Suppport conditional PLIST inclusion of the expected egg file, because
python2.4 distutils does not create them but later versions do. (Note
that PRINT_PLIST_AWK production is not working, probably due to order
of substitution.)
Earlier patch reviewed by wiz@.
egg info files in non-egg python distributions.
Add a comment that PY_NO_EGG appears wrong. Obviously there is a
reason it was added, but we need to figure out why and if it's still
true.
What's New in Python 2.5.4?
===========================
*Release date: 23-Dec-2008*
Core and builtins
-----------------
- Revert patch for #1706039, as it can crash the interpreter.
- Added test case to ensure attempts to read from a file opened for writing
fail.
What's New in Python 2.5.3?
===========================
*Release date: 19-Dec-2008*
Build
-----
- In the OSX installer, update SQLite to 3.6.7, and change bsddb URL.
Build against system Tcl framework.
What's New in Python 2.5.3c1?
=============================
*Release date: 13-Dec-2008*
Core and builtins
-----------------
- Issue #1706039: Support continued reading from a file even after
EOF was hit.
- Issue #1683: prevent forking from interfering in threading storage.
- Issue #4597: Fixed several opcodes that weren't always propagating
exceptions.
- Issue #4589: Propagated an exception thrown by a context manager's
__exit__ method's result while it's being converted to bool.
- Issue #4317: Fixed a crash in the imageop.rgb2rgb8() function.
- Issue #4230: If ``__getattr__`` is a descriptor, it now functions correctly.
- Issue #4048: The parser module now correctly validates relative imports.
- Issue #4176: Fixed a crash when pickling an object which ``__reduce__``
method does not return iterators for the 4th and 5th items.
- Issue #3967: Fixed a crash in the count() and find() methods of string-like
objects, when the "start" parameter is a huge value.
- Issue #3936: The parser warnings for using "as" and "with" as variable names
didn't fire after import statements.
- Issue #3751: str.rpartition would perform a left-partition when called with
a unicode argument.
- Issue #3537: Fix an assertion failure when an empty but presized dict
object was stored in the freelist.
- Apply security patches from Apple.
- Issue #2620: Overflow checking when allocating or reallocating memory
was not always being done properly in some python types and extension
modules. PyMem_MALLOC, PyMem_REALLOC, PyMem_NEW and PyMem_RESIZE have
all been updated to perform better checks and places in the code that
would previously leak memory on the error path when such an allocation
failed have been fixed.
- Issue #2242: Fix a crash when decoding invalid utf-7 input on certain
Windows / Visual Studio versions.
- Issue #3360: Fix incorrect parsing of '020000000000.0', which
produced a ValueError instead of giving the correct float.
- Issue #3242: Fix a crash inside the print statement, if sys.stdout is
set to a custom object whose write() method happens to install
another file in sys.stdout.
- Issue #3088: Corrected a race condition in classes derived from
threading.local: the first member set by a thread could be saved in
another thread's dictionary.
- Issue #3100: Corrected a crash on deallocation of a subclassed weakref which
holds the last (strong) reference to its referent.
- Issue #1686386: Tuple's tp_repr did not take into account the possibility of
having a self-referential tuple, which is possible from C code. Nor did
object's tp_str consider that a type's tp_str could do something that could
lead to an inifinite recursion. Py_ReprEnter() and Py_EnterRecursiveCall(),
respectively, fixed the issues. (Backport of r58288 from trunk.)
- Patch #1442: properly report exceptions when the PYTHONSTARTUP file
cannot be executed.
- The compilation of a class nested in another class used to leak one
reference on the outer class name.
- Issue #1477: With narrow Unicode builds, the unicode escape sequence
\Uxxxxxxxx did not accept values outside the Basic Multilingual Plane. This
affected raw unicode literals and the 'raw-unicode-escape' codec. Now
UTF-16 surrogates are generated in this case, like normal unicode literals
and the 'unicode-escape' codec.
- Issue #2321: use pymalloc for unicode object string data to reduce
memory usage in some circumstances.
- Issue #2238: Some syntax errors in *args and **kwargs expressions could give
bogus error messages.
- Issue #2587: In the C API, PyString_FromStringAndSize() takes a signed size
parameter but was not verifying that it was greater than zero. Values
less than zero will now raise a SystemError and return NULL to indicate a
bug in the calling C code.
