Will backport to release23-maint.

* Relaxed the argument restrictions for non-operator methods.  They now
  allow any iterable instead of requiring a set.  This makes the module
  a little easier to use and paves the way for an efficient C
  implementation which can take better advantage of iterable arguments
  while screening out immutables.

* Deprecated Set.update() because it now duplicates Set.union_update()

* Adapted the tests and docs to include the above changes.

* Added more test coverage including testing identities and checking
  to make sure non-restartable generators work as arguments.

Will backport to Py2.3.1 so that the interface remains consistent
across versions.  The deprecation of update() will be changed to
a FutureWarning.

throughout the documentation.

* Positive wording for the description of why < and > and = can all
  be False.

* Move to a three column table format that puts long method names
  side-by-side with their operator equivalents

* Mention that KeyError can be raised by Set.pop() and Set.remove().

* Minor tweaks to the examples.

Will backport as soon as Fred rebuilds the docs so I can confirm
the tables formatted properly

If opening a directory, the exception would leak.

unchanged (and sys.gettotalrefcount() remains constant).

Fix a few typos.

[ 784825 ] fix obscure crash in descriptor handling

Should be applied to release23-maint and in all likelyhood
release22-maint, too.

Certainly doesn't apply to release21-maint.

Fix, by not using n at all in that case.

Needs to be applied to release23-maint, too.

number.  This accounts for the 2 refcount leaks per test_complex run
Michael Hudson discovered (I figured only I would have the stomach to
look for leaks in floating-point code <wink>).

usually preferred.

When parsing the constructor arguments failed, a
reference to the argument tuple was leaked.

when an encoding error occurs and the callback name is unknown,
i.e. when the callback has to be called. The problem was that
the fact that the callback has already been looked up was only
recorded in a local variable in charmap_encoding_error(), because
charmap_encoding_error() got it's own copy of the errorHandler
pointer instead of a pointer to the pointer in
PyUnicode_EncodeCharmap().

(Contributed by George Yoshida.)

Check for a[:] = a _before_ calling PySequence_Fast on a.
release23-maint candidate
Reference leak doesn't happen with head of release22-maint.

the output was consumed would cause and exception to be raise in
logmerge; suppress this specific error, because it's not helpful.

on W2K and XP.  Python Bugs 780451, 784183

Backported to 2.2-maint

UnicodeTranslateError message.

If there is only one bad character it will now be printed in a
form that is a valid Python string.

Now test_descr only appears to leak two references & I think this
are in fact illusory (it's to do with things getting resurrected in
__del__ methods & it's easy to be believe confusion occurs when that
happens <wink>).  Woohoo!

docs

the tp_name is not correct, but what's exposed to users is known visibly as
"StringIO", not "StringI" or "StringO".

This should go onto release23-maint, too.