Metadata-Version: 2.1
Name: zope.testing
Version: 4.1.0
Summary: Zope testing helpers
Home-page: http://pypi.python.org/pypi/zope.testing
Author: Zope Foundation and Contributors
Author-email: zope-dev@zope.org
License: ZPL 2.1
Platform: UNKNOWN
Classifier: Development Status :: 5 - Production/Stable
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: Zope Public License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 2.4
Classifier: Programming Language :: Python :: 2.5
Classifier: Programming Language :: Python :: 2.6
Classifier: Programming Language :: Python :: 2.7
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.1
Classifier: Programming Language :: Python :: 3.2
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Classifier: Topic :: Software Development :: Testing
License-File: LICENSE.txt

************
zope.testing
************

.. contents::

This package provides a number of testing frameworks.

cleanup
  Provides a mixin class for cleaning up after tests that
  make global changes.

formparser
  An HTML parser that extracts form information.

  **Python 2 only**

  This is intended to support functional tests that need to extract
  information from HTML forms returned by the publisher.

  See formparser.txt.

loggingsupport
  Support for testing logging code

  If you want to test that your code generates proper log output, you
  can create and install a handler that collects output.

loghandler
  Logging handler for tests that check logging output.

module
  Lets a doctest pretend to be a Python module.

  See module.txt.

renormalizing
  Regular expression pattern normalizing output checker.
  Useful for doctests.

server
  Provides a simple HTTP server compatible with the zope.app.testing
  functional testing API.  Lets you interactively play with the system
  under test.  Helpful in debugging functional doctest failures.

  **Python 2 only**

setupstack
  A simple framework for automating doctest set-up and tear-down.
  See setupstack.txt.

wait
  A small utility for dealing with timing non-determinism
  See wait.txt.

Getting started
***************

zope.testing uses buildout.  To start, run ``python bootstrap.py``.  It will
create a number of directories and the ``bin/buildout`` script.  Next, run
``bin/buildout``.  It will create a test script for you.  Now, run ``bin/test``
to run the zope.testing test suite.

Detailed Documentation
**********************

Parsing HTML Forms
==================

Sometimes in functional tests, information from a generated form must
be extracted in order to re-submit it as part of a subsequent request.
The `zope.testing.formparser` module can be used for this purpose.

NOTE
   formparser doesn't support Python 3.

The scanner is implemented using the `FormParser` class.  The
constructor arguments are the page data containing the form and
(optionally) the URL from which the page was retrieved:

  >>> import zope.testing.formparser

  >>> page_text = '''\
  ... <html><body>
  ...   <form name="form1" action="/cgi-bin/foobar.py" method="POST">
  ...     <input type="hidden" name="f1" value="today" />
  ...     <input type="submit" name="do-it-now" value="Go for it!" />
  ...     <input type="IMAGE" name="not-really" value="Don't."
  ...            src="dont.png" />
  ...     <select name="pick-two" size="3" multiple>
  ...       <option value="one" selected>First</option>
  ...       <option value="two" label="Second">Another</option>
  ...       <optgroup>
  ...         <option value="three">Third</option>
  ...         <option selected="selected">Fourth</option>
  ...       </optgroup>
  ...     </select>
  ...   </form>
  ...
  ...   Just for fun, a second form, after specifying a base:
  ...   <base href="http://www.example.com/base/" />
  ...   <form action = 'sproing/sprung.html' enctype="multipart/form">
  ...     <textarea name="sometext" rows="5">Some text.</textarea>
  ...     <input type="Image" name="action" value="Do something."
  ...            src="else.png" />
  ...     <input type="text" value="" name="multi" size="2" />
  ...     <input type="text" value="" name="multi" size="3" />
  ...   </form>
  ... </body></html>
  ... '''

  >>> parser = zope.testing.formparser.FormParser(page_text)
  >>> forms = parser.parse()

  >>> len(forms)
  2
  >>> forms.form1 is forms[0]
  True
  >>> forms.form1 is forms[1]
  False

More often, the `parse()` convenience function is all that's needed:

  >>> forms = zope.testing.formparser.parse(
  ...     page_text, "http://cgi.example.com/somewhere/form.html")

  >>> len(forms)
  2
  >>> forms.form1 is forms[0]
  True
  >>> forms.form1 is forms[1]
  False

Once we have the form we're interested in, we can check form
attributes and individual field values:

  >>> form = forms.form1
  >>> form.enctype
  'application/x-www-form-urlencoded'
  >>> form.method
  'post'

  >>> keys = form.keys()
  >>> keys.sort()
  >>> keys
  ['do-it-now', 'f1', 'not-really', 'pick-two']

  >>> not_really = form["not-really"]
  >>> not_really.type
  'image'
  >>> not_really.value
  "Don't."
  >>> not_really.readonly
  False
  >>> not_really.disabled
  False

Note that relative URLs are converted to absolute URLs based on the
``<base>`` element (if present) or using the base passed in to the
constructor.

