Return None rather than raise an exception if os.device_error is given an invalid file descriptor.

Fix to the last checkin, correctly raise an exception on failure.

Someone may have closed the file descriptor, with something like
f = open('test.test', 'w')
os.close(f.fileno())
f.close()
Protect against this by checking fd on windows before closing.

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

........
  r70574 | benjamin.peterson | 2009-03-23 20:11:37 -0500 (Mon, 23 Mar 2009) | 1 line

  fix typo
........

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

........
  r70558 | benjamin.peterson | 2009-03-23 17:29:45 -0500 (Mon, 23 Mar 2009) | 4 lines

  comply with the evilJavaNamingScheme for attribute names

  It seems my love of PEP 8 overrode the need for consistentcy
........
  r70561 | benjamin.peterson | 2009-03-23 18:10:14 -0500 (Mon, 23 Mar 2009) | 1 line

  refactor unittest docs
........
  r70562 | benjamin.peterson | 2009-03-23 18:13:36 -0500 (Mon, 23 Mar 2009) | 1 line

  forgot to document that setUp can be skipped (silly me...)
........
  r70563 | benjamin.peterson | 2009-03-23 18:19:03 -0500 (Mon, 23 Mar 2009) | 1 line

  update from CVS
........
  r70568 | benjamin.peterson | 2009-03-23 19:35:20 -0500 (Mon, 23 Mar 2009) | 1 line

  some cleanup and modernization
........
  r70569 | benjamin.peterson | 2009-03-23 19:36:16 -0500 (Mon, 23 Mar 2009) | 1 line

  remove special metadata
........
  r70570 | benjamin.peterson | 2009-03-23 19:37:12 -0500 (Mon, 23 Mar 2009) | 1 line

  update docstring
........
  r70571 | benjamin.peterson | 2009-03-23 19:39:24 -0500 (Mon, 23 Mar 2009) | 1 line

  add new skipping things to __all__
........

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

........
  r70555 | benjamin.peterson | 2009-03-23 16:50:21 -0500 (Mon, 23 Mar 2009) | 4 lines

  implement test skipping and expected failures

  patch by myself #1034053
........

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

........
  r70550 | antoine.pitrou | 2009-03-23 20:17:00 +0100 (lun., 23 mars 2009) | 3 lines

  The tracking statistics were actually too pessimistic
........

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

........
  r70546 | antoine.pitrou | 2009-03-23 19:41:45 +0100 (lun., 23 mars 2009) | 9 lines

  Issue #4688: Add a heuristic so that tuples and dicts containing only
  untrackable objects are not tracked by the garbage collector. This can
  reduce the size of collections and therefore the garbage collection overhead
  on long-running programs, depending on their particular use of datatypes.

  (trivia: this makes the "binary_trees" benchmark from the Computer Language
  Shootout 40% faster)
........

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

........
  r70542 | mark.dickinson | 2009-03-23 18:25:13 +0000 (Mon, 23 Mar 2009) | 14 lines

  Issue #5512: speed up the long division algorithm for Python longs.
  The basic algorithm remains the same; the most significant speedups
  come from the following three changes:

    (1) normalize by shifting instead of multiplying and dividing
    (2) the old algorithm usually did an unnecessary extra iteration of
        the outer loop; remove this.  As a special case, this means that
        long divisions with a single-digit result run twice as fast as
        before.
    (3) make inner loop much tighter.

  Various benchmarks show speedups of between 50% and 150% for long
  integer divisions and modulo operations.
........

(how annoying to have the same docs in two places)

(three if you count the docstring...)

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

........
  r70523 | lars.gustaebel | 2009-03-22 21:09:33 +0100 (Sun, 22 Mar 2009) | 5 lines

  Issue #5068: Fixed the tarfile._BZ2Proxy.read() method that would loop
  forever on incomplete input. That caused tarfile.open() to hang when used
  with mode 'r' or 'r:bz2' and a fileobj argument that contained no data or
  partial bzip2 compressed data.
........

........
  r70463 | benjamin.peterson | 2009-03-18 15:52:15 -0500 (Wed, 18 Mar 2009) | 1 line

  fix strange errors when setting attributes on tracebacks #4034
........

