Writing Python Tests
Use Python as the first choice
Samba provides extensive Python bindings and test infrastructure, so if possible please write tests in Python as a first choice. Python provides exceptions and early returns, ensuring that subsequent code doesn't run after prerequisite checks have failed, many helper functions as well as setUp() and tearDown() routines allowing unit tests to be easily built.
Likewise LDAP is easily accessed via LDB.
Writing Python Tests for Samba
Python unittest.TestCase is the standard basis on which all of Samba's python based tests should be written.
This provides helpful routines such as assertTrue() etc, which provide a cosnsitent pattern in the tests.
Samba's subclass samba.tests.TestCase, provides additional helper routines such as env_loadparm()
You can explore the methods provided using pydoc:
PYTHONPATH=bin/python pydoc samba.tests.TestCase
Testing Samba binaries
When testing Samba binaries, please use:
Obtaining credentials for a sub process
A pattern like this is typical for passing credentials to a subprocess
creds = self.get_credentials() cmd_line_auth = "-U%s/%s%%%s" % (creds.get_domain(), creds.get_username(), creds.get_password())
Running the subprocess
Use self.check_run or self.check_output
self.check_run("ndrdump samr samr_CreateUser in %s" % (self.data_path("samr-CreateUser-in.dat")))
samba-tool is a special case, as it is written in python. To avoid a fork()/exec() of python from python, use
The runcmd() and runsubcmd() methods allows testing of a samba-tool command in the same process, which also allows the use of
This example from source4/torture/drs/python/fsmo.py shows how to run an authenticated subcommand without forcing a new kinit:
ccache_name = self.get_creds_ccache_name() cmd_line_auth = "--krb5-ccache=%s" % ccache_name (result, out, err) = self.runsubcmd("fsmo", "transfer", "--role=%s" % role, "-H", "ldap://%s:389" % DC, cmd_line_auth)
If you need a temp dir, use
descend from this class, so the self.tempdir attribute is available with a temporary directory. The framework will assert that directory is empty when the test is done.
Where to add new tests
Samba Tests are scattered around the code. However, in general, unless already related to or derived from other tests, new tests should be added in
Tests should be run using subunitrun if possible, declared in tests.py via planpythontestsuite(), not planoldpythontestsuite()
Try not to make your new test special. The more our tests look like each other, the easier it is for new developers to follow the same patterns.
Mysterious error messages
AttributeError: 'module' object has no attribute 'foo'
A traceback that ends like this:
Traceback (most recent call last): [...] File "bin/python/samba/subunit/run.py", line 603, in createTests self.module) File "/usr/lib/python2.7/unittest/loader.py", line 130, in loadTestsFromNames suites = [self.loadTestsFromName(name, module) for name in names] File "/usr/lib/python2.7/unittest/loader.py", line 100, in loadTestsFromName parent, obj = obj, getattr(obj, part) AttributeError: 'module' object has no attribute 'foo'
usually means there was an exception raised when the test framework was importing your
foo.py test module. This means an error at the module's top level, or a syntax error in compilation, not an error in the actual tests (which are not run at this stage).
If the test runs fine for you locally but not in autobuild, the problem could be an
ImportError, caused by the use of a module that is not in the Python standard library. If the module is in Samba's
thirdparty directory, make sure you use this pattern:
import samba samba.ensure_third_party_module("dns", "dnspython") import dns.resolver
Debugging Python tests
Python tests that use subunit can be run directly by executing:
PYTHONPATH=bin/python python source4/scripting/bin/subunitrun TEST
PYTHONPATH=bin/python python source4/scripting/bin/subunitrun samba.tests.docs