  >>> form.action
  'http://cgi.example.com/cgi-bin/foobar.py'
  >>> not_really.src
  'http://cgi.example.com/somewhere/dont.png'

  >>> forms[1].action
  'http://www.example.com/base/sproing/sprung.html'
  >>> forms[1]["action"].src
  'http://www.example.com/base/else.png'

Fields which are repeated are reported as lists of objects that
represent each instance of the field::

  >>> field = forms[1]["multi"]
  >>> isinstance(field, list)
  True
  >>> [o.value for o in field]
  ['', '']
  >>> [o.size for o in field]
  [2, 3]

The ``<textarea>`` element provides some additional attributes:

  >>> ta = forms[1]["sometext"]
  >>> print ta.rows
  5
  >>> print ta.cols
  None
  >>> ta.value
  'Some text.'

The ``<select>`` element provides access to the options as well:

  >>> select = form["pick-two"]
  >>> select.multiple
  True
  >>> select.size
  3
  >>> select.type
  'select'
  >>> select.value
  ['one', 'Fourth']

  >>> options = select.options
  >>> len(options)
  4
  >>> [opt.label for opt in options]
  ['First', 'Second', 'Third', 'Fourth']
  >>> [opt.value for opt in options]
  ['one', 'two', 'three', 'Fourth']

Support for testing logging code
================================

If you want to test that your code generates proper log output, you
can create and install a handler that collects output:

  >>> from zope.testing.loggingsupport import InstalledHandler
  >>> handler = InstalledHandler('foo.bar')

The handler is installed into loggers for all of the names passed. In
addition, the logger level is set to 1, which means, log
everything. If you want to log less than everything, you can provide a
level keyword argument.  The level setting effects only the named
loggers.

  >>> import logging
  >>> handler_with_levels = InstalledHandler('baz', level=logging.WARNING)

Then, any log output is collected in the handler:

  >>> logging.getLogger('foo.bar').exception('eek')
  >>> logging.getLogger('foo.bar').info('blah blah')

  >>> for record in handler.records:
  ...     print_(record.name, record.levelname)
  ...     print_(' ', record.getMessage())
  foo.bar ERROR
    eek
  foo.bar INFO
    blah blah

A similar effect can be gotten by just printing the handler:

  >>> print_(handler)
  foo.bar ERROR
    eek
  foo.bar INFO
    blah blah

After checking the log output, you need to uninstall the handler:

  >>> handler.uninstall()
  >>> handler_with_levels.uninstall()

At which point, the handler won't get any more log output.
Let's clear the handler:

  >>> handler.clear()
  >>> handler.records
  []

And then log something:

  >>> logging.getLogger('foo.bar').info('blah')

and, sure enough, we still have no output:

  >>> handler.records
  []

Regular expression pattern normalizing output checker
=====================================================

The pattern-normalizing output checker extends the default output checker with
an option to normalize expected and actual output.

You specify a sequence of patterns and replacements.  The replacements are
applied to the expected and actual outputs before calling the default outputs
checker.  Let's look at an example.  In this example, we have some times and
addresses:

    >>> want = '''\
    ... <object object at 0xb7f14438>
    ... completed in 1.234 seconds.
    ... <BLANKLINE>
    ... <object object at 0xb7f14440>
    ... completed in 123.234 seconds.
    ... <BLANKLINE>
    ... <object object at 0xb7f14448>
    ... completed in .234 seconds.
    ... <BLANKLINE>
    ... <object object at 0xb7f14450>
    ... completed in 1.234 seconds.
    ... <BLANKLINE>
    ... '''

