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

........
  r69324 | tarek.ziade | 2009-02-06 01:31:59 +0100 (Fri, 06 Feb 2009) | 1 line

  Fixed #1276768: verbose option was not used in the code.
........

implicit hooks are handled properly.

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

........
  r69316 | tarek.ziade | 2009-02-05 23:52:52 +0100 (Thu, 05 Feb 2009) | 1 line

  Fixed #5132: enable extensions to link on Solaris
........

importlib.machinery.PathFinder.

source directory.  Mainly, use 'srcdir' rather than os.getcwd() or
'.'.

directory other than the source directory, simplify the test code in
test_sysconfig.py.

the source.  Also, define 'srcdir' on non-posix platforms.

........
  r69295 | georg.brandl | 2009-02-05 12:23:47 +0100 (Do, 05 Feb 2009) | 1 line

  PyErr_PrintEx is also in 2.x...
........

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

........
  r69285 | tarek.ziade | 2009-02-05 10:06:23 +0100 (Thu, 05 Feb 2009) | 1 line

  Fix comment for #1835
........

importlib.machinery. Also make the methods either class or staticmethods.

........
  r69276 | raymond.hettinger | 2009-02-04 13:25:17 -0600 (Wed, 04 Feb 2009) | 1 line

  Clarify that named tuples do not have to subclass tuple.
........

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

........
  r69260 | thomas.heller | 2009-02-03 18:07:40 +0100 (Di, 03 Feb 2009) | 9 lines

  This refactoring should make it easier to add new calling conventions.

  Replace ffi_call_STDCALL and ffi_call_SYSV by a ffi_call_x86 function
  that cleans up the stack when FFI_SYSV is used, and does nothing for
  FFI_STDCALL.

  Remove libffi_msvc\win32.S, which is out of date and also unused; it
  was only used for building ctypes with the MingW compiler.
........

r69209 caused the fp of HTTPResponse objects to be a io.BufferedReader instead of a socket.SocketIO.  This moves the underlying socket.socket object from r.fp._sock to r.fp.raw._sock.  Since _sock is an internal object, this should be ok. The change is for the testsuite only, to assert socket state.

........
  r69252 | brett.cannon | 2009-02-02 20:58:29 -0800 (Mon, 02 Feb 2009) | 3 lines

  Make importlib a package. This allows using svn:externals in the sandbox to
  package up the code for separate distribution.
........

........
  r69242 | raymond.hettinger | 2009-02-02 21:37:03 -0600 (Mon, 02 Feb 2009) | 1 line

  Register decimals as numbers.Number
........

........
  r69237 | raymond.hettinger | 2009-02-02 20:23:19 -0600 (Mon, 02 Feb 2009) | 1 line

  Validate that __length_hint__ returns a usable result.
........

........
  r68532 | kristjan.jonsson | 2009-01-11 10:23:37 -0600 (Sun, 11 Jan 2009) | 1 line

  Issue 4879: Allow buffering for HTTPResponse
........
  r69161 | gregory.p.smith | 2009-01-31 18:24:21 -0600 (Sat, 31 Jan 2009) | 2 lines

  wording for for issue4903.
........
  r69227 | raymond.hettinger | 2009-02-02 15:50:13 -0600 (Mon, 02 Feb 2009) | 1 line

  Issue 1242657: list(obj) can swallow KeyboardInterrupt.
........

to renaming of tp_compare.

svn+ssh://pythondev/python/trunk

........
  r69217 | guilherme.polo | 2009-02-02 19:08:32 -0200 (Mon, 02 Feb 2009) | 1 line

  Fix for issue #1581476
........
  r69219 | guilherme.polo | 2009-02-02 19:14:14 -0200 (Mon, 02 Feb 2009) | 1 line

  NEWS entry for issue #1581476
........

of cmpfunc.

type of tp_compare in a separate commit, for ease of reversion
should things go wrong.

Add some notes for importlib to package up what is in Python 2.7 as well as modules that can probably get replaced by importlib at some point.

an ``svn revert`` (gee, maybe more atomic commits by use of something like a
DVCS would help with this? Wish someone would get on that ...), I am checking
in an untested, work-in-progress meta path finder to handle sys.path.

Enable buffering for HTTPResponse's fp.  read() behaves identically for buffered and non-buffered IO.  read(n) also won't block if n bytes are availble on the socket.  There is therefore no reason not to use buffering.  The reason 2.x disables buffering by default, that some clients may be accessing the underlying socket directly and so bypass the buffering buffer, doesn't apply in 3.x with its redesigned IO library.
See issue 4448 and issue 4879