value.  This affects TALCompiler.py and TALInterpreter.py, and
  affects the output of the non-local-macro expansion tests.  The
  TALVisitor.py module is *not* yet fixed to do this.

- In order to do this, I had to change the "program code" format for
  the attribute list: the 3rd item of an attribute description tuple,
  if present, now names an action, and the remaining tuple items are
  arguments for that action.  Two actions are recognized: "replace"
  replaces the attribute with the outcome of evaluating the expression
  given by the next tuple  item; "macroHack" replaces the attribute
  name with "use-macro" and its value with the current macro, but
  only if we're inside macro expansion and the attribute name ends in
  ":define-macro".

- The test suite is now error-free for "./runtest.py -c" but shows some
  errors for "./runtest.py".

- Added a primitive debugging flag to the TALInterpreter class.

- Update the README.

- Should I check for spurious TAL attributes even when doing just
  METAL (macro) expansion?

- Should I turn all the other errors into exceptions too?  (Currently
  they are all print statements.)

lock-refresh request.

to minidom.  (Note that at the moment, only the latest Python CVS
version of minidom works; 2.1b1 will work too when it is released, but
2.1a2 or anything before won't.)

- Lose redundant semicolon.

- Rearrange pickle imports.

built and a dictionary was being found where a list was expected (the
dictionary was in the form of {'rolename':1}, used to set the default
roles for a permission.  Changed the line to get the keys of the
dictionary).

Document running the regression test.

Add TO DO list.

attribute list.

constructor.

This uses Tim Peters's ndiff to produce nice diffs without invoking an
external program.

* Merge hardening of ZMI contents view against subobjects with
  flaky 'get_size' (Collector #1900).

It has more informative default output useful for testing things that take out the interpreter halfway through the test script (like segfaulting C extensions).  Instead of printing dots as tests are run, it prints the testname.  It also has other features, listed below.

Changes from 1.2.0 to 1.3.0
---------------------------

* Clearer and more verbose text output format
* Tests run in text mode can now be interrupted using ctrl-c
* New FunctionTestCase class provides support for wrapping legacy test
  functions into PyUnit test case instances
* Code is now compatible with JPython (new example: examples/withjpython.py)
* Support for short descriptions of tests, taken from __doc__ strings
  by default
* Updated and expanded documentation
* Tested with Python 2
* Changed module reloading mechanism in GUI test runner to fix a problem
  with Python 2 on Win32 reported by Henrik Weber (bug 125463)
* Convenient new unittest.main() function for use by all test modules

For more information, see http://pyunit.sourceforge.net

The output of both has the same form, and most be interpreted using
the TALInterpreter class.  However the METALCompiler output only
does (METAL) macro expansion and leaves everything else (i.e. TAL
statements) alone.