    >>> got = '''\
    ... <object object at 0xb7f14458>
    ... completed in 1.235 seconds.
    ...
    ... <object object at 0xb7f14460>
    ... completed in 123.233 seconds.
    ...
    ... <object object at 0xb7f14468>
    ... completed in .231 seconds.
    ...
    ... <object object at 0xb7f14470>
    ... completed in 1.23 seconds.
    ...
    ... '''

We may wish to consider these two strings to match, even though they differ in
actual addresses and times.  The default output checker will consider them
different:

    >>> import doctest
    >>> doctest.OutputChecker().check_output(want, got, 0)
    False

We'll use the zope.testing.renormalizing.OutputChecker to normalize both the
wanted and gotten strings to ignore differences in times and
addresses:

    >>> import re
    >>> from zope.testing.renormalizing import OutputChecker
    >>> checker = OutputChecker([
    ...    (re.compile('[0-9]*[.][0-9]* seconds'), '<SOME NUMBER OF> seconds'),
    ...    (re.compile('at 0x[0-9a-f]+'), 'at <SOME ADDRESS>'),
    ...    ])

    >>> checker.check_output(want, got, 0)
    True

Usual OutputChecker options work as expected:

    >>> want_ellided = '''\
    ... <object object at 0xb7f14438>
    ... completed in 1.234 seconds.
    ... ...
    ... <object object at 0xb7f14450>
    ... completed in 1.234 seconds.
    ... <BLANKLINE>
    ... '''

    >>> checker.check_output(want_ellided, got, 0)
    False

    >>> checker.check_output(want_ellided, got, doctest.ELLIPSIS)
    True

When we get differencs, we output them with normalized text:

    >>> source = '''\
    ... >>> do_something()
    ... <object object at 0xb7f14438>
    ... completed in 1.234 seconds.
    ... ...
    ... <object object at 0xb7f14450>
    ... completed in 1.234 seconds.
    ... <BLANKLINE>
    ... '''

    >>> example = doctest.Example(source, want_ellided)

    >>> print_(checker.output_difference(example, got, 0))
    Expected:
        <object object at <SOME ADDRESS>>
        completed in <SOME NUMBER OF> seconds.
        ...
        <object object at <SOME ADDRESS>>
        completed in <SOME NUMBER OF> seconds.
        <BLANKLINE>
    Got:
        <object object at <SOME ADDRESS>>
        completed in <SOME NUMBER OF> seconds.
        <BLANKLINE>
        <object object at <SOME ADDRESS>>
        completed in <SOME NUMBER OF> seconds.
        <BLANKLINE>
        <object object at <SOME ADDRESS>>
        completed in <SOME NUMBER OF> seconds.
        <BLANKLINE>
        <object object at <SOME ADDRESS>>
        completed in <SOME NUMBER OF> seconds.
        <BLANKLINE>
    <BLANKLINE>

    >>> print_(checker.output_difference(example, got,
    ...                                 doctest.REPORT_NDIFF))
    Differences (ndiff with -expected +actual):
        - <object object at <SOME ADDRESS>>
        - completed in <SOME NUMBER OF> seconds.
        - ...
          <object object at <SOME ADDRESS>>
          completed in <SOME NUMBER OF> seconds.
          <BLANKLINE>
        + <object object at <SOME ADDRESS>>
        + completed in <SOME NUMBER OF> seconds.
        + <BLANKLINE>
        + <object object at <SOME ADDRESS>>
        + completed in <SOME NUMBER OF> seconds.
        + <BLANKLINE>
        + <object object at <SOME ADDRESS>>
        + completed in <SOME NUMBER OF> seconds.
        + <BLANKLINE>
    <BLANKLINE>

    If the wanted text is empty, however, we don't transform the actual output.
    This is usful when writing tests.  We leave the expected output empty, run
    the test, and use the actual output as expected, after reviewing it.

