been increased from 100 to 500 and the cache replacement policy has changed
from simply clearing the entire cache on overflow to randomly forgetting 20%
of the existing cached compiled regular expressions. This is a performance
win for applications that use a lot of regular expressions and limits the
impact of the performance hit anytime the cache is exceeded.
Changes the previously private attributes to make them public, increasing the potential for extending the library in user code. Backward-compatible and documented.
The EPIPE error occurs when the server closes the socket and the client sends a
"big" XML-RPC request (I don't know exactly the size threshold).
request() just have to ignore the error because single_request() closes the
socket on error, and so the next call to single_request() will open a new
socket.
Remove also a comment in the HTTP client because it's now wrong: see r70643
and issue #5542.
with the fact that getgroups(2) might return
more that MAX_GROUPS on OSX.
See the issue (and python-dev archives) for the
gory details. Summarized: OSX behaves rather oddly
and Apple says this is intentional.
the Crash Reporter on OSX test_subprocess will trigger
the reporter.
This patch prints a warning when the Crash Reporter will
get triggered intentionally, which should avoid confusing
people.
in distutils.sysconfig matches that in the
toplevel sysconfig module.
Without this patch universal builds on OSX are
broken.
Als add a test that checks that the two version
of get_config_vars agree on important values.
is wrong when PY_LDFLAGS is not empty.
The bug was caused by LDSHARED getting expanded *before* sysconfig
renamed PY_LDSHARED (and simular values) to names without a PY_
prefix.
The patch tries to maintain the intended behaviour of allowing users
to set LDFLAGS in the environment and have that affect the build.
Without this patch a universal build on OSX cannot build universal
(fat binary) extensions.
configure to append to Python's default values for those variables, and
similarly allow users to set $XXFLAGS on the make command line to append to the
values set by configure.
In the makefile, this renames the variables that used to be $XXFLAGS to
$PY_XXFLAGS, and renames the old $PY_CFLAGS to $PY_CORE_CFLAGS. To compensate,
sysconfig now aliases $XXFLAGS=$PY_XXFLAGS so that scripts using it keep
working. I see that as the right interface, not a backward-compatibility hack,
since these are logically the $XXFLAGS variables; we just use a different name
in the makefile to deal with make's semantics.
using WindowsError in a try/except. Only add WindowsError to the list of
exceptions to catch when we are actually running on Windows.
Additionally, add a call that was left out in test_posixpath.
Thanks Amaury, Antoine, and Jason.
Added Windows support for os.symlink when run on Windows 6.0 or greater,
aka Vista. Previous Windows versions will raise NotImplementedError
when trying to symlink.
Includes numerous test updates and additions to test_os, including
a symlink_support module because of the fact that privilege escalation
is required in order to run the tests to ensure that the user is able
to create symlinks. By default, accounts do not have the required
privilege, so the escalation code will have to be exposed later (or
documented on how to do so). I'll be following up with that work next.
Note that the tests use ctypes, which was agreed on during the PyCon
language summit.
This required moving the class from importlib/abc.py into
importlib/_bootstrap.py and jiggering some code to work better with the class.
This included changing how the file finder worked to better meet import
semantics. This also led to fixing importlib to handle the empty string from
sys.path as import currently does (and making me wish we didn't support that
instead just required people to insert '.' instead to represent cwd).
It also required making the new set_data abstractmethod create
any needed subdirectories implicitly thanks to __pycache__ (it was either this
or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir
method or have set_data with no data arg mean to create a directory).
Lastly, as an optimization the file loaders cache the file path where the
finder found something to use for loading (this is thanks to having a
sourceless loader separate from the source loader to simplify the code and
cut out stat calls).
Unfortunately test_runpy assumed a loader would always work for a module, even
if you changed from underneath it what it was expected to work with. By simply
dropping the previous loader in test_runpy so the proper loader can be returned
by the finder fixed the failure.
At this point importlib deviates from import on two points:
1. The exception raised when trying to import a file is different (import does
an explicit file check to print a special message, importlib just says the path
cannot be imported as if it was just some module name).
2. the co_filename on a code object is not being set to where bytecode was
actually loaded from instead of where the marshalled code object originally
came from (a solution for this has already been agreed upon on python-dev but has
not been implemented yet; issue8611).
