to make sure it fell within [-1, 1] just in case someone implementing
strftime() in libc was stupid enough to assume this. Turns out, though, some
OSs (e.g. zOS) are stupid enough to use values outside of this range for time
structs created by the system itself. So instead of throwing a ValueError,
tm_isdst is now normalized before being passed to strftime().
Fixes issue #6823. Thanks Robert Shapiro for diagnosing the problem and
contributing an initial patch.
to the libedit emulation of the readline API on OSX 10.5 or later.
This also adds a minimal testsuite for readline to check that the
history manipuation functions have the same interface with both
C libraries.
acquiring the import lock around fork() calls. This prevents other threads
from having that lock while the fork happens, and is the recommended way of
dealing with such issues. There are two other locks we care about, the GIL
and the Thread Local Storage lock. The GIL is obviously held when calling
Python functions like os.fork(), and the TLS lock is explicitly reallocated
instead, while also deleting now-orphaned TLS data.
This only fixes calls to os.fork(), not extension modules or embedding
programs calling C's fork() directly. Solving that requires a new set of API
functions, and possibly a rewrite of the Python/thread_*.c mess. Add a
warning explaining the problem to the documentation in the mean time.
This also changes behaviour a little on AIX. Before, AIX (but only AIX) was
getting the import lock reallocated, seemingly to avoid this very same
problem. This is not the right approach, because the import lock is a
re-entrant one, and reallocating would do the wrong thing when forking while
holding the import lock.
Will backport to 2.6, minus the tiny AIX behaviour change.
The filter argument must be a function that takes a TarInfo
object argument, changes it and returns it again. If the
function returns None the TarInfo object will be excluded
from the archive.
The exclude argument is deprecated from now on, because it
does something similar but is not as flexible.
canonical version, sw_vers, leaves off trailing zeros in the version number
(e.g. 10.6 instead of 10.6.0). Test now compensates by tacking on extra zeros
for the test comparison.
Fixes issue #6806.
No longer use tarfile.normpath() on pathnames. Store pathnames
unchanged, i.e. do not remove "./", "../" and "//" occurrences.
However, still convert absolute to relative paths.
Add an "exc_value" attribute to the _AssertRaisesContext context manager in the unittest package. This allows further tests on the exception that was raised after the context manager exits.
occur when writing to a BufferedRandom object (e.g. a file opened in "rb+" or
"wb+" mode) after having buffered a certain amount of data for reading. This
bug was not present in the pure Python implementation.
Yes, this is a serious issue.
'single' statements with partial sentences (so they test things like "try:" in
interactive mode). Others tests hit areas that previously failed in Jython. A
couple still fail in Jython, mainly due to the difficulty of parsing partial
sentences (but should be fixed by Jython 2.6).
R. David Murray
Skip Montanaro
Ezio Melotti
Andrew M. Kuchling
Kristján Valur Jónsson
Brett Cannon
Ronald Oussoren
Georg Brandl
Benjamin Peterson
Mark Dickinson
Thomas Wouters
Michael Foord
Lars Gustäbel
Armin Rigo
Amaury Forgeot d'Arc
Tarek Ziadé
Frank Wierzbicki
Gregory P. Smith
Antoine Pitrou
Raymond Hettinger