    >>> source = '''\
    ... >>> do_something()
    ... '''

    >>> example = doctest.Example(source, '\n')
    >>> print_(checker.output_difference(example, got, 0))
    Expected:
    <BLANKLINE>
    Got:
        <object object at 0xb7f14458>
        completed in 1.235 seconds.
        <BLANKLINE>
        <object object at 0xb7f14460>
        completed in 123.233 seconds.
        <BLANKLINE>
        <object object at 0xb7f14468>
        completed in .231 seconds.
        <BLANKLINE>
        <object object at 0xb7f14470>
        completed in 1.23 seconds.
        <BLANKLINE>
    <BLANKLINE>

If regular expressions aren't expressive enough, you can use arbitrary Python
callables to transform the text.  For example, suppose you want to ignore
case during comparison:

    >>> checker = OutputChecker([
    ...    lambda s: s.lower(),
    ...    lambda s: s.replace('<blankline>', '<BLANKLINE>'),
    ...    ])

    >>> want = '''\
    ... Usage: thundermonkey [options] [url]
    ... <BLANKLINE>
    ... Options:
    ...     -h    display this help message
    ... '''

    >>> got = '''\
    ... usage: thundermonkey [options] [URL]
    ...
    ... options:
    ...     -h    Display this help message
    ... '''

    >>> checker.check_output(want, got, 0)
    True

Suppose we forgot that <BLANKLINE> must be in upper case:

    >>> checker = OutputChecker([
    ...    lambda s: s.lower(),
    ...    ])

    >>> checker.check_output(want, got, 0)
    False

The difference would show us that:

    >>> source = '''\
    ... >>> print_help_message()
    ... ''' + want
    >>> example = doctest.Example(source, want)
    >>> print_(checker.output_difference(example, got,
    ...                                 doctest.REPORT_NDIFF))
    Differences (ndiff with -expected +actual):
          usage: thundermonkey [options] [url]
        - <blankline>
        + <BLANKLINE>
          options:
              -h    display this help message
    <BLANKLINE>


It is possible to combine OutputChecker checkers for easy reuse:

    >>> address_and_time_checker = OutputChecker([
    ...    (re.compile('[0-9]*[.][0-9]* seconds'), '<SOME NUMBER OF> seconds'),
    ...    (re.compile('at 0x[0-9a-f]+'), 'at <SOME ADDRESS>'),
    ...    ])
    >>> lowercase_checker = OutputChecker([
    ...    lambda s: s.lower(),
    ...    ])
    >>> combined_checker = address_and_time_checker + lowercase_checker
    >>> len(combined_checker.transformers)
    3

Combining a checker with something else does not work:

    >>> lowercase_checker + 5 #doctest: +ELLIPSIS
    Traceback (most recent call last):
        ...
    TypeError: unsupported operand type(s) for +: ...


Stack-based test setUp and tearDown
===================================

Writing doctest setUp and tearDown functions can be a bit tedious,
especially when setUp/tearDown functions are combined.

the zope.testing.setupstack module provides a small framework for
automating test tear down.  It provides a generic setUp function that
sets up a stack. Normal test setUp functions call this function to set
up the stack and then use the register function to register tear-down
functions.

To see how this works we'll create a faux test:

    >>> class Test:
    ...     def __init__(self):
    ...         self.globs = {}
    >>> test = Test()

We'll register some tearDown functions that just print something:

    >>> import sys
    >>> import zope.testing.setupstack
    >>> zope.testing.setupstack.register(
    ...     test, lambda : sys.stdout.write('td 1\n'))
    >>> zope.testing.setupstack.register(
    ...     test, lambda : sys.stdout.write('td 2\n'))

Now, when we call the tearDown function:

    >>> zope.testing.setupstack.tearDown(test)
    td 2
    td 1

The registered tearDown functions are run. Note that they are run in
the reverse order that they were registered.


Extra positional arguments can be passed to register:

    >>> zope.testing.setupstack.register(
    ...    test, lambda x, y, z: sys.stdout.write('%s %s %s\n' % (x, y, z)),
    ...    1, 2, z=9)
    >>> zope.testing.setupstack.tearDown(test)
    1 2 9


Temporary Test Directory
------------------------

Often, tests create files as they demonstrate functionality.  They
need to arrange for the removeal of these files when the test is
cleaned up.

The setUpDirectory function automates this.  We'll get the current
directory first:

    >>> import os
    >>> here = os.getcwd()

We'll also create a new test:

    >>> test = Test()

Now we'll call the setUpDirectory function:

    >>> zope.testing.setupstack.setUpDirectory(test)

We don't have to call zope.testing.setupstack.setUp, because
setUpDirectory calls it for us.

Now the current working directory has changed:

    >>> here == os.getcwd()
    False
    >>> setupstack_cwd = os.getcwd()

We can create files to out heart's content:

    >>> foo = open('Data.fs', 'w').write('xxx')
    >>> os.path.exists(os.path.join(setupstack_cwd, 'Data.fs'))
    True

We'll make the file read-only. This can cause problems on Windows, but
setupstack takes care of that by making files writable before trying
to remove them.

