Hey cool, I've just submitted my first PR against GH, and it's *so* nice to be able to use git, I might have to do this more often <wink>. https://github.com/python/cpython/pull/138 But I have questions about the coverage reports. In bbdef4c, coverage failed because my diff wasn't 100% covered, and the module itself was missing some coverage in code I didn't touch. If you click on the red X next to that commit, you can see the codecov/patch report 94.73% of diff hit. It's expected that some lines may be missed, due to underlying platform support, so the natural thing would be to say so in the code. I.e. suppress some misses so coverage doesn't false positive. This leads to questions: * How is codecov.io actually performing coverage? It seems difficult to find out exactly based on their documentation, and python/cpython's .codecov.yml doesn't reveal any details. Over in #python-dev, nedbat guessed it would be coverage.py (big surprise! :), and that would both make sense and be great, since it's a well-known and well-loved tool. * On that guess, I figured I'd add a few #pragmas for lines which are okay not to be covered. I pushed the new commit and didn't see a codecov.io failure so I thought that did the trick, but I'm actually not so sure. If you look at the above PR, you can see that commit bbdef4c has a red X, clicking on which gives you details that codecov/patch failed but travis-ci/pr succeeded. However, if you hover over the green check next to ccefb3a, all you see is that the travis-ci/pr check succeeded. There's no indication that coverage actually ran at all on ccefb3a. That doesn't make sense to me. * How can you run the same coverage tests locally? It's hard to know, given the apparent dearth of details from codecov.io. The top-level Makefile does have a `coverage` target, but that's a different thing. It would be nice if you didn't have to push a new commit just to see if you've fixed your coverage issues. There needs to be a local way to run the same tests. As I'm writing this, suddenly ccefb3a gained another check! So it seems like codecov.io is quite delayed in reporting its coverage, and the PR results can be misleading in the meantime, because it only reports that there is 1 check until codecov.io completes. I would prefer to see a yellow dot (i.e. the test is still running). Cheers, -Barry