Merged revisions 70342,70385-70387,70389-70390,70392-70393,70395,70400,70405-70406,70418,70438,70464,70468 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/trunk

........
  r70342 | georg.brandl | 2009-03-13 14:03:58 -0500 (Fri, 13 Mar 2009) | 1 line

  #5486: typos.
........
  r70385 | benjamin.peterson | 2009-03-15 09:38:55 -0500 (Sun, 15 Mar 2009) | 1 line

  fix tuple.index() error message #5495
........
  r70386 | georg.brandl | 2009-03-15 16:32:06 -0500 (Sun, 15 Mar 2009) | 1 line

  #5496: fix docstring of lookup().
........
  r70387 | georg.brandl | 2009-03-15 16:37:16 -0500 (Sun, 15 Mar 2009) | 1 line

  #5493: clarify __nonzero__ docs.
........
  r70389 | georg.brandl | 2009-03-15 16:43:38 -0500 (Sun, 15 Mar 2009) | 1 line

  Fix a small nit in the error message if bool() falls back on __len__ and it returns the wrong type: it would tell the user that __nonzero__ should return bool or int.
........
  r70390 | georg.brandl | 2009-03-15 16:44:43 -0500 (Sun, 15 Mar 2009) | 1 line

  #5491: clarify nested() semantics.
........
  r70392 | georg.brandl | 2009-03-15 16:46:00 -0500 (Sun, 15 Mar 2009) | 1 line

  #5488: add missing struct member.
........
  r70393 | georg.brandl | 2009-03-15 16:47:42 -0500 (Sun, 15 Mar 2009) | 1 line

  #5478: fix copy-paste oversight in function signature.
........
  r70395 | georg.brandl | 2009-03-15 16:51:48 -0500 (Sun, 15 Mar 2009) | 1 line

  #5276: document IDLESTARTUP and .Idle.py.
........
  r70400 | georg.brandl | 2009-03-15 16:59:37 -0500 (Sun, 15 Mar 2009) | 3 lines

  Fix markup in re docs and give a mail address in regex howto, so that
  the recommendation to send suggestions to the author can be followed.
........
  r70405 | georg.brandl | 2009-03-15 17:11:07 -0500 (Sun, 15 Mar 2009) | 7 lines

  Move the previously local import of threading to module level.

  This is cleaner and avoids lockups in obscure cases where a Queue
  is instantiated while the import lock is already held by another thread.

  OKed by Tim Peters.
........
  r70406 | hirokazu.yamamoto | 2009-03-15 17:43:14 -0500 (Sun, 15 Mar 2009) | 1 line

  Added skip for old MSVC.
........
  r70418 | georg.brandl | 2009-03-16 14:42:03 -0500 (Mon, 16 Mar 2009) | 1 line

  Add token markup.
........
  r70438 | benjamin.peterson | 2009-03-17 15:29:51 -0500 (Tue, 17 Mar 2009) | 1 line

  I thought this was begging for an example
........
  r70464 | benjamin.peterson | 2009-03-18 15:58:09 -0500 (Wed, 18 Mar 2009) | 1 line

  a much better example
........
  r70468 | benjamin.peterson | 2009-03-18 22:04:31 -0500 (Wed, 18 Mar 2009) | 1 line

  close files after comparing them
........

Blocked revisions 70286,70290,70292-70293,70295-70296,70298,70300,70305,70315,70319,70466,70470,70472-70473,70475,70477,70479,70485 via svnmerge

........
  r70286 | raymond.hettinger | 2009-03-09 19:06:05 -0500 (Mon, 09 Mar 2009) | 1 line

  Fix markup.
........
  r70290 | raymond.hettinger | 2009-03-09 20:07:30 -0500 (Mon, 09 Mar 2009) | 1 line

  Update url for the spec.
........
  r70292 | raymond.hettinger | 2009-03-09 23:40:24 -0500 (Mon, 09 Mar 2009) | 1 line

  Clarify the meaning of normal and subnormal.
........
  r70293 | raymond.hettinger | 2009-03-09 23:49:21 -0500 (Mon, 09 Mar 2009) | 1 line

  Add a version tag to the decimal module.
........
  r70295 | raymond.hettinger | 2009-03-10 03:16:05 -0500 (Tue, 10 Mar 2009) | 1 line

  Update the decimal FAQ for the from_float() classmethod and improve the recipe for remove_exponent() to make it cut and pasteable.
........
  r70296 | raymond.hettinger | 2009-03-10 04:31:48 -0500 (Tue, 10 Mar 2009) | 1 line