    >>> import stat
    >>> os.chmod('Data.fs', stat.S_IREAD)

On Unix systems, broken symlinks can cause problems because the chmod
attempt by the teardown hook will fail; let's set up a broken symlink as
well, and verify the teardown doesn't break because of that:

    >>> if hasattr(os, 'symlink'):
    ...     os.symlink('NotThere', 'BrokenLink')

When tearDown is called:

    >>> zope.testing.setupstack.tearDown(test)

We'll be back where we started:

    >>> here == os.getcwd()
    True

and the files we created will be gone (along with the temporary
directory that was created:

    >>> os.path.exists(os.path.join(setupstack_cwd, 'Data.fs'))
    False

Context-manager support
-----------------------

You can leverage context managers using the ``contextmanager`` method.
The result of calling the content manager's __enter__ method will be
returned. The context-manager's __exit__ method will be called as part
of test tear down:

    >>> class Manager(object):
    ...     def __enter__(self):
    ...         print_('enter')
    ...         return 42
    ...     def __exit__(self, *args):
    ...         print_('exit', args)

    >>> manager = Manager()
    >>> test = Test()

    >>> zope.testing.setupstack.context_manager(test, manager)
    enter
    42

    >>> zope.testing.setupstack.tearDown(test)
    exit (None, None, None)

globs
-----

Doctests have ``globs`` attributes used to hold test globals.
``setupstack`` was originally designed to work with doctests, but can
now work with either doctests, or other test objects, as long as the
test objects have either a ``globs`` attribute or a ``__dict__``
attribute.  The ``zope.testing.setupstack.globs`` function is used to
get the globals for a test object:

    >>> zope.testing.setupstack.globs(test) is test.globs
    True

Here, because the test object had a ``globs`` attribute, it was
returned. Because we used the test object above, it has a setupstack:

    >>> '__zope.testing.setupstack' in test.globs
    True

If we remove the ``globs`` attribute, the object's instance dictionary
will be used:

    >>> del test.globs
    >>> zope.testing.setupstack.globs(test) is test.__dict__
    True
    >>> zope.testing.setupstack.context_manager(test, manager)
    enter
    42

    >>> '__zope.testing.setupstack' in test.__dict__
    True

The ``globs`` function is used internally, but can also be used by
setup code to support either doctests or other test objects.

Wait until a condition holds (or until a time out)
==================================================

Often, in tests, you need to wait until some condition holds.  This
may be because you're testing interaction with an external system or
testing threaded (threads, processes, greenlet's, etc.) interactions.

You can add sleeps to your tests, but it's often hard to know how
long to sleep.

``zope.testing.wait`` provides a convenient way to wait until
some condition holds.  It will test a condition and, when true,
return.  It will sleep a short time between tests.

Here's a silly example, that illustrates it's use:

    >>> from zope.testing.wait import wait
    >>> wait(lambda : True)

Since the condition we passed is always True, it returned
immediately.  If the condition doesn't hold, then we'll get a timeout:

    >>> wait((lambda : False), timeout=.01)
    Traceback (most recent call last):
    ...
    TimeOutWaitingFor: <lambda>

``wait`` has some keyword options:

timeout
   How long, in seconds, to wait for the condition to hold

   Defaults to 9 seconds.

wait
   How long to wait between calls.

   Defaults to .01 seconds.

message
   A message (or other data) to pass to the timeout exception.

   This defaults to ``None``.  If this is false, then the callable's
   doc string or ``__name__`` is used.

``wait`` can be used as a decorator:

    >>> @wait
    ... def ok():
    ...     return True

    >>> @wait(timeout=.01)
    ... def no_way():
    ...     pass
    Traceback (most recent call last):
    ...
    TimeOutWaitingFor: no_way

    >>> @wait(timeout=.01)
    ... def no_way():
    ...     "never true"
    Traceback (most recent call last):
    ...
    TimeOutWaitingFor: never true

.. more tests

    >>> import time
    >>> now = time.time()
    >>> @wait(timeout=.01, message='dang')
    ... def no_way():
    ...     "never true"
    Traceback (most recent call last):
    ...
    TimeOutWaitingFor: dang

    >>> .01 < (time.time() - now) < .03
    True


Customization
-------------

``wait`` is an instance of ``Wait``.  With ``Wait``,
you can create you're own custom ``wait`` utilities.  For
example, if you're testing something that uses getevent, you'd want to
use gevent's sleep function:

    >>> import zope.testing.wait
    >>> wait = zope.testing.wait.Wait(getsleep=lambda : gevent.sleep)

Wait takes a number of customization parameters:

exception
  Timeout exception class

getnow
  Function used to get a function for getting the current time.