- Issue #2588, #2589: Fix potential integer underflow and overflow
conditions in the PyOS_vsnprintf C API function.
- Issue #1204: The configure script now tests for additional libraries
that may be required when linking against readline. This fixes issues
with x86_64 builds on some platforms (a few Linux flavors and OpenBSD).
- Issue #3678: Correctly pass LDFLAGS and LDLAST to the linker on shared
library targets in the Makefile.
Library
-------
- Issue #3767: Convert Tk object to string in tkColorChooser.
- Issue #4342: Always convert Text.index result to string.
- Issue 3248: Allow placing ScrolledText in a PanedWindow.
- Issue #4084: Fix max, min, max_mag and min_mag Decimal methods to
give correct results in the case where one argument is a quiet NaN
and the other is a finite number that requires rounding.
- Issue #1776581 and #4302. Minor corrections to smtplib.
- Issue #3774: Fixed an error when create a Tkinter menu item without command
and then remove it.
- Assigning methods to ctypes.Structure and ctypes.Union subclasses
after creation of the class does now work correctly. See Issue #1700288.
- Issue #3895: _lsprof could be crashed with an external timer that did not
return a float when a Profiler object is garbage collected.
- Issues #3968 and #3969: two minor turtle problems.
- Issue #3547: Fixed ctypes structures bitfields of varying integer
sizes.
- Issue #3762: platform.architecture() fails if python is lanched via
its symbolic link.
- Issue #3554: ctypes.string_at and ctypes.wstring_at did call Python
api functions without holding the GIL, which could lead to a fatal
error when they failed.
- Issue #2234: distutils failed for some versions of the cygwin compiler. The
version reported by these tools does not necessarily follow the python
version numbering scheme, so the module is less strict when parsing it.
- Issue #2222: Fixed reference leak when occured os.rename()
fails unicode conversion on 2nd parameter. (windows only)
- Issue #3134: shutil referenced undefined WindowsError symbol.
- Issue #1342811: Fix leak in Tkinter.Menu.delete. Commands associated to
menu entries were not deleted.
- Issue #799428: Fix Tkinter.Misc._nametowidget to unwrap Tcl command objects.
- Issue #3339: dummy_thread.acquire() could return None which is not a valid
return value.
- Issue #3116 and #1792: Fix quadratic behavior in marshal.dumps().
- Issue #2682: ctypes callback functions no longer contain a cyclic
reference to themselves.
- Issue #2670: Fix a failure in urllib2.build_opener(), when passed two
handlers that derive the same default base class.
- Issue #2495: tokenize.untokenize now inserts a space between two consecutive
string literals; previously, ["" ""] was rendered as [""""], which is
incorrect python code.
- Issue #2482: Make sure that the coefficient of a Decimal is always
stored as a str instance, not as a unicode instance. This ensures
that str(Decimal) is always an instance of str. This fixes a
regression from Python 2.5.1 to Python 2.5.2.
- Issue #2478: fix failure of decimal.Decimal(0).sqrt()
- Issue #2432: give DictReader the dialect and line_num attributes
advertised in the docs.
- Issue #1747858: Fix chown to work with large uid's and gid's on 64-bit
platforms.
- Bug #2220: handle rlcompleter attribute match failure more gracefully.
- Bug #1725737: In distutil's sdist, exclude RCS, CVS etc. also in the
root directory, and also exclude .hg, .git, .bzr, and _darcs.
- Bug #1389051: imaplib causes excessive memory fragmentation when reading
large messages.
- Bug #1389051, 1092502: fix excessively large memory allocations when
calling .read() on a socket object wrapped with makefile().
- Bug #1433694: minidom's .normalize() failed to set .nextSibling for
last child element.
- Issue #2791: subprocess.Popen.communicate explicitly closes its
stdout and stderr fds rather than leaving them open until the
instance is destroyed.
- Issue #2632: Prevent socket.read(bignumber) from over allocating memory
in the common case when the data is returned from the underlying socket
in increments much smaller than bignumber.
- Issue #1857: subprocess.Popen.poll gained an additional _deadstate keyword
argument in python 2.5, this broke code that subclassed Popen to include its
own poll method. Fixed my moving _deadstate to an _internal_poll method.
