Stefan Krah [Mon, 2 Apr 2012 17:10:20 +0000 (19:10 +0200)]
Testing the implicit initialization of the thread local context on first
access fails (expectedly) if other modules have already used decimal. The
only option is to remove the test.
Antoine Pitrou [Sun, 1 Apr 2012 15:25:49 +0000 (17:25 +0200)]
Issue #14151: Raise a ValueError, not a NameError, when trying to create
a multiprocessing Client or Listener with an AF_PIPE type address under
non-Windows platforms. Patch by Popa Claudiu.
Antoine Pitrou [Sun, 1 Apr 2012 15:19:09 +0000 (17:19 +0200)]
Issue #14151: Raise a ValueError, not a NameError, when trying to create
a multiprocessing Client or Listener with an AF_PIPE type address under
non-Windows platforms. Patch by Popa Claudiu.
Eli Bendersky [Sun, 1 Apr 2012 14:40:17 +0000 (17:40 +0300)]
Removing the test of Element that causes ref-leak in GC (issue #14464).
I will now continue investigating the cause of the ref-leak, but I wanted
to remove the test so that the refcount test in the buildbots could be clean.
The whole change (adding GC to Element) is not reverted because it improved
the situation (GC works for immediate cycles) and didn't cause regressions
(the test is new and was added together with the fix).
R David Murray [Sat, 31 Mar 2012 16:08:59 +0000 (12:08 -0400)]
Merge #14434: make tutorial link in 'help' banner version-specific
Without this fix, both 2.7 and 3.x would always point to the "current"
docs...which means that before this fix python 3.2 'help' pointed to the 2.7
tutorial.
R David Murray [Sat, 31 Mar 2012 16:06:35 +0000 (12:06 -0400)]
#14434: make tutorial link in 'help' banner version-specific
Without this fix, both 2.7 and 3.x would always point to the "current"
docs...which means that before this fix python 3.2 'help' pointed to the 2.7
tutorial.
Stefan Krah [Fri, 30 Mar 2012 12:12:20 +0000 (14:12 +0200)]
Use abort() rather than exit() to appease tools like rpmlint. abort() is used
in libmpdec to prevent undefined behavior if an invalid context is used. This
cannot occur for the _decimal module since user input for the context is
validated.
Issue #14435: Remove special block allocation code from floatobject.c
PyFloatObjects are now allocated using PyObject_MALLOC like all other
internal types, but maintain a limited freelist of objects at hand for
performance. This will result in more consistent memory usage by Python.
Eli Bendersky [Fri, 30 Mar 2012 08:44:15 +0000 (11:44 +0300)]
Issue #14006: improve the documentation of xml.etree.ElementTree
Removed the useless explanation of the Element data structure that started the
documentation page. Instead, the documentation now starts with a brief tutorial
skimming some of the capabilities of the module. The tutorial can be followed
by additional topic-specific sections (such as XPath support), and eventually
by a reference that goes over the module's classes and functions, as usual.
R David Murray [Thu, 29 Mar 2012 11:15:45 +0000 (07:15 -0400)]
#14416: conditionally add LOG_AUTHPRIV facility and LOG_ODELAY to syslog.
Unlike the other facilities, we don't use a fallback for AUTHPRIV if it
doesn't exist. Because it is intended for logging sensitive log messages, it
is better that a program trying to log such messages fail than that it log
them insecurely.
Victor Stinner [Wed, 28 Mar 2012 00:54:15 +0000 (02:54 +0200)]
time.time() now uses clock_gettime(CLOCK_REALTIME) if available
clock_gettime(CLOCK_REALTIME) has a better resolution than gettimeofday().
time.time() falls back on gettimeofday() (and then on other functions) on
error.