  Default: lambda : time.time

getsleep
  Function used to get a sleep function.

  Default: lambda : time.sleep

timeout
  Default timeout

  Default: 9

wait
  Default time to wait between attempts

  Default: .01


.. more tests

    >>> def mysleep(t):
    ...     print_('mysleep', t)
    ...     time.sleep(t)

    >>> def mynow():
    ...     print_('mynow')
    ...     return time.time()

    >>> wait = zope.testing.wait.Wait(
    ...    getnow=(lambda : mynow), getsleep=(lambda : mysleep),
    ...    exception=ValueError, timeout=.02, wait=.0001)

    >>> @wait
    ... def _(state=[]):
    ...     if len(state) > 1:
    ...        return True
    ...     state.append(0)
    mynow
    mysleep 0.0001
    mynow
    mysleep 0.0001

    >>> @wait(wait=.002)
    ... def _(state=[]):
    ...     if len(state) > 1:
    ...        return True
    ...     state.append(0)
    mynow
    mysleep 0.002
    mynow
    mysleep 0.002

    >>> @wait(timeout=0)
    ... def _(state=[]):
    ...     if len(state) > 1:
    ...        return True
    ...     state.append(0)
    Traceback (most recent call last):
    ...
    ValueError: _

    >>> wait = zope.testing.wait.Wait(timeout=0)
    >>> @wait(timeout=0)
    ... def _(state=[]):
    ...     if len(state) > 1:
    ...        return True
    ...     state.append(0)
    Traceback (most recent call last):
    ...
    TimeOutWaitingFor: _

zope.testing Changelog
**********************

4.1.0 (2012-01-29)
==================

- Added context-manager support to ``zope.testing.setupstack``

- Made ``zope.testing.setupstack`` usable with all tests, not just
  doctests and added ``zope.testing.setupstack.globs``, which makes it
  easier to write test setup code that workes with doctests and other
  kinds of tests.

- Added the ``wait`` module, which makes it easier to deal with
  non-deterministic timing issues.

- Renamed ``zope.testing.renormalizing.RENormalizing`` to
  ``zope.testing.renormalizing.OutputChecker``. The old name is an
  alias.

- Updated tests to run with Python 3.

- More clearly labeled which features were supported by Python 3.

- Reorganized documentation.

4.0.0 (2011-11-09)
==================

- Removes the deprecated zope.testing.doctest.

- Adds Python 3 support.

- Fixed test which fails if there is a file named `Data.fs` in the current
  working directory.


3.10.2 (2010-11-30)
===================

- Fix test of broken symlink handling to not break on Windows.


3.10.1 (2010-11-29)
===================

- Fix removal of broken symlinks on Unix.


3.10.0 (2010-07-21)
===================

- Removed zope.testing.testrunner, which now is moved to zope.testrunner.

- Update fix for LP #221151 to a spelling compatible with Python 2.4.

3.9.5 (2010-05-19)
==================

- LP #579019: When layers were run in parallel, their tearDown was not
  called. Additionally, the first layer which was run in the main
  thread did not have it's tearDown called either.

- Deprecated zope.testing.testrunner and zope.testing.exceptions. They have
  been moved to a separate zope.testrunner module, and will be removed from
  zope.testing in 4.0.0, together with zope.testing.doctest.

3.9.4 (2010-04-13)
==================

- LP #560259: Fix subunit output formatter to handle layer setup
  errors.

- LP #399394:  Added a ``--stop-on-error`` / ``--stop`` / ``-x`` option to
  the testrunner.

- LP #498162:  Added a ``--pdb`` alias for the existing ``--post-mortem``
  / ``-D`` option to the testrunner.

- LP #547023:  Added a ``--version`` option to the testrunner.

