Closed arkwave closed 7 years ago
Problem: the test suite will now get run twice, once without and once with coverage, under Python 3.5. Can we change it so that it only is run once?
We probably also need a command to upload the coverage report to cov.io?
@stefanv I'm not entirely sure, this seems to suggest that integration is handled automatically: https://github.com/codecov/codecov-python (specifically the Configuration section). Thoughts?
It looks like they have the following:
after_success:
- codecov
Hmm I am a little confused in that case. Initially, that same line after_success: codecov
was marked for deletion because it led to the test suite running twice. However, in the scikit-image
repo, additional dependencies are tested with coverage in the same way as they're done here in travis_script.sh
, but after succcess: codecov
is still included in the travis.yml
file. Wouldn't this also lead to coverage running twice, or is there something I'm misunderstanding?
In most cases, that command just won't do anything, since there's no coverage information to upload: https://travis-ci.org/scikit-image/scikit-image/jobs/174340273#L7563
Alright, I have added that line back in. Thanks for catching it. Also, does the cesium-ml repo might have to be registered at codecov.io
for coverage reports to be published? Or is that more to centralize the coverage reports at a single location?
I've registered cesium with them now, so we should be good to go!
Looks like codecov is missing (no pip install?): https://travis-ci.org/cesium-ml/cesium/jobs/175605412#L1617
Also, code coverage is enabled for two items in the matrix, instead of just one.
Excellent, thanks @arkwave !
This does not appear to actually work (see: all the PRs since this was merged).
1) updated travis.yml to run codecov after build passes. 2) added in a .coveragerc file to stop
coverage.py
from testing scikit-learn imports 3) updated requirements.txt to include coverage.