- Issue #2113: Fix error in subprocess.Popen if the select system call is
interrupted by a signal.
- Issue #874900: after an os.fork() call the threading module state is cleaned
up in the child process to prevent deadlock and report proper thread counts
if the new process uses the threading module.
- Issue #3309: Fix bz2.BZFile iterator to release its internal lock
properly when raising an exception due to the bz2file being closed.
Prevents a deadlock.
Extension Modules
-----------------
- Issue #1040026: Fix os.times result on systems where HZ is incorrect.
- Issue #4228: Pack negative values the same way as 2.4 in struct's L format.
- Security Issue #2: imageop did not validate arguments correctly and could
segfault as a result.
- Issue 3886: [CVE-2008-2316] Possible integer overflow in the _hashopenssl
module was closed.
- Issue 1179: [CVE-2007-4965] Integer overflow in imageop module.
Also fixes rgbimg module.
- Issue #3205: When iterating over a BZ2File fails allocating memory, raise
a MemoryError rather than silently stop the iteration.
- Patch #2111: Avoid mmap segfault when modifying a PROT_READ block.
- zlib.decompressobj().flush(value) no longer crashes the interpreter when
passed a value less than or equal to zero.
- issue2858: Fix potential memory corruption when bsddb.db.DBEnv.lock_get
and other bsddb.db object constructors raised an exception.
- Issue #3120: On 64-bit Windows the subprocess module was truncating handles.
- Issue #1471: Arguments to fcntl.ioctl are no longer broken on 64-bit OpenBSD
and similar platforms due to sign extension.
- Issue #3312: Fix two crashes in sqlite3.
Tests
-----
- Issue #3863: Disabled a unit test of fork being called from a thread
when running on platforms known to exhibit OS bugs when attempting that.
- Issue #3261: test_cookielib had an improper file encoding specified.
- Patch #2232: os.tmpfile might fail on Windows if the user has no
permission to create files in the root directory.
Documentation
-------------
Build
-----
- Issue #4368: Don't define _XOPEN_SOURCE on FreeBSD 4.*.
on some platforms that lacked shared library support in the past. The
list hasn't been maintained at all and the gain is very limited, so just
get rid of it.
match py([0-9][0-9])-.*. Fixes breakage when building gnome which then
tries to build hamster-applet and chokes due to PYPKGPREFIX ending up
undefined.
releases is enough" rule of thumb. (python 2.3 was released 5 years ago.)
Keep python 1.5 and 2.1 though, because there are a handful of packages
that still need them.
anyway, simplify logic a bit:
Add Python 2.5 to the default list, but also mark it as incompatible if
wip/python25 is not present. Move the Darwin handling after setting a
default value.
Provide a new variable _PYTHON_VERSIONS_ACCEPTED that is filtered by
PYTHON_VERSIONS_INCOMPATIBLE. This helps to avoid providing broken
dependencies when a version is not supported as PYPKGPREFIX wouldn't be
set in that case.
We use the flat installation mode (e.g. no multiple versions by
default) by default. That avoids having to run DEINSTALL/INSTALL
scripts and allows DESTDIR installation.
Always prefer the default version and fallback to the first entry
of PYTHON_VERSIONS_ACCEPTED, which is supported by the current
system. Also honour PYTHON_VERSIONS_ACCEPTED and _INCOMPATIBLE,
when PYTHON_VERSION_REQD is used.
Python 2.4 is now in bugfix-only mode, no new features are being added. At
least 50 bugs have been squashed since Python 2.4.2, including a number of
bugs and potential bugs found by Coverity.
RECOMMENDED is removed. It becomes ABI_DEPENDS.
BUILDLINK_RECOMMENDED.foo becomes BUILDLINK_ABI_DEPENDS.foo.
BUILDLINK_DEPENDS.foo becomes BUILDLINK_API_DEPENDS.foo.
BUILDLINK_DEPENDS does not change.
IGNORE_RECOMMENDED (which defaulted to "no") becomes USE_ABI_DEPENDS
which defaults to "yes".
Added to obsolete.mk checking for IGNORE_RECOMMENDED.
