* bpo-30796: Fix failures in signal delivery stress test
setitimer() can have a poor minimum resolution on some machines,
this would make the test reach its deadline (and a stray signal
could then kill a subsequent test).
* Make sure to clear the itimer after the test
* Improve signal delivery
Avoid using Py_AddPendingCall from signal handler, to avoid calling signal-unsafe functions.
* Remove unused function
* Improve comments
* Add stress test
* Adapt for --without-threads
* Add second stress test
* Add NEWS blurb
* Address comments @haypo
* Add Lib/test/signalinterproctester.py
* Don't disable the garbage collector anymore
* Don't use os.fork() with a subprocess to not inherit existing signal handlers
or threads: start from a fresh process
* Don't use UNIX kill command to send a signal but Python os.kill()
* Use a timeout of 10 seconds to wait for the signal instead of 1 second
* Always use signal.pause(), instead of time.wait(1), to wait for a signal
* Use context manager on subprocess.Popen
* remove code to retry on EINTR: it's no more needed since the PEP 475
* remove unused function exit_subprocess()
* Cleanup the code
I have compared output between pre- and post-patch runs of these tests
to make sure there's nothing missing and nothing broken, on both
Windows and Linux. The only differences I found were actually tests
that were previously *not* run.
when interrupted by a signal not in the *sigset* parameter, if the signal
handler does not raise an exception. signal.sigtimedwait() recomputes the
timeout with a monotonic clock when it is retried.
Remove test_signal.test_sigwaitinfo_interrupted() because sigwaitinfo() doesn't
raise InterruptedError anymore if it is interrupted by a signal not in its
sigset parameter.
retried with the recomputed delay, except if the signal handler raises an
exception (PEP 475).
Modify also test_signal to use a monotonic clock instead of the system clock.
- Use _testcapi.raise_signal() in test_signal
- close also os.pipe() file descriptors in some test_signal tests where they
were not closed properly
- Remove faulthandler._sigill() and faulthandler._sigbus(): reuse
_testcapi.raise_signal() in test_faulthandler
When signals are unblocked, pending signal ared delivered in the reverse order
of their number (also on Linux, not only on FreeBSD 6).
Don't sort signals by their number if signals were not blocked (test_signum).
* Don't create a thread at startup anymore to initialize the pthread library:
it changes the behaviour of many functions related to signal handling like
sigwait()
* Reenable test_sigtimedwait_poll() on FreeBSD 6
Move also the "ready" trigger after the installation of the signal handler and
the call to siginterrupt().
Use a timeout of 5 seconds instead of 3. Two seconds are supposed to be enough,
but some of our buildbots are really slow (especially the FreeBSD 6 VM).
Move also the "ready" trigger after the installation of the signal handler and
the call to siginterrupt().
Use a timeout of 5 seconds instead of 3. Two seconds are supposed to be enough,
but some of our buildbots are really slow (especially the FreeBSD 6 VM).
Backport commits 968b9ff9a059 and aff0a7b0cb12 from the default branch to 3.2
branch. Extract of the changelog messages:
"The previous tests used time.sleep() to synchronize two processes. If the host
was too slow, the test could fail.
The new tests only use one process, but they use a subprocess to:
- have only one thread
- have a timeout on the blocking read (select cannot be used in the test,
select always fail with EINTR, the kernel doesn't restart it)
- not touch signal handling of the parent process"
and
"Add a basic synchronization code between the child and the parent processes:
the child writes "ready" to stdout."
I replaced .communicate(timeout=3.0) by an explicit waiting loop using
Popen.poll().
... instead of using fork(): sig*wait*() functions behave differently (not
correctly) after a fork, especially on FreeBSD 6.
Skip also test_sigtimedwait_poll() on FreeBSD 6 because of a kernel bug.
On FreeBSD6, pthread_kill() doesn't work on the main thread before the creation
of the first thread. Create therefore a dummy thread (no-op) a startup to
initialize the pthread library.
Add also a test for this use case, test written by Charles-François Natali.
The previous tests used time.sleep() to synchronize two processes. If the host
was too slow, the test could fail.
The new tests only use one process, but they use a subprocess to:
- have only one thread
- have a timeout on the blocking read (select cannot be used in the test,
select always fail with EINTR, the kernel doesn't restart it)
- not touch signal handling of the parent process
The _tkinter module loads the Tcl library which creates a thread waiting events
in select(). This thread receives signals blocked by all other threads. We
cannot test blocked signals if the _tkinter module is loaded.
called. The test checks that SIG_UNBLOCK calls immediatly the signal handler of
the pending SIGUSR1. Improve also the tests using an exception (division by
zero) instead of a flag (a function attribute).
and/or change the signal mask of the calling thread.
