1. Armin Rigo
  2. cpython-withatomic

Commits

Guido van Rossum  committed 37c0445

News for 2.2.2b1. I'm exhausted -- this was a multi-hour job of
poring over the logs. Quite likely I've forgotten some things,
introduced typos, and organized things less than ideally.

  • Participants
  • Parent commits 000d155
  • Branches 2.2

Comments (0)

Files changed (1)

File Misc/NEWS

View file
 Release date: 7-Oct-2002
 =============================
 
+Almost everything in this release is a pure bugfix and is backported
+from a corresponding bugfix alrady applied to Python 2.3.  While at
+the time of writing, Python 2.3 is still in pre-alpha form, only
+accessible via CVS, it receives continuous and extensive testing by
+its developers.  The list below is not a complete list of fixed bugs;
+it only lists fixed bugs that someone might be interested in hearing
+about.  Documentation fixes are not listed.
+
+Tip: to quickly find SourceForge bug or patch NNNNNN, use an URL of
+the form www.python.org/sf/NNNNNN.
+
 Core and builtins
 
 - Changed new-style class instantiation so that when C's __new__
   behavior.  If problems with this are reported within a week of the
   release of 2.2.2 beta 1, we may revert this change.)
 
+- Fix a core dump in type_new() when looking for the tp_init() slot.
+  This could call a garbage pointer when e.g. an ExtensionClass was
+  given.
+
+- A variety of very obscure end-case bugs in new-style classes were
+  fixed, some of which could be made to trigger core dumps with absurd
+  input.
+
 - u'%c' will now raise a ValueError in case the argument is an
   integer outside the valid range of Unicode code point ordinals.
 
+- Several small patches were applied that aren't bugfixes (and aren't
+  even backported from 2.3!) but make life easier for tools like Armin
+  Rigo's Psyco.  [SF patches 617309, 617311, 617312]
+
+- Made conversion failure error message consistent between types.
+
+- The complex() built-in now finds __complex__() in new-style
+  classes.  [SF bug 563740]
+
+- Fixed a problem in the UTF-8 decoder where a Unicode literal
+  containing a "lone surrogate" would cause a .pyc file to be written
+  that could not be read.  [SF bug 610783]
+
+- Fixed a problem with code objects whose stacksize is >= 2**15.
+  These cannot be marshalled correctly.  As a work-around, don't write
+  a .pyc file in this case.  [SF bug 561858]
+
+- Fixed several bugs that could cause issubclass() and isinstance() to
+  leave an exception lingering behind while returning a non-error
+  value.
+
+- The Unicode replace() method  would do the wrong thing for a unicode
+  subclass when there were zero string replacements.  [SF bug 599128]
+
+- Fixed some endcase bugs in Unicode rfind()/rindex() and endswith().
+  [SF bug 595350]
+
 - When x is an object whose class implements __mul__ and __rmul__,
   1.0*x would correctly invoke __rmul__, but 1*x would erroneously
   invoke __mul__.  This was due to the sequence-repeat code in the int
   type.  This has been fixed now.
 
-- If a dying instance of a new-style class got resurrected by its class's
-  __del__ method, Python aborted with a fatal error.
+- The __delete__ method wrapper wasn't supported.  [SF patch 588728]
+
+- If a dying instance of a new-style class got resurrected by its
+  class's __del__ method, Python aborted with a fatal error.
 
 - Source that creates parse nodes with an extremely large number of
   children (e.g.,  test_longexp.py) triggers problems with the
 - Classes using __slots__ are now properly garbage collected.
   [SF bug 519621]
 
+- Fixed an inefficiency in clearing the stack frame of new frame
+  objects.
+
 - Repaired a slow memory leak possible only in programs creating a
-  great many cyclic structures involving frames.  Reported on
-  SourceForge as bug 543148.
+  great many cyclic structures involving frames [SF bug 543148].
+
+- Fixed an esoteric performance glitch in GC.  [SF bug 574132]
 
 - A method zfill() was added to str and unicode, that fills a numeric
   string to the left with zeros.  For example,
   these make no sense.  Since this was documented, they're being
   deprecated now.
 
-- String methods lstrip(), rstrip() and strip() now take an optional
-  argument that specifies the characters to strip.  For example,
-  "Foo!!!?!?!?".rstrip("?!") -> "Foo". In addition, "200L".strip("L")
-  will return "200". This is useful for replacing code that assumed
-  longs will always be printed with a trailing "L".
+- String and Unicode methods lstrip(), rstrip() and strip() now take
+  an optional argument that specifies the characters to strip.  For
+  example, "Foo!!!?!?!?".rstrip("?!") -> "Foo". In addition,
+  "200L".strip("L") will return "200". This is useful for replacing
+  code that assumed longs will always be printed with a trailing "L".
+
+- A change to how new-style classes deal with __doc__: you can now
+  supply a __doc__ descriptor that returns something different for a
+  class than for instances of that class.
 