This required moving the class from importlib/abc.py into
importlib/_bootstrap.py and jiggering some code to work better with the class.
This included changing how the file finder worked to better meet import
semantics. This also led to fixing importlib to handle the empty string from
sys.path as import currently does (and making me wish we didn't support that
instead just required people to insert '.' instead to represent cwd).
It also required making the new set_data abstractmethod create
any needed subdirectories implicitly thanks to __pycache__ (it was either this
or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir
method or have set_data with no data arg mean to create a directory).
Lastly, as an optimization the file loaders cache the file path where the
finder found something to use for loading (this is thanks to having a
sourceless loader separate from the source loader to simplify the code and
cut out stat calls).
Unfortunately test_runpy assumed a loader would always work for a module, even
if you changed from underneath it what it was expected to work with. By simply
dropping the previous loader in test_runpy so the proper loader can be returned
by the finder fixed the failure.
At this point importlib deviates from import on two points:
1. The exception raised when trying to import a file is different (import does
an explicit file check to print a special message, importlib just says the path
cannot be imported as if it was just some module name).
2. the co_filename on a code object is not being set to where bytecode was
actually loaded from instead of where the marshalled code object originally
came from (a solution for this has already been agreed upon on python-dev but has
not been implemented yet; issue8611).
This required moving the class from importlib/abc.py into
importlib/_bootstrap.py and jiggering some code to work better with the class.
This included changing how the file finder worked to better meet import
semantics. This also led to fixing importlib to handle the empty string from
sys.path as import currently does (and making me wish we didn't support that
instead just required people to insert '.' instead to represent cwd).
It also required making the new set_data abstractmethod create
any needed subdirectories implicitly thanks to __pycache__ (it was either this
or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir
method or have set_data with no data arg mean to create a directory).
Lastly, as an optimization the file loaders cache the file path where the
finder found something to use for loading (this is thanks to having a
sourceless loader separate from the source loader to simplify the code and
cut out stat calls).
Unfortunately test_runpy assumed a loader would always work for a module, even
if you changed from underneath it what it was expected to work with. By simply
dropping the previous loader in test_runpy so the proper loader can be returned
by the finder fixed the failure.
At this point importlib deviates from import on two points:
1. The exception raised when trying to import a file is different (import does
an explicit file check to print a special message, importlib just says the path
cannot be imported as if it was just some module name).
2. the co_filename on a code object is not being set to where bytecode was
actually loaded from instead of where the marshalled code object originally
came from (a solution for this has already been agreed upon on python-dev but has
not been implemented yet; issue8611).
svn+ssh://pythondev@svn.python.org/python/trunk
........
r82492 | victor.stinner | 2010-07-03 15:36:19 +0200 (sam., 03 juil. 2010) | 3 lines
Issue #7673: Fix security vulnerability (CVE-2010-2089) in the audioop module,
ensure that the input string length is a multiple of the frame size
........
1) #8271: when a byte sequence is invalid, only the start byte and all the
valid continuation bytes are now replaced by U+FFFD, instead of replacing
the number of bytes specified by the start byte.
See http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf (pages 94-95);
2) 5- and 6-bytes-long UTF-8 sequences are now considered invalid (no changes
in behavior);
3) Change the error messages "unexpected code byte" to "invalid start byte"
and "invalid data" to "invalid continuation byte";
4) Add an extensive set of tests in test_unicode;
5) Fix test_codeccallbacks because it was failing after this change.
svn+ssh://pythondev@svn.python.org/python/trunk
........
r82403 | benjamin.peterson | 2010-06-30 12:11:08 -0500 (Wed, 30 Jun 2010) | 1 line
mark test depending on ref counting
........
Georg Brandl
Ezio Melotti
Victor Stinner
Alexander Belopolsky
Florent Xicluna
Gregory P. Smith
Brian Curtin
Andrew M. Kuchling
Michael Foord
Ronald Oussoren
Richard Jones
Brett Cannon
Martin v. Löwis
Tarek Ziadé
Antoine Pitrou
Doug Hellmann
Benjamin Peterson
Senthil Kumaran
Stefan Krah
R. David Murray
Mark Dickinson
Jeffrey Yasskin
Nick Coghlan
Giampaolo Rodolà