a non-infinite timeout. Also make that method friendlier with applications
wanting to continue using the socket in clear-text mode, by disabling
OpenSSL's internal readahead. Thanks to Darryl Miles for guidance.
Issue #8108: test_ftplib's non-blocking SSL server now has proper handling
of SSL shutdowns.
loaded machines.
2) Even with the increased timeout, the OS does not guarantee that
a process will get a certain amount of virtual time in 60s, so
the failure is changed to a diagnostic.
Two itimer tests and an interprocess signal test fail on FreeBSD 6 if
any test that starts a thread runs before test_signal. Since FreeBSD7
does not show this behavior, the bug is most likely a platform bug,
so this patch just skips the failing tests on freebsd6.
Rather than depending on some sleep value, start up an interpreter
as a subprocess and communicate with it. Because subprocess pipes
can't be read from until EOF and I want to read from them before that,
use ctypes to peek by using PeekNamedPipe. Once the subprocess has
written the message, then it is ready to roll and accept signals.
After that, kill it.
David Bolen's buildbot isn't know for it's speed, and it seems that we
may have been trying to kill the subprocess before it was fully
initialized. I ran with this change on the bot itself and it seemed to work.
Instead of spaces between the filename and date (or whatever the string
is that follows the filename, if any) use tabs. This is what the unix
'diff' command does, for example, and difflib was intended to follow
the 'standard' way of doing diffs. This improves compatibility with
patch tools. The docs and examples are also changed to recommended that
the date format used be the ISO 8601 format, which is what modern diff
tools emit by default.
Patch by Anatoly Techtonik.
1. The assumptions in check_node() were too restrictive:
- Hardware addresses with universal_local_bit=1 are valid
(locally administered).
- Many of the tested functions (including uuid.getnode())
may return valid RFC 4122 random node IDs. These are pretty
much random 48-bit values with the multicast bit set to 1.
2. _unixdll_getnode() calls _uuid_generate_time(), which may be None
on some platforms. The resulting TypeError is now caught.
The test_classify_oldstyle and test_classify_newstyle methods of
test.test_inspect.TestClassesAndFunctions were previously almost
identical (aside from irrelevant whitespace and one semantic
difference). They now share a single helper.
Fixes issue #8363.
On slower Windows machines, waiting 0.1 seconds can sometimes not be
enough for a subprocess to start and be ready to accept signals, causing
the test to fail. One buildbot is also choking on input()/EOFError so
that was changed to not depend on input.
- rename to _test_structmembersType to avoid the class being automatically
called by test_capi
- allow space for trailing NUL in inplace_member field of all_structmembers
- use T_STRING_INPLACE instead of T_INPLACE_STRING as keyword argument
to _test_structmembersType initializer
- don't attempt to initialize inplace_member field if T_STRING_INPLACE
argument wasn't supplied.
os.kill takes one of two newly added signals, CTRL_C_EVENT and
CTRL_BREAK_EVENT, or any integer value. The events are a special case
which work with subprocess console applications which implement a
special console control handler. Any other value but those two will
cause os.kill to use TerminateProcess, outright killing the process.
This change adds win_console_handler.py, which is a script to implement
SetConsoleCtrlHandler and applicable handler function, using ctypes.
subprocess also gets another attribute which is a necessary flag to
creationflags in Popen in order to send the CTRL events.
*ReflectionKey functions used to not be documented or tested, but they are
now sufficiently documented and tested on platforms where they apply.
Additionally, fixed a bug in QueryReflectionKey which was returning an
incorrect value.
All tests pass from XP through Windows 7, on 32 and 64-bit platforms.
Also rewrite the Decimal __hash__ method so that it doesn't rely on
float('inf') being valid: float('inf') could raise an exception on
platforms not using IEEE 754 arithmetic.
Float to decimal comparison operations now return a result based on
the numeric values of the operands. Decimal.__hash__ has also been
fixed so that Decimal and float values that compare equal have equal
hash value.
Antoine Pitrou
Florent Xicluna
Senthil Kumaran
Ronald Oussoren
Victor Stinner
Tarek Ziadé
Stefan Krah
Raymond Hettinger
Martin v. Löwis
R. David Murray
Benjamin Peterson
Brian Curtin
Mark Dickinson
Philip Jenvey
Jean-Paul Calderone
Nick Coghlan
Michael Foord
Ezio Melotti
Eric Smith