 Extension modules
 
-- If the size passed to mmap.mmap() is larger than the length of the
-  file on non-Windows platforms, a ValueError is raised. [SF bug 585792]
+- In readline.c: change completion to avoid appending a space
+  character; this is usually more useful when editing Python code.
+
+- Fixed a crash in debug builds for marshal.dumps([128] * 1000).  [SF
+  bug 588452]
+
+- In cPickle.c: more robust test of whether global objects are
+  accessible.  Added recursion limit to pickling [SF bug 576084].  Try
+  the persistent id code *before* calling save_global().
+
+- In mmapmpdule.c: if the size passed to mmap() is larger than the
+  length of the file on non-Windows platforms, a ValueError is
+  raised. [SF bug 585792]
+
+- In socketmodule.c: improve robustness of IPv6 code.
+
+- In _hotshot.c: fix broken logic in the logreader object.
+
+- In zlibmodule.c: fix for crash on second flush() call.  [SF bug
+  544995]
 
 Library
 
   API, so you should read the section in the library manual about the
   changes since email v1.
 
-- random.gauss() uses a piece of hidden state used by nothing else,
-  and the .seed() and .whseed() methods failed to reset it.  In other
-  words, setting the seed didn't completely determine the sequence of
-  results produced by random.gauss().  It does now.  Programs repeatedly
-  mixing calls to a seed method with calls to gauss() may see different
-  results now.
-
-- Some fixes in the copy module: when an object is copied through its
-  __reduce__ method, there was no check for a __setstate__ method on
-  the result [SF patch 565085]; deepcopy should treat instances of
-  custom metaclasses the same way it treats instances of type 'type'
-  [SF patch 560794].
+- In pydoc.py: Extend stripid() to handle strings ending in more than
+  one '>'; add resolve() to handle looking up objects and names (fix
+  SF bug 586931); add a nicer error message when given a filename that
+  doesn't exist.  Pretend that the docstring for non-callable objects
+  is always None; this makes for less confusing output and fixes the
+  problem reported in SF patch 550290.  Change the way 'less' is
+  invoked as a browser (on Unix) to make it more robust.
+
+- In pickle.py: Whichmodule() now skips dummy (None) package entries
+  in sys.modules.  Try the persistent id code *before* calling
+  save_global().
+
+- A variety of fixes were applied to the compiler package.
+
+- In distutils/: Fix distutils.sysconfig to understand that the
+  running Python is part of the build tree and needs to use the
+  appropriate "shape" of the tree [SF patch 547734].  Prefer rpmbuild
+  over rpm if available [SF patch 619493].   util.convert_path()
+  failed with empty pathname.  [SF bug 574235]
+
+- In posixpath.py and user.py: fixed SF bug 555779, "import user
+  doesn't work with CGIs."
+
+- In site.py: fixed a problem which triggered when sys.path was empty.
+
+- In smtpd.py: print the refused list to the DEBUGSTREAM [SF 515021];
+  removed an embarrassing debug line from smtp_RCPT().
+
+- In smtplib.py: fix multiline string in sendmail example [SF patch
+  586999]; handle empty addresses [SF bug 602029].
+
+- In urllib.py: treat file://localhost/ as local too (same as file:/
+  and file:///).  [SF bug 607789]
+
+- In warnings.py: ignore IOError when writing the message.
+
+- In ConfigParser.py: allow internal whitespace in keys [SF bug
+  583248]; use option name transform consistently in has_option() [SF
+  bug 561822]; misc other patches.
+
+- In sre_compile.py (the compile() function for the re module):
+  Disable big charsets in UCS-4 builds. [SF bug 599377]
+
+- In pre.py (the deprecated, *old* implementation of the re module):
+  fix broken sub() and subn().  [SF bug 570057]
+
+- In weakref.py: The WeakKeyDictionary constructor didn't work when a
+  dict arg was given.  [SF patch 564549]
+
+- In xml/: various fixes tracking PyXML.
+
+- In urllib2.py: fix proxy config with user+pass authentication.  [SF
+  patch 527518]
+
+- In pdb.py: Increase the maxstring value of _saferepr.  Add exit as
+  an alias for quit [SF bug 543674].  Fix crash on input line
+  consisting of one or more spaces [SF bug 579701].
+
+- In test/regrtest.py: added some sys.stdout.flush() calls.
+
+- In random.py:
+
+  - Deprecate (in comment) cunifvariate().  [SF bug 506647]
+
+  - Loosened the acceptable 'start' and 'stop' arguments to
+    randrange() so that any Python (bounded) ints can be used.  So,
+    e.g., randrange(-sys.maxint-1, sys.maxint) no longer blows up.
+    [SF bug 594996]
+
+  - The gauss() method uses a piece of hidden state used by nothing
+    else, and the .seed() and .whseed() methods failed to reset it.
+    In other words, setting the seed didn't completely determine the
+    sequence of results produced by random.gauss().  It does now.
+    Programs repeatedly mixing calls to a seed method with calls to
+    gauss() may see different results now.
+
+  - The randint() method is rehabilitated (i.e. no longer deprecated).
+  
+-  In copy.py: when an object is copied through its __reduce__ method,
+   there was no check for a __setstate__ method on the result [SF
+   patch 565085]; deepcopy should treat instances of custom
+   metaclasses the same way it treats instances of type 'type' [SF
+   patch 560794].
+
+- In turtle.py: update canvas before computing width; draw turtle when
+  done drawing circle.  [SF bug 612595]
+
+- In Tkinter.py: Canvas.select_item() now returns the selected item,
+  if any.  [SF patch 581396]
+
+- In multifile.py: *backed out* the change that stripped a trailing
+  \r\n.  This caused more problems than it fixed.  [SF bug 514676]
+
+- In rexec.py: fixed several security problems.  *This does not mean
+  that rexec is now considered safe!*
+
+- In os.py: security fixes for _execvpe().
+
+- In gzip.py: open files in binary mode.
+
+- In CGIHTTPServer.py: update os.environ regardless of hos it tries to
+  handle calls (fork, popen*, etc.).  Also fixed a flush() of a
+  read-only file (can't do that on MacOS X).
+
+- In urllib.py: in splituser(), allow @ in the userinfo field.  This
+  is not allowed by RFC 2396; however, other tools support unescaped
+  @'s so we should also.  [SF patch 596581, bug 581529]
+
+- In base64.py: decodestring('') should return '' instead of raising
+  an exception.  [SF bug 595671]
+
+- atexit.py: keep working if sys.exitfunc is already set when this is
+  first imported.
+
+- In copy.py: Make sure that *any* object whose id() is used as a memo
+  key is kept alive in the memo.  [SF bug 592567]
+
+- In httplib.py: fixed a variety of bugs.  The httplib.py in Python
+  2.2.2 is identical to that in the CVS head (at the time of the
+  release of 2.2.2).
+
+- In rfc822.py: change the default for Message.get() back to None.
+
+- In bdb.py: fix an old bug that made it impossible to continue after
+  hitting a breakpoint while in the bottom frame.
+
+- In Queue.py: use try/finally to ensure that all locks are properly
+  released.  [SF bug 544473]
+
+- In SocketServer.py: the correct initialization of self.wfile is
+  StringIO.StringIO(), not StringIO.StringIO(self.packet).  [SF bug
+  543318]
 