- Added tests for LP #144569 and #69988.

  https://bugs.launchpad.net/bugs/69988

  https://bugs.launchpad.net/zope3/+bug/144569


3.9.3 (2010-03-26)
==================

- zope.testing.renormalizer no longer imports zope.testing.doctest, which
  caused deprecation warnings.

- Fix testrunner-layers-ntd.txt to suppress output to sys.stderr.

- Suppress zope.testing.doctest deprecation warning when running
  zope.testing's own test suite.


3.9.2 (2010-03-15)
==================

- Fixed broken ``from zope.testing.doctest import *``

3.9.1 (2010-03-15)
==================

- No changes; reuploaded to fix broken 3.9.0 release on PyPI.

3.9.0 (2010-03-12)
==================

- Modified the testrunner to use the standard Python doctest module instead of
  the deprecated zope.testing.doctest.

- Fix testrunner-leaks.txt to use the run_internal helper, so that
  sys.exit() isn't triggered during the test run.

- Added support for conditionally using a subunit-based output
  formatter upon request if subunit and testtools are available. Patch
  contributed by Jonathan Lange.

3.8.7 (2010-01-26)
==================

- Downgraded the zope.testing.doctest deprecation warning into a
  PendingDeprecationWarning.

3.8.6 (2009-12-23)
==================

- Added MANIFEST.in and reuploaded to fix broken 3.8.5 release on PyPI.


3.8.5 (2009-12-23)
==================

- Added DocFileSuite, DocTestSuite, debug_src and debug back BBB imports
  back into zope.testing.doctestunit; apparently many packages still import
  them from there!

- Made zope.testing.doctest and zope.testing.doctestunit emit deprecation
  warnings: use the stdlib doctest instead.


3.8.4 (2009-12-18)
==================

- Fixed missing imports and undefined variables reported by pyflakes,
  adding tests to exercise the blind spots.

- Cleaned up unused imports reported by pyflakes.

- Added two new options to generate randomly ordered list of tests and to
  select a specific order of tests.

- RENormalizing checkers can be combined via ``+`` now:
  ``checker1 + checker2`` creates a checker with the transformations of both
  checkers.

- Test fixes for Python 2.7.

3.8.3 (2009-09-21)
==================

- Avoid a split() call or we get test failures when running from a directory
  with spaces in it.

- Fix testrunner behavior on Windows for -j2 (or greater) combined with -v
  (or greater).

3.8.2 (2009-09-15)
==================

- Removing hotshot profiler when using Python 2.6. That makes zope.testing
  compatible with Python 2.6


3.8.1 (2009-08-12)
==================

- Avoid hardcoding sys.argv[0] as script;
  allow, for instance, Zope 2's `bin/instance test` (LP#407916).

- Produce a clear error message when a subprocess doesn't follow the
  zope.testing.testrunner protocol (LP#407916).

- Do not unnecessarily squelch verbose output in a subprocess when there are
  not multiple subprocesses.

- Do not unnecessarily batch subprocess output, which can stymie automated and
  human processes for identifying hung tests.

- Include incremental output when there are multiple subprocesses and a
  verbosity of -vv or greater is requested.  This again is not batched,
  supporting automated processes and humans looking for hung tests.


3.8.0 (2009-07-24)
==================

- Testrunner automatically picks up descendants of unittest.TestCase in test
  modules, so you don't have to provide a test_suite() anymore.


3.7.7 (2009-07-15)
==================

- Clean up support for displaying tracebacks with supplements by turning it
  into an always-enabled feature and making the dependency on zope.exceptions
  explicit.

- Fix #251759: Test runner descended into directories that aren't Python
  packages.

- Code cleanups.


3.7.6 (2009-07-02)
==================

- Add zope-testrunner console_scripts entry point. This exposes a
  zope-testrunner binary with default installs allowing the testrunner to be
  run from the command line.

3.7.5 (2009-06-08)
==================

- Fix bug when running subprocesses on Windows.

- The option REPORT_ONLY_FIRST_FAILURE (command line option "-1") is now
  respected even when a doctest declares its own REPORTING_FLAGS, such as
  REPORT_NDIFF.

- Fixed bug that broke readline with pdb when using doctest
  (see http://bugs.python.org/issue5727).

- Made tests pass on Windows and Linux at the same time.


