be answered with the comments removed.

There are many places that require checks when doing arithmetic for memory
sizes when allocating memory.  Otherwise, overflow is possible with
a subsequent crash.

Fix SF #1777057 which was a result of not initializing the new BufferError
properly.  Had to update the test for exceptions for BufferError too.

(testNewlinesInput and testNewlinesOutput are mine, not Tony's.)

and implements .newlines attribute in a 2.x-compatible fashion.

*restoring* selected registries instead of *resetting* them all.  Also
change the default for the huntrleaks parameter to None everywhere, as
its other value is a list of three elements, so it is not a bool.

methods.  Reset the negative cache *before* resetting the invalidation
counter, hoping this may plug a race condition (but then again, this
whole module isn't coded to be thread-safe).

docstring when there's a command line syntax error.

codec changes in trubk and branch).  Hopefully the Py3k glossary wasn't
different from the trunk one.

on big endian machines.

Update documentation: UTF-32 codecs will be in 2.6.

svn+ssh://pythondev@svn.python.org/python/trunk

........
  r57127 | neal.norwitz | 2007-08-16 21:10:55 -0700 (Thu, 16 Aug 2007) | 1 line

  Update the tools before building the docs
........

svn+ssh://pythondev@svn.python.org/python/trunk

........
  r57123 | guido.van.rossum | 2007-08-16 20:30:08 -0700 (Thu, 16 Aug 2007) | 2 lines

  Ignore *.py[co] files.
........
  r57124 | guido.van.rossum | 2007-08-16 20:36:03 -0700 (Thu, 16 Aug 2007) | 2 lines

  Ignore a few more files.
........

using the message generated by svnmerge, because it contains a lot of
stuff about the Doc tree, which I'm not merging this time due to the
way the Doc tree was initially added.  I am however adding roman.py
which was added later to Doc/tools.  I'll try to diff the two Doc
trees separately to see if there's stuff I missed.

svn+ssh://pythondev@svn.python.org/python/trunk

ut-32-be). On narrow builds the codecs combine surrogate pairs in the unicode
object into one codepoint on encoding and create surrogate pairs for
codepoints outside the BMP on decoding. Lone surrogates are passed through
unchanged in all cases.

Backport to the trunk will follow.

Refactor the urllib open_http() and open_https() to use the same
internal logic.  The only difference between the two is the call to
constructor an HTTPConnection object.