While unittests work well for testing facets of an implementation, they fail to
provide assurances that the user-visible functions work in practice. Here, we
complement the unittests with functional tests that drive the system as a user
would to verify user visible functionality. These functional tests are run as
part of the unittests.
So, we use Twill to verify Trac's functionality as served by tracd (and in the
future, other frontends).
Unlike most unittests, we setup a single fixture against which we run all the
testcases. This is for two reasons: Primarily, that provides us with a more
complex set of data to test against and thus more room for triggering bugs.
Secondarily, the cost of setting up a new Trac environment and Subversion
repository is significant, so recreating the fixture for each test would be
There are two primary objects involved in the testing, the
FunctionalTestEnvironment and the FunctionalTester.
FunctionalTestEnvironment represents the Trac environment, the Subversion
repository, and the server. The server will be run on a random local port in
the range 8000-8999. A subdirectory named 'tracenv' will be created containing
the Trac environment, Subversion repository, and the user authentication
information. An 'admin' user is created and given TRAC_ADMIN privs early in
the testing. There are other users added as well. All accounts are setup with
a password equalling the username. The test environment is left behind after
the testing has completed to assist in debugging.
FunctionalTester provides code reuse for the testcases to allow a higher-level
description of the more complicated bugs. For example, creating a new ticket
is the first step in regression testing many things, so FunctionalTester
provides a create_ticket() method. That method is written as if it were itself
a testcase for creating a ticket, so there is a testcase that simply calls that
method, and other testcases that use it as a higher-level step don't have to
worry about basic issues such as if the ticket was successfully created.
- Twill (http://twill.idyll.org/)
- lxml for XHTML validation (optional)
from trac.tests.functional.compat import close_fds, rmtree
# Handle missing twill so we can print a useful 'SKIP'
# message. We import subprocess first to allow customizing it on Windows
# to select pywin32 in favor of _subprocess for low-level calls. If Twill
# is allowed to load first, its (unmodified) copy will always be loaded.
from better_twill import twill, b, tc, ConnectError
# This is the first indicator of whether the subversion bindings are
# correctly installed.
from svn import core
has_svn = True
has_svn = False
from datetime import datetime, timedelta
from trac.tests.contentgen import random_sentence, random_page, random_word, \
from trac.test import TestSetup, TestCaseSetup
internal_error = 'Trac detected an internal error:'
trac_source_tree = os.path.normpath(os.path.join(trac.__file__, '..', '..'))
# testing.log gets any unused output from subprocesses
logfile = open(os.path.join(trac_source_tree, 'testing.log'), 'w')
# functional-testing.log gets the twill output
from trac.tests.functional.testenv import FunctionalTestEnvironment
from trac.tests.functional.svntestenv import SvnFunctionalTestEnvironment
from trac.tests.functional.tester import FunctionalTester
"""TestSuite that provides a test fixture containing a
FunctionalTestEnvironment and a FunctionalTester.
env_class = SvnFunctionalTestEnvironment
env_class = FunctionalTestEnvironment
def setUp(self, port=None):
"""If no port is specified, use a semi-random port and subdirectory
'testenv'; but if a port is specified, use that port and
if port is None:
port = 8000 + os.getpid() % 1000
dirname = "testenv"
dirname = "testenv%s" % port
dirname = os.path.join(trac_source_tree, dirname)
baseurl = "http://127.0.0.1:%s" % port
self._testenv = self.env_class(dirname, port, baseurl)
self._tester = FunctionalTester(baseurl)
self.fixture = (self._testenv, self._tester)
"""Convenience class to expand the fixture into the _testenv and
self._testenv, self._tester = self.fixture
failureException = twill.errors.TwillAssertionError
# We're going to have to skip the functional tests
# Twill's find command accepts regexes; some convenient but complex regexes
# & regex factories are provided here (only one so far):
return '(Owned by:(<[^>]*>|\\n| )*%s)' % username
from trac.tests.functional.testcases import suite
suite = suite()
diagnostic = "SKIP: functional tests"
if not twill:
diagnostic += " (no twill installed)"
# No tests to run, provide an empty suite.
suite = unittest.TestSuite()
if __name__ == '__main__':