I did not manually go through and fix any aesthetic tab/spacing issues.
I have tested the above patch on DragonFly building and packaging
subversion and pkglint and their many dependencies.
I have also tested USE_ABI_DEPENDS=no on my NetBSD workstation (where I
have used IGNORE_RECOMMENDED for a long time). I have been an active user
of IGNORE_RECOMMENDED since it was available.
As suggested, I removed the documentation sentences suggesting bumping for
"security" issues.
As discussed on tech-pkg.
I will commit to revbump, pkglint, pkg_install, createbuildlink separately.
Note that if you use wip, it will fail! I will commit to pkgsrc-wip
later (within day).
without underscores (REPLACE.*.old, REPLACE.*.new, and REPLACE_FILES.*).
Also convert REPLACE.*.new= ${SH:Q} back to ${SH}, as it should not be quoted
here, if at all.
Ok with rillig.
is not available for the platform.
Mark packages as "unavailable" instead of "broken" when they depend on
a python version that is not available for the platform.
by other package Makefiles, and with the deprecation of USE_BUILDLINK3
support in the infrastructure files, these had the potential to break
existing packages.
bin/python wrapper. Bump their PKGREVISION to 1.
Also remove the python package. In order to preserve a similar behavior to
this one using the new framework, start by installing pkg_alternatives.
Then just open its manpage, scroll down to the EXAMPLES section, Applying
filters subsection, and follow the (rather short) directions given there.
python*-pth packages into meta-packages which will install the non-pth
packages. Bump PKGREVISIONs on the non-pth versions to propagate the
thread change, but leave the *-pth versions untouched to not affect
existing installations.
Sync all PYTHON_VERSIONS_AFFECTED lines in package Makefiles.
Note that this cannot be changed to ${INSTALL} because it is not the install
command but rather an argument to setup.py. While adding the quotes makes
no difference to the use of the file, it does keep pkglint quiet. This is
even more important because this file is included by many other packages
and causes warnings when trying to check new Python extensions.
by package Makefiles as well as buildlink3s to prevent usage of particular
versions of Python -- even if they appear in PYTHON_VERSIONS_ACCEPTED.
(And make Darwin exclusion use this variable instead, so that it effectively
overrides any pkg's definition of _ACCEPTED.)
to the preferred version of python (as specified by PYTHON_VERSION_DEFAULT).
This means that cross-platform python scripts with "#!/usr/bin/env python"
actually function on pkgsrc systems without nasty hackery.
Approved by Alistair.
scheme only accepts `python', so if we have a script which uses:
#!/path/to/python23
PYTHON_PATCH_SCRIPTS will fail.
Fix provided by Min Sik Kim in PR pkg/24069.
This is a bug-fix release for Python 2.3 that fixes a number of bugs,
including a couple of serious errors with weakrefs and the cyclic garbage
collector. There are also a number of fixes to the standard library
- see the release notes ( http://www.python.org/2.3.3/NEWS.html )for details.
Inspired by FreeBSD "ports".
Fix the PLISTs accordingly.
Also, while at it, remove now obsolete compileall.py calls in post-install
targets and insure that extension.mk is in included before builinlinks of
other Python modules.
Discussed with/ok'ed by drochner@.
Change the order of PYSETUP*AGS and the corresponding target.
Eg.:
s/${PYSETUPINSTALLARGS} install/install ${PYSETUPINSTALLARGS}/
This allows us to automatically install .pyo files with a
PYSETUPINSTALLARGS=-O1 -c and thus drop the compileall.py stuff in the
post-install target.
ok'ed by drochner@
This is a bugfix release.
Also, use the LINK_ALL_LIBGCC_HACK just introduced to link the
interpreter against libgcc. This should fix PR pkg/16144 in the
Python2.2 (and NetBSD-1.5) case.
in subdirectory and need to execute setup.py in that directory, but
still need WRKSRC set to the base directory for configure/patch targets
to handle this, add PYSETUPSUBDIR variable (default empty), and
change do-build+do-install targets to use working directory
${WRKSRC}/${PYSETUPSUBDIR} when executing setup.py
versions of Python may be installed, and packages that require an
earlier version can always specify that, this should cause no problem,
and it seems sensible that people who don't care should get the latest.
(drochner concurs)
