Updates to the msi builder and PCbuild9 directory. msi.py can now create installers from the PCbuild9 directory with MSVCR90.dll. PCbuild and MSVCR71.dll are still the default.

Initial import of new PCbuild9 for VS 2008. It partly based on PCbuild and partly hand crafted with some idea from PCbuild8. I've recreated all the extension module projects.

The new directory needs some more love and care but it works. I'm not able to test the AMD64 build.

The new tree is heavily using the *.vcprops property sheets. Please set any global settings in the property sheets.

to ssl.py (no need to test whether we can dup any more).
Regular sockets no longer have a _base, but we still have explicit
reference counting of socket objects for the benefit of makefile();
using duplicate sockets won't work for SSLSocket.

be sure to import a fresh module by removing it from sys.modules

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

........
  r58952 | christian.heimes | 2007-11-12 10:58:08 -0800 (Mon, 12 Nov 2007) | 6 lines

  readline module cleanup
  fixed indention to tabs
  use Py_RETURN_NONE macro
  added more error checks to on_completion_display_matches_hook

  open question: Does PyList_SetItem(l, i, o) steal a reference to o in the case of an error?
........
  r58956 | guido.van.rossum | 2007-11-12 12:06:40 -0800 (Mon, 12 Nov 2007) | 2 lines

  Add the test from issue 1704621 (the issue itself is already fixed here).
........
  r58963 | amaury.forgeotdarc | 2007-11-13 13:54:28 -0800 (Tue, 13 Nov 2007) | 23 lines

  Merge from py3k branch:
  Correction for issue1265 (pdb bug with "with" statement).

  When an unfinished generator-iterator is garbage collected, PyEval_EvalFrameEx
  is called with a GeneratorExit exception set.  This leads to funny results
  if the sys.settrace function itself makes use of generators.
  A visible effect is that the settrace function is reset to None.
  Another is that the eventual "finally" block of the generator is not called.

  It is necessary to save/restore the exception around the call to the trace
  function.

  This happens a lot with py3k: isinstance() of an ABCMeta instance runs
      def __instancecheck__(cls, instance):
          """Override for isinstance(instance, cls)."""
          return any(cls.__subclasscheck__(c)
                     for c in {instance.__class__, type(instance)})
  which lets an opened generator expression each time it returns True.

  Backport candidate, even if the case is less frequent in 2.5.
........
  r58968 | georg.brandl | 2007-11-14 05:59:09 -0800 (Wed, 14 Nov 2007) | 2 lines

  Remove dead link from random docs.
........
  r58971 | raymond.hettinger | 2007-11-14 14:56:16 -0800 (Wed, 14 Nov 2007) | 1 line

  Make __fields__ read-only.  Suggested by Issac Morland
........
  r58972 | raymond.hettinger | 2007-11-14 15:02:30 -0800 (Wed, 14 Nov 2007) | 1 line

  Add test for __fields__ being read-only
........
  r58975 | raymond.hettinger | 2007-11-14 18:44:53 -0800 (Wed, 14 Nov 2007) | 6 lines

  Accept Issac Morland's suggestion for __replace__ to allow multiple replacements
  (suprisingly, this simplifies the signature, improves clarity, and is comparably fast).
  Update the docs to reflect a previous change to the function name.
  Add an example to the docs showing how to override the default __repr__ method.
........
  r58976 | raymond.hettinger | 2007-11-14 18:55:42 -0800 (Wed, 14 Nov 2007) | 1 line

  Small improvement to the implementation of __replace__().
........
  r58977 | raymond.hettinger | 2007-11-14 18:58:20 -0800 (Wed, 14 Nov 2007) | 1 line

  Fixup example in docs.
........
  r58978 | raymond.hettinger | 2007-11-14 19:16:09 -0800 (Wed, 14 Nov 2007) | 1 line

  Example of multiple replacements.
........
  r58998 | raymond.hettinger | 2007-11-15 14:39:34 -0800 (Thu, 15 Nov 2007) | 1 line

  Add example for use cases requiring default values.
........
  r59000 | bill.janssen | 2007-11-15 15:03:03 -0800 (Thu, 15 Nov 2007) | 1 line

  add the certificate for the Python SVN repository for testing SSL
........
  r59004 | guido.van.rossum | 2007-11-15 16:24:44 -0800 (Thu, 15 Nov 2007) | 8 lines

  A patch from issue 1378 by roudkerk:

  Currently on Windows set_error() make use of a large array which maps
  socket error numbers to error messages.

  This patch removes that array and just lets PyErr_SetExcFromWindowsErr()
  generate the message by using the Win32 function FormatMessage().
........

docstrings conform to PEP 8 to someone else.)

Although WS2_32.dll doesn't have inet_pton the definition conflicts with ws2tcpip.h.

and utf-8, crashed when parsing a multiline string, or a line longer that 512
columns.

Seems to start compiling.

without calling PyType_Ready().

Question 1: Should the interpreter register all types with PyType_Ready()?
Many types seem to avoid it.

Question 2: To reproduce the problem, run the following code:
    def f():
        while True:
           for a in iter(range(0,1,10**20)):
              pass
    f()
And watch the memory used by the process.
How do we test this in a unittest?

#define for PyObject_Unicode in object.h.

In anticipation of the upcoming ssl patch I've added _ssl to PCbuild8. It's based on a copy of _socket.vcproj. My German version of VS 2005 alters almost every line of pcbuild.sln. Could somebody else please add _ssl.vcproj to the solution?

(I also cleared out all trailing whitespace in the C file.)

external symbol PyObject_Unicode'.

Building _ssl than hangs the build server because starting python.exe
displayes a message box that 'Python30.dll' cannot be found.

Temporary (?) solution: Look for existance of python30.dll and fail
the _ssl build when it is missing.

Tested on Windows by Christian Heimes.
I changed the code slightly, renaming decref_socketios() to
_decref_socketios(), and moving it closer to the close() method
that it calls.  Hopefully Bill can now submit his SSL port to 3.0.

Python failed to report an error when it wasn't able to open the PYTHONSTARTUP file.

On Windows _PySys_Init() could raise an uninitialized exception which leads to a seg fault.

Paths and file names from the fs should always be decoded with PyUnicode_DecodeFSDefault and not PyUnicode_FromString

print() mustn't raise an exception when sys.stdout is None.

After a long discussion about the problem with Windows GUI apps Guido decided that sys.stdin, stdout and stderr should be None when the C runtime library returns invalid file descriptors for the standard streams.

So far the only known cases are Windows GUI apps and scripts started with pythonw on Windows. The OS restrictions are tight enough to catch the problem on other OSes.

When an unfinished generator-iterator is garbage collected, PyEval_EvalFrameEx
is called with a GeneratorExit exception set.  This leads to funny results
if the sys.settrace function itself makes use of generators.
A visible effect is that the settrace function is reset to None.
Another is that the eventual "finally" block of the generator is not called.

It is necessary to save/restore the exception around the call to the trace
function.

This happens a lot with py3k: isinstance() of an ABCMeta instance runs
    def __instancecheck__(cls, instance):
        """Override for isinstance(instance, cls)."""
        return any(cls.__subclasscheck__(c)
                   for c in {instance.__class__, type(instance)})
which lets an opened generator expression each time it returns True.

Seems a backport candidate, even if the case is less frequent in 2.5.

The pre-commit hook doesn't allow a trailing newline