😄

Codecov Report

Merging #4041 into master will decrease coverage by 0.12%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master    #4041      +/-   ##
==========================================
- Coverage   94.45%   94.33%   -0.13%     
==========================================
  Files         109      109              
  Lines       23788    23792       +4     
  Branches     2357     2359       +2     
==========================================
- Hits        22470    22445      -25     
- Misses       1006     1029      +23     
- Partials      312      318       +6

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b1fbb2a...a089a95. Read the comment docs.

Coverage increased (+0.02%) to 93.811% when pulling a089a95 on Zac-HD:user-properties-type into b1fbb2a on pytest-dev:master.

@nicoddemus - are these coverage tools ever useful? Because the diff clearly doesn't change coverage at all, but one tool has it going up and the other going down!

are these coverage tools ever useful?

We are still evaluating them... can't really answer that at the moment. 🤔

Fair enough! My thoughts:

• Coverage tools that give inconsistent results are just as bad as any other flaky test.
• The important thing about coverage measurement is not percentage, it's "which lines are not covered?"
• 100% is therefore the only percentage target that makes sense; perhaps for a subset of files.

@nicoddemus - Appveyor failure looks like a job failed to upload coverage data, not failing tests 😌

Yeah unfortunately this happens somewhat frequently. 😞

Fair enough! My thoughts:

Definitely, agree with all your points. If codecov keeps behaving like that, I'm afraid we will have to revert back to coveralls (this was discussed briefly in tox-dev/tox#1019 (comment) also).