Instead of globally ignoring pyflakes and
hacking warnings, only blacklist those that trigger
very frequently so far, in order to clean them
up in followup commits. Fix and start gating
on the rest already.
Tox can manage virtualenvs, and is currently doing so for running
tests in Jenkins. It's just as, or more, useful for running tests
locally, so this starts the migration from the run_tests system to
tox. The goal is to reduce duplicate testing infrastructure, and
get what's running locally on developer workstations as close to
what is run by Jenkins as possible.
This patch removes run_tests.py, and the scripts that manage .venv.
It makes run_tests.sh call tox to facilitate the transition for
developers used to typing "run_tests.sh".
Developers will need tox installed on their workstations. It can
be installed from PyPI with "pip install tox". run_tests.sh outputs
those instructions if tox is not present.
New facilities are available using tox directly, including:
tox -e py26 # run tests under python 2.6
tox -e py27 # run tests under python 2.7
tox -e pep8 # run pep8 tests
tox # run all of the above
tox -e venv foo # run the command "foo" inside a virtualenv
The configuration of the openstack nose plugin is removed from
setup.cfg and added to the nosetests command line arguments in tox.
It is used when running tox from the command line, so the enhanced,
colorized output is visible to developers running the test suite
locally. However, when Jenkins runs tox, the xunit plugin will be
used instead, providing output natively understood by jenkins which
is much more readable in that context.