 Build
 
+- Various platform-specific problems were fixed, including most open
+  64-bit platform specific issues.
+
+- Updated Misc/RPM for Python 2.2.2b1; added Makefile.pre.in to -devel.
+
 - The fpectl module is not built by default; it's dangerous or useless
-  except in the hands of experts.
+  except in the hands of experts.  (At the same time, a fix for DEC
+  Alpha under Linux was applied.)
+
+- Better check for C++ linkage.  [SF bug 559429]
+
+- The errno module needs to be statically linked, since it is now
+  needed during the extension building phase.
 
 - A bug was fixed that could cause COUNT_ALLOCS builds to segfault, or
   get into infinite loops, when a new-style class got garbage-collected.
   that new-style classes be immortal in COUNT_ALLOCS builds.  Note that
   COUNT_ALLOCS is not enabled by default, in either release or debug
   builds, and that new-style classes are immortal only in COUNT_ALLOCS
-  builds.  SourceForge bug 578752.
+  builds.  [SF bug 578752]
+
+- In order to avoid problems with binutils 2.12 and later, test for
+  --export-dynamic in its help output.
 
 C API
 
 
 Windows
 
+- Improve handling of ^C on Windows.  [SF bug 439992]
+
+- Provide a fallback version of ntpath.abspath() when the nt module
+  can't be imported.
+
+- Fixed asyncore on Windows to avoid calling select() with three empty
+  lists.  Use time.sleep() instead, to match what happens on
+  Unix/Linux in that case.  [SF item 611464]
+
+- Fixed selectmodule.c to call WSAGetLastError() to retrieve the error
+  number.
+
+- Fixed the test for mmap so that it passes on Windows too.
+
 - SF bug 595919:  popenN return only text mode pipes
   popen2() and popen3() created text-mode pipes even when binary mode
   was asked for.  This was specific to Windows.
   of that has been fixed in the installer (disabled Wise's "delete in-
   use files" uninstall option).
 
+Other
+
+- Most changes to IDLE were backported, including some featurettes.
+
 
 What's New in Python 2.2.1 final?
 Release date: 10-Apr-2002