nova/HACKING.rst

Nova Style Commandments

• Step 1: Read the OpenStack Style Commandments https://docs.openstack.org/hacking/latest/
• Step 2: Read on

Nova Specific Commandments

• [N307] nova.db imports are not allowed in nova/virt/*
• [N309] no db session in public API methods (disabled) This enforces a guideline defined in oslo.db.sqlalchemy.session
• [N310] timeutils.utcnow() wrapper must be used instead of direct calls to datetime.datetime.utcnow() to make it easy to override its return value in tests
• [N311] importing code from other virt drivers forbidden Code that needs to be shared between virt drivers should be moved into a common module
• [N312] using config vars from other virt drivers forbidden Config parameters that need to be shared between virt drivers should be moved into a common module
• [N313] capitalize help string Config parameter help strings should have a capitalized first letter
• [N316] Change assertTrue(isinstance(A, B)) by optimal assert like assertIsInstance(A, B).
• [N317] Change assertEqual(type(A), B) by optimal assert like assertIsInstance(A, B)
• [N319] Validate that logs are not translated.
• [N320] Setting CONF.* attributes directly in tests is forbidden. Use self.flags(option=value) instead.
• [N322] Method's default argument shouldn't be mutable
• [N323] Ensure that the _() function is explicitly imported to ensure proper translations.
• [N324] Ensure that jsonutils.%(fun)s must be used instead of json.%(fun)s
• [N326] Translated messages cannot be concatenated. String should be included in translated message.
• [N332] Check that the api_version decorator is the first decorator on a method
• [N334] Change assertTrue/False(A in/not in B, message) to the more specific assertIn/NotIn(A, B, message)
• [N335] Check for usage of deprecated assertRaisesRegexp
• [N336] Must use a dict comprehension instead of a dict constructor with a sequence of key-value pairs.
• [N337] Don't import translation in tests
• [N338] Change assertEqual(A in B, True), assertEqual(True, A in B), assertEqual(A in B, False) or assertEqual(False, A in B) to the more specific assertIn/NotIn(A, B)
• [N339] Check common raise_feature_not_supported() is used for v2.1 HTTPNotImplemented response.
• [N340] Check nova.utils.spawn() is used instead of greenthread.spawn() and eventlet.spawn()
• [N341] contextlib.nested is deprecated
• [N342] Config options should be in the central location nova/conf/
• [N343] Check for common double word typos
• [N348] Deprecated library function os.popen()
• [N349] Check for closures in tests which are not used
• [N350] Policy registration should be in the central location nova/policies/
• [N351] Do not use the oslo_policy.policy.Enforcer.enforce() method.
• [N352] LOG.warn is deprecated. Enforce use of LOG.warning.
• [N353] Validate that context objects is not passed in logging calls.
• [N355] Enforce use of assertTrue/assertFalse
• [N356] Enforce use of assertIs/assertIsNot
• [N357] Use oslo_utils.uuidutils or uuidsentinel(in case of test cases) to generate UUID instead of uuid4().
• [N358] Return must always be followed by a space when returning a value.
• [N359] Check for redundant import aliases.
• [N360] Yield must always be followed by a space when yielding a value.
• [N361] Check for usage of deprecated assertRegexpMatches and assertNotRegexpMatches
• [N362] Imports for privsep modules should be specific. Use "import nova.privsep.path", not "from nova.privsep import path". This ensures callers know that the method they're calling is using priviledge escalation.
• [N363] Disallow (not_a_tuple) because you meant (a_tuple_of_one,).
• [N364] Check non-existent mock assertion methods and attributes.
• [N365] Check misuse of assertTrue/assertIsNone.
• [N366] The assert_has_calls is a method rather than a variable.
• [N367] Disallow aliasing the mock.Mock and similar classes in tests.
• [N368] Reject if the mock.Mock class is used as a replacement value instead of and instance of a mock.Mock during patching in tests.
• [N369] oslo_concurrency.lockutils.ReaderWriterLock() or fasteners.ReaderWriterLock() does not function correctly with eventlet patched code. Use nova.utils.ReaderWriterLock() instead.
• [N370] Don't use or import six
• [N371] You must explicitly import python's mock: from unittest import mock

Creating Unit Tests

For every new feature, unit tests should be created that both test and (implicitly) document the usage of said feature. If submitting a patch for a bug that had no unit test, a new passing unit test should be added. If a submitted bug fix does have a unit test, be sure to add a new one that fails without the patch and passes with the patch.

For more information on creating unit tests and utilizing the testing infrastructure in OpenStack Nova, please read nova/tests/unit/README.rst.

Running Tests

The testing system is based on a combination of tox and stestr. The canonical approach to running tests is to simply run the command tox. This will create virtual environments, populate them with dependencies and run all of the tests that OpenStack CI systems run. Behind the scenes, tox is running stestr run, but is set up such that you can supply any additional stestr arguments that are needed to tox. For example, you can run: tox -- --analyze-isolation to cause tox to tell stestr to add --analyze-isolation to its argument list.

Python packages may also have dependencies that are outside of tox's ability to install. Please refer to Development Quickstart for a list of those packages on Ubuntu, Fedora and Mac OS X.

To run a single or restricted set of tests, pass a regex that matches the class name containing the tests as an extra tox argument; e.g. tox -- TestWSGIServer (note the double-hypen) will test all WSGI server tests from nova/tests/unit/test_wsgi.py; -- TestWSGIServer.test_uri_length_limit would run just that test, and -- TestWSGIServer|TestWSGIServerWithSSL would run tests from both classes.

It is also possible to run the tests inside of a virtual environment you have created, or it is possible that you have all of the dependencies installed locally already. In this case, you can interact with the stestr command directly. Running stestr run will run the entire test suite. stestr run --concurrency=1 will run tests serially (by default, stestr runs tests in parallel). More information about stestr can be found at: http://stestr.readthedocs.io/

Since when testing locally, running the entire test suite on a regular basis is prohibitively expensive, the tools/run-tests-for-diff.sh script is provided as a convenient way to run selected tests using output from git diff. For example, this allows running only the test files changed/added in the working tree:

tools/run-tests-for-diff.sh

However since it passes its arguments directly to git diff, tests can be selected in lots of other interesting ways, e.g. it can run all tests affected by a single commit at the tip of a given branch:

tools/run-tests-for-diff.sh mybranch^!

or all those affected by a range of commits, e.g. a branch containing a whole patch series for a blueprint:

tools/run-tests-for-diff.sh gerrit/master..bp/my-blueprint

It supports the same -HEAD invocation syntax as flake8wrap.sh (as used by the fast8 tox environment):

tools/run-tests-for-diff.sh -HEAD

By default tests log at INFO level. It is possible to make them log at DEBUG level by exporting the OS_DEBUG environment variable to True.

Building Docs

Normal Sphinx docs can be built via the setuptools build_sphinx command. To do this via tox, simply run tox -e docs, which will cause a virtualenv with all of the needed dependencies to be created and then inside of the virtualenv, the docs will be created and put into doc/build/html.

Building a PDF of the Documentation

If you'd like a PDF of the documentation, you'll need LaTeX and ImageMagick installed, and additionally some fonts. On Ubuntu systems, you can get what you need with:

apt-get install texlive-full imagemagick

Then you can use the build_latex_pdf.sh script in tools/ to take care of both the sphinx latex generation and the latex compilation. For example:

tools/build_latex_pdf.sh

The script must be run from the root of the Nova repository and it'll copy the output pdf to Nova.pdf in that directory.