1) Stop using --skip-missing-interpreters with tox.
Since all interpreters should now be present, and using this pref hides any issues with the tox config.
2) Run the tests for each Python version in parallel
Previously the Travis tests were run in a single ~33 minute job. With this change the wall time for the entire test run is now only ~9 minutes, and the cumulative time across all sub-jobs is still only ~33-34 minutes (not counting the new Python 3.5 job).
The TOXENV variable is set using a bit of a hack, since tox does not support wildcards (either with TOXENV or when using -e). The output of --listenvs is filtered by the current Python version (in the form pyNN) and then converted from multiple lines to a comma-separated list.
3) Test peep on Python 3.5
And add the appropriate PyPI classifier.
erikrose
commented
8 years ago
1) Stop using --skip-missing-interpreters with tox. Since all interpreters should now be present, and using this pref hides any issues with the tox config.
2) Run the tests for each Python version in parallel Previously the Travis tests were run in a single ~33 minute job. With this change the wall time for the entire test run is now only ~9 minutes, and the cumulative time across all sub-jobs is still only ~33-34 minutes (not counting the new Python 3.5 job).
The TOXENV variable is set using a bit of a hack, since tox does not support wildcards (either with TOXENV or when using
-e). The output of--listenvsis filtered by the current Python version (in the formpyNN) and then converted from multiple lines to a comma-separated list.3) Test peep on Python 3.5 And add the appropriate PyPI classifier.