1. Alex Willmer
  2. trac-ticketlinks


trac-ticketlinks / TESTING-README

rblank 4022c4b 

 = Testing Readme =

So, you want to see what's broken?  Eeeexcellent.

If you are running python < 2.4.4, please see the troubleshooting section.

 == Quick Start ==
First thing to do is run the tests.
If you have genshi and twill installed on your system, you should be able to run the tests like this:
PYTHONPATH=. ./trac/test.py

 == Slow Start ==
If you want to test against specific versions of genshi, twill, pygments, etc., you can set those up like this:

Run `python setup egg_info` in those subdirectories as needed.
Then you can run:
PYTHONPATH=.:../twill-0.9:../genshi-0.4.4:../pygments-0.8 ./trac/test.py

If you want to run just the functional tests, you can do that by running
PYTHONPATH=.:../twill-0.9:../genshi-0.4.4:../pygments-0.8 ./trac/tests/functional/testcases.py

And to run everything except the functional tests,
PYTHONPATH=. ./trac/test.py --skip-functional-tests

NOTE: Unlike most unittests, the functional tests share a test fixture across tests.  This means that you can't(*) run just one of the tests by itself.
But you can run a sub-set of the functional tests:
PYTHONPATH=.:../twill-0.9:../genshi-0.4.4:../pygments-0.8 ./trac/tests/functional/__init__.py
PYTHONPATH=.:../twill-0.9:../genshi-0.4.4:../pygments-0.8 ./trac/ticket/tests/functional.py
PYTHONPATH=.:../twill-0.9:../genshi-0.4.4:../pygments-0.8 ./trac/versioncontrol/tests/functional.py
Note that that there is a set of functional tests that are run regardless of what subset you choose; these tests setup the shared fixture.

(*) Unless you modify the code to comment out the other functional tests.  The test fixture is setup and torn down by `FunctionalTestSuite`, and it runs the tests added to it with `_tester` and `_testenv` set in the testcase objects.

The functional tests require subversion, and use a random local port 8000-8999 for the test web server.

 == Testing output and byproducts ==
There is some logging done:
 - testing.log
    output from trac environment creation, tracd, and some svn commands
 - functional-testing.log
    output from twill

The test fixture is left behind in 'testenv' so you can inspect it when debugging a problem.
testenv/htpasswd   -- the password/authentication for the test fixture.  password = username
       /repo       -- the Subversion repository
       /trac       -- the Trac environment
(Note that running the tests again will automatically delete this test environment and create a new one.  If you want to save a test environment, you will need to rename this directory before running the tests again.)

The command to serve the test environment is:
PYTHONPATH=.:../twill-0.9:../genshi-0.4.4:../pygments-0.8 \
./trac/web/standalone.py --basic-auth=trac,testenv/htpasswd, -s \
    --port=8888 testenv/trac
This is particularly useful when a test fails and you want to explore the environment to debug the problem.

 == Test Coverage ==
You can now determine statement coverage of unittests and functional tests.  But you'll have to run them separately for now.

`figleaf` will need to be on your `PATH`.  Set the `FIGLEAF` environment variable to the figleaf command.

Create a figleaf-exclude file with filename patterns to exclude.  For example:

Then run the tests something like this:
export FIGLEAF=figleaf
figleaf ./trac/test.py -v --skip-functional-tests
mv .figleaf .figleaf.unittests
python trac/tests/functional/testcases.py -v
mv .figleaf .figleaf.functional
figleaf2html --exclude-patterns=../figleaf-exclude .figleaf.functional .figleaf.unittests
Also, this is very slow; on a decent machine, 10 minutes for the functional tests is normal.

To run without figleaf, be sure to unset `FIGLEAF`.


 1. trac-admin is failing on initenv with this exception:
    raise Exception('Failed with exitcode %s running trac-admin with %r' % (retval, args))
Exception: Failed with exitcode 1 running trac-admin with ('initenv', 'testenv', 'sqlite:db/trac.db', 'svn', '..../testenv/repo')
    This can be caused by not having run `python setup.py egg_info` in the genshi tree.

 2. Windows needs an implementation of crypt or fcrypt.  Carey Evans' pure
    python version works, but prints warnings on Python 2.3 (they can be
    ignored).  See http://carey.geek.nz/code/python-fcrypt/

 3. Python versions compatibility notes
   * If using Python >=2.4.0, <2.4.4, you need to backport `unittest` from
     2.4.4.  If you do not do this, you will get an `AttributeError` regarding
     `_fixture` being unset on every test.
   * If using Python >=2.3.0, <2.3.5, you need to backport a few modules for
     `setuptools` to work properly (`httplib2`, `cookielib`, `_*CookieJar`) and
     remove the tuple imports to replace with explicit line continuations.
   * If using Python >=2.3.0, <2.4.0, you need `subprocess`, `unittest`, and
     `traceback` from 2.4.4.  On Windows you also need to modify `subprocess.py`
     to use `pywin32` instead of `_subprocess`.  Twill includes a `subprocess`
     module that will not work in this situation because it is not modified.