3.7.4 (2009-05-01)
==================

- Filenames of doctest examples now contain the line number and not
  only the example number. So a stack trace in pdb tells the exact
  line number of the current example. This fixes
  https://bugs.launchpad.net/bugs/339813

- Colorization of doctest output correctly handles blank lines.


3.7.3 (2009-04-22)
==================

- Better deal with rogue threads by always exiting with status so even
  spinning daemon threads won't block the runner from exiting. This deprecated
  the ``--with-exit-status`` option.


3.7.2 (2009-04-13)
==================

- fix test failure on Python 2.4 because of slight difference in the way
  coverage is reported (__init__ files with only a single comment line are now
  not reported)
- fixed bug that caused the test runner to hang when running subprocesses (as a
  result Python 2.3 is no longer supported).
- there is apparently a bug in Python 2.6 (related to
  http://bugs.python.org/issue1303673) that causes the profile tests to fail.
- added explanitory notes to buildout.cfg about how to run the tests with
  multiple versions of Python


3.7.1 (2008-10-17)
==================

- The setupstack temporary-directory support now properly handles
  read-only files by making them writable before removing them.


3.7.0 (2008-09-22)
==================

- Added an alterate setuptools / distutils commands for running all tests
  using our testrunner.  See 'zope.testing.testrunner.eggsupport:ftest'.

- Added a setuptools-compatible test loader which skips tests with layers:
  the testrunner used by 'setup.py test' doesn't know about them, and those
  tests then fail.  See 'zope.testing.testrunner.eggsupport:SkipLayers'.

- Added support for Jython, when a garbage collector call is sent.

- Added support to bootstrap on Jython.

- Fixed NameError in StartUpFailure.

- Open doctest files in universal mode, so that packages released on Windows
  can be tested on Linux, for example.


3.6.0 (2008/07/10)
==================

- Added -j option to parallel tests run in subprocesses.

- RENormalizer accepts plain Python callables.

- Added --slow-test option.

- Added --no-progress and --auto-progress options.

- Complete refactoring of the test runner into multiple code files and a more
  modular (pipeline-like) architecture.

- Unified unit tests with the layer support by introducing a real unit test
  layer.

- Added a doctest for ``zope.testing.module``. There were several bugs
  that were fixed:

  * ``README.txt`` was a really bad default argument for the module
    name, as it is not a proper dotted name. The code would
    immediately fail as it would look for the ``txt`` module in the
    ``README`` package. The default is now ``__main__``.

  * The tearDown function did not clean up the ``__name__`` entry in the
    global dictionary.

- Fix a bug that caused a SubprocessError to be generated if a subprocess
  sent any output to stderr.

- Fix a bug that caused the unit tests to be skipped if run in a subprocess.


3.5.1 (2007/08/14)
==================

Bugs Fixed:
-----------

- Post-mortem debugging wasn't invoked for layer-setup failures.

3.5.0 (2007/07/19)
==================

New Features
------------

- The test runner now works on Python 2.5.

- Added support for cProfile.

- Added output colorizing (-c option).

- Added --hide-secondary-failures and --show-secondary-failures options
  (https://bugs.launchpad.net/zope3/+bug/115454).

Bugs Fixed:
-----------

- Fix some problems with Unicode in doctests.

- Fix "Error reading from subprocess" errors on Unix-like systems.

3.4 (2007/03/29)
================

New Features
------------

- Added exit-with-status support (supports use with buildbot and
  zc.recipe.testing)

- Added a small framework for automating set up and tear down of
  doctest tests. See setupstack.txt.

Bugs Fixed:
-----------

- Fix testrunner-wo-source.txt and testrunner-errors.txt to run with a
  read-only source tree.

3.0 (2006/09/20)
================

- Updated the doctest copy with text-file encoding support.

- Added logging-level support to loggingsuppport module.

- At verbosity-level 1, dots are not output continuously, without any
  line breaks.

- Improved output when the inability to tear down a layer causes tests
  to be run in a subprocess.

- Made zope.exception required only if the zope_tracebacks extra is
  requested.

2.x.y (???)
===========

- Fix the test coverage. If a module, for example `interfaces`, was in an
  ignored directory/package, then if a module of the same name existed in a
  covered directory/package, then it was also ignored there, because the
  ignore cache stored the result by module name and not the filename of the
  module.

2.0 (2006/01/05)
================

- Corresponds to the version of the zope.testing package shipped as part of
  the Zope 3.2.0 release.


