nova/doc/source/devref/unit_tests.rst
Duncan McGreggor 5f72723b41 Add missing documentation for shared folder issue with unit tests and Python
lock file.

Addresses documentation need raised in bug 897155

Change-Id: Idc95c12f939948bbbefd84a79b3785e3c2fd752c
2011-12-05 09:54:31 -08:00

160 lines
5.5 KiB
ReStructuredText

Unit Tests
==========
Nova contains a suite of unit tests, in the nova/tests directory.
Any proposed code change will be automatically rejected by the OpenStack
Jenkins server [#f1]_ if the change causes unit test failures.
Running the tests
-----------------
Run the unit tests by doing::
./run_tests.sh
This script is a wrapper around the `nose`_ testrunner and the `pep8`_ checker.
.. _nose: http://code.google.com/p/python-nose/
.. _pep8: https://github.com/jcrocholl/pep8
Flags
-----
The ``run_tests.sh`` script supports several flags. You can view a list of
flags by doing::
run_tests.sh -h
This will show the following help information::
Usage: ./run_tests.sh [OPTION]...
Run Nova's test suite(s)
-V, --virtual-env Always use virtualenv. Install automatically if not present
-N, --no-virtual-env Don't use virtualenv. Run tests in local environment
-s, --no-site-packages Isolate the virtualenv from the global Python environment
-r, --recreate-db Recreate the test database (deprecated, as this is now the default).
-n, --no-recreate-db Don't recreate the test database.
-x, --stop Stop running tests after the first error or failure.
-f, --force Force a clean re-build of the virtual environment. Useful when dependencies have been added.
-p, --pep8 Just run pep8
-P, --no-pep8 Don't run pep8
-c, --coverage Generate coverage report
-h, --help Print this usage message
--hide-elapsed Don't print the elapsed time for each test along with slow test list
Because ``run_tests.sh`` is a wrapper around nose, it also accepts the same
flags as nosetests. See the `nose options documentation`_ for details about
these additional flags.
.. _nose options documentation: http://readthedocs.org/docs/nose/en/latest/usage.html#options
Running a subset of tests
-------------------------
Instead of running all tests, you can specify an individual directory, file,
class, or method that contains test code.
To run the tests in the ``nova/tests/scheduler`` directory::
./run_tests.sh scheduler
To run the tests in the ``nova/tests/test_libvirt.py`` file::
./run_tests.sh test_libvirt
To run the tests in the `HostStateTestCase` class in
``nova/tests/test_libvirt.py``::
./run_tests.sh test_libvirt:HostStateTestCase
To run the `ToPrimitiveTestCase.test_dict` test method in
``nova/tests/test_utils.py``::
./run_tests.sh test_utils:ToPrimitiveTestCase.test_dict
Suppressing logging output when tests fail
------------------------------------------
By default, when one or more unit test fails, all of the data sent to the
logger during the failed tests will appear on standard output, which typically
consists of many lines of texts. The logging output can make it difficult to
identify which specific tests have failed, unless your terminal has a large
scrollback buffer or you have redirected output to a file.
You can suppress the logging output by calling ``run_tests.sh`` with the nose
flag::
--nologcapture
Virtualenv
----------
By default, the tests use the Python packages installed inside a
virtualenv [#f2]_. (This is equivalent to using the ``-V, --virtualenv`` flag).
If the virtualenv does not exist, it will be created the first time the tests are run.
If you wish to recreate the virtualenv, call ``run_tests.sh`` with the flag::
-f, --force
Recreating the virtualenv is useful if the package dependencies have changed
since the virtualenv was last created. If the ``tools/pip-requires`` or
``tools/install_venv.py`` files have changed, it's a good idea to recreate the
virtualenv.
By default, the unit tests will see both the packages in the virtualenv and
the packages that have been installed in the Python global environment. In
some cases, the packages in the Python global environment may cause a conflict
with the packages in the virtualenv. If this occurs, you can isolate the
virtualenv from the global environment by using the flag::
-s, --no-site packages
If you do not wish to use a virtualenv at all, use the flag::
-N, --no-virtual-env
Database
--------
Some of the unit tests make queries against an sqlite database [#f3]_. By
default, the test database (``tests.sqlite``) is deleted and recreated each
time ``run_tests.sh`` is invoked (This is equivalent to using the
``-r, --recreate-db`` flag). To reduce testing time if a database already
exists it can be reused by using the flag::
-n, --no-recreate-db
Reusing an existing database may cause tests to fail if the schema has
changed. If any files in the ``nova/db/sqlalchemy`` have changed, it's a good
idea to recreate the test database.
Gotchas
-------
**Running Tests from Shared Folders**
If you are running the unit tests from a shared folder, you may see tests start
to fail or stop completely as a result of Python lockfile issues [#f4]_. You
can get around this by manually setting or updating the following line in
``nova/tests/fake_flags.py``::
FLAGS['lock_path'].SetDefault('/tmp')
Note that you may use any location (not just ``/tmp``!) as long as it is not
a shared folder.
.. rubric:: Footnotes
.. [#f1] See :doc:`jenkins`.
.. [#f2] See :doc:`development.environment` for more details about the use of
virtualenv.
.. [#f3] There is an effort underway to use a fake DB implementation for the
unit tests. See https://lists.launchpad.net/openstack/msg05604.html
.. [#f4] See Vish's comment in this bug report: https://bugs.launchpad.net/nova/+bug/882933