  Small optimization for corner case where maxlen==0.
........
  r70298 | raymond.hettinger | 2009-03-10 07:50:59 -0500 (Tue, 10 Mar 2009) | 1 line

  For collections.deque() objects, expose the maxlen parameter as a read-only attribute.
........
  r70300 | raymond.hettinger | 2009-03-10 08:04:30 -0500 (Tue, 10 Mar 2009) | 1 line

  Fix typo.
........
  r70305 | brett.cannon | 2009-03-10 23:51:06 -0500 (Tue, 10 Mar 2009) | 5 lines

  Require implementations for warnings.showwarning() support the 'line' argument.
  Was a DeprecationWarning for not supporting it since Python 2.6.

  Closes issue #3652.
........
  r70315 | raymond.hettinger | 2009-03-11 19:25:03 -0500 (Wed, 11 Mar 2009) | 1 line

  Add reference to solution for a commonly asked question.
........
  r70319 | raymond.hettinger | 2009-03-11 19:31:58 -0500 (Wed, 11 Mar 2009) | 1 line

  Issue 5477: Fix buglet in the itertools documentation.
........
  r70466 | raymond.hettinger | 2009-03-18 17:13:20 -0500 (Wed, 18 Mar 2009) | 1 line

  Use mixin methods where possible. (2.7 only -- these don't all exist in 3.0)
........
  r70470 | raymond.hettinger | 2009-03-19 10:21:10 -0500 (Thu, 19 Mar 2009) | 6 lines

  Improve implementation with better underlying data structure
  for O(1) deletions.  Big-Oh performance now the same as regular
  dictionaries.  Uses a doubly-linked list instead of a list/seq
  to track insertion order.
........
  r70472 | raymond.hettinger | 2009-03-19 14:24:43 -0500 (Thu, 19 Mar 2009) | 1 line

  Silence a compiler warning.
........
  r70473 | raymond.hettinger | 2009-03-19 14:59:58 -0500 (Thu, 19 Mar 2009) | 6 lines

  * Add clearer comment to initialization code.
  * Add optional argument to popitem() -- modeled
    after Anthon van der Neut's C version.
  * Fix method markup in docs.
........
  r70475 | raymond.hettinger | 2009-03-19 18:12:41 -0500 (Thu, 19 Mar 2009) | 6 lines

  * Add implementation notes.
  * Re-order methods so that those touching the underlying data
    structure come first and the derived methods come last.
........
  r70477 | raymond.hettinger | 2009-03-19 18:22:25 -0500 (Thu, 19 Mar 2009) | 1 line

  Fix typo
........
  r70479 | mark.dickinson | 2009-03-20 10:51:55 -0500 (Fri, 20 Mar 2009) | 3 lines

  Issue #4258:  Use 30-bit digits for Python longs, on 64-bit platforms.
  Backport of r70459.
........
  r70485 | raymond.hettinger | 2009-03-20 13:25:49 -0500 (Fri, 20 Mar 2009) | 1 line

  Add MutableSet example.
........

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

........
  r70499 | hirokazu.yamamoto | 2009-03-21 19:32:52 +0900 | 1 line

  There is no macro named SIZEOF_SSIZE_T. Should use SIZEOF_SIZE_T instead.
........

  when packing an integer: for example, struct.pack('=L', -1) now
  raises struct.error instead of returning b'\xff\xff\xff\xff'.

  Thanks Andreas Schawo for the patch.

in missing details for the sys module.

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

........
  r70489 | mark.dickinson | 2009-03-20 23:16:14 +0000 (Fri, 20 Mar 2009) | 4 lines

  Rewrite Py_ARITHMETIC_RIGHT_SHIFT so that it's valid for all signed
  integer types T, not just those for which "unsigned T" is legal.
........

whatsnew/2.7.rst to keep it in sync with the trunk version, and
replace SHIFT with PyLong_SHIFT in #error message

........
  r70479 | mark.dickinson | 2009-03-20 15:51:55 +0000 (Fri, 20 Mar 2009) | 3 lines

  Issue #4258:  Use 30-bit digits for Python longs, on 64-bit platforms.
  Backport of r70459.
........

 - new configure option --enable-big-digits
 - new structseq sys.int_info giving information about the internal format
By default, 30-bit digits are enabled on 64-bit machines but
disabled on 32-bit machines.