Trac Unit Tests
Most of the Python modules in the Trac codebase are accompanied by unit tests. You should run the tests whenever making changes, to be confident you haven't broken anything. Note though that the coverage of application code by the unit tests is incomplete, so not having broken the unit tests does not mean you haven't broken the application! Unit tests do not replace manual testing.
Ideally, also include new unit tests for a change or enhancement, even if you are just submitting a patch. Patches that break the unit tests are a lot less likely to get integrated than patches that add unit tests for the new or changed functionality.
Running the tests
You can run all of the unit tests from the command line using
$ make unit-test
or by invoking
$ python -m trac.test --skip-functional-tests
or the Unix way:
$ PYTHONPATH=. python trac/test.py --skip-functional-tests
This assumes the current working directory is where you checked out the Trac code from the SubversionRepository.
You can also run only the tests for a specific package, module, class or method.
For example, to run the unit tests for the
trac.versioncontrol package, execute:
$ make test=trac.versioncontrol.tests
$ python -m unittest trac.versioncontrol.tests.__init__
or if you prefer bash:
$ PYTHONPATH=. trac/versioncontrol/tests/__init__.py
To run the unit tests for the
trac.versioncontrol.cache module, execute:
$ make test=trac.versioncontrol.tests.cache
$ python -m unittest trac.versioncontrol.tests.cache
or if you prefer bash:
$ PYTHONPATH=. trac/versioncontrol/tests/cache.py
To run the unit tests for the test class
$ python -m unittest trac.versioncontrol.tests.cache.CacheTestCase
To run the test case
$ python -m unittest trac.versioncontrol.tests.cache.TestCase.test_initial_sync
If you've made larger changes, before running the tests please make sure you've cleaned all
.pyc files that may be left after removed or renamed source
$ make clean
$ find . -name *.pyc | xargs rm
If you are developing on a database different from SQLite, you may want to specify its URI using the
TRAC_TEST_DB_URI environment variable.
If you use the Makefile the database can be specified using the
$ make db=postgres test $ make db=mysql test $ make db=sqlite test
make db=sqlite test runs the tests with an on-disk SQLite database.
make test uses an in-memory SQLite database.
The Trac unit tests are also run by the AutomaticBuilds.
If you're adding a new module, or you want to add tests for a module that doesn't have any unit tests yet, you'll need to create a new Python module for the unit tests.
For example, say you want to add tests for the module
trac.foo (which maps to
trac/foo.py). You'll need to create a new module at
trac/tests/foo.py and put the tests there. Also, you'll have to edit the
__init__.py in the
tests package so that your new unit tests get executed with the others.
WritingUnitTests is a tutorial for writing your own unit tests.
Utility code for unit tests
The module trac.test contains a couple of functions and classes that can help writing unit tests. In particular, it provides an
EnvironmentStub class which will allow the tests to run faster than using a real
Environment. There's also a very simple factory for mock objects, which you can use to create quick substitutes of the "real" objects for testing.
Some unit-tests depend on:
In addition, the figleaf package can be used to provide code coverage information:
These can be installed with
pip install pytz pip install Pygments pip install figleaf
If these dependencies are not present, certain tests will be skipped.
For general advice about Trac debugging, see TracTroubleshooting.
ImportError: no module named tests
If you try to run the tests and you receive the following message:
ImportError: No module named tests
It may mean that you have a version of Trac installed in
/usr/local/lib/pythonX.X/site-packages but you are testing a different version installed elsewhere on your machine.
If so, uninstalling the system version of Trac in /usr/lib/pythonX.X should allow you to run the unit tests, testing your private version.
Installing your virtual environment with
--no-site-packages should eliminate this problem.