Fix also tests of test_io using threads and an alarm: use pthread_sigmask() to
ensure that the SIGALRM signal is received by the main thread.
Original patch written by Jean-Paul Calderone.
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r85586 | gregory.p.smith | 2010-10-16 17:17:24 -0700 (Sat, 16 Oct 2010) | 2 lines
fix for netbsd.
........
r85587 | gregory.p.smith | 2010-10-16 17:43:10 -0700 (Sat, 16 Oct 2010) | 3 lines
applying netbsd-wizs-mod.patch from issue5510 -
fixes for netbsd (and dragonflybsd?)
........
r85596 | gregory.p.smith | 2010-10-16 19:14:36 -0700 (Sat, 16 Oct 2010) | 6 lines
Fix multiprocessing Semaphore's on netbsd5. SEM_VALUE_MAX is defined
as (~0U) on NetBSD which was causing it to appear as -1 when used as
a signed int for _multprocessing.SemLock.SEM_VALUE_MAX. This works
around the problem by substituting INT_MAX on systems where it appears
negative when used as an int.
........
r85597 | gregory.p.smith | 2010-10-16 19:57:19 -0700 (Sat, 16 Oct 2010) | 2 lines
skip test_itimer_virtual on NetBSD to prevent the test suite from hanging.
........
r85598 | gregory.p.smith | 2010-10-16 20:09:12 -0700 (Sat, 16 Oct 2010) | 2 lines
Avoid hanging the test on netbsd5.
........
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r85140 | brian.curtin | 2010-10-01 09:49:24 -0500 (Fri, 01 Oct 2010) | 4 lines
Fix#10003. Add SIGBREAK to the set of valid signals on Windows.
This fixes a regression noticed by bzr, introduced by issue #9324.
........
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r84556 | brian.curtin | 2010-09-06 11:04:10 -0500 (Mon, 06 Sep 2010) | 7 lines
Clean up the fix to #9324 with some of the suggestions raised on python-dev
in response to the original checkin.
Move the validation from the original loop into a switch statement,
and adjust a platform check in the tests.
........
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r83771 | brian.curtin | 2010-08-06 22:47:21 -0500 (Fri, 06 Aug 2010) | 3 lines
Fix an assertRaises situation and typo. Also pass all tests to run_unittest
rather than do it by platform -- the proper skips are in place already.
........
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r83763 | brian.curtin | 2010-08-06 14:27:32 -0500 (Fri, 06 Aug 2010) | 3 lines
Fix#9324: Add parameter validation to signal.signal on Windows in order
to prevent crashes.
........
svn+ssh://pythondev@svn.python.org/python/branches/py3k
................
r81188 | stefan.krah | 2010-05-15 11:41:27 +0200 (Sat, 15 May 2010) | 10 lines
Merged revisions 81185 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/trunk
........
r81185 | stefan.krah | 2010-05-15 11:31:08 +0200 (Sat, 15 May 2010) | 4 lines
If the timeout is exceeded, count the tests as skipped instead of just
issuing a warning.
........
................
svn+ssh://pythondev@svn.python.org/python/branches/py3k
................
r80296 | r.david.murray | 2010-04-20 21:51:57 -0400 (Tue, 20 Apr 2010) | 14 lines
Merged revisions 80144 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/trunk
........
r80144 | r.david.murray | 2010-04-17 01:26:26 -0400 (Sat, 17 Apr 2010) | 8 lines
Issue #3864: Skip three test_signal tests on freebsd6 due to platform bug.
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.
........
................
svn+ssh://pythondev@svn.python.org/python/branches/py3k
................
r80240 | stefan.krah | 2010-04-20 10:13:03 +0200 (Tue, 20 Apr 2010) | 15 lines
Merged revisions 80238 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/trunk
........
r80238 | stefan.krah | 2010-04-20 09:59:10 +0200 (Tue, 20 Apr 2010) | 9 lines
1) The timeout in the itimer tests was too low for slow or heavily
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.
........
................
svn+ssh://pythondev@www.python.org/python/branches/py3k
................
r76185 | mark.dickinson | 2009-11-09 20:08:57 +0000 (Mon, 09 Nov 2009) | 9 lines
Merged revisions 76182 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/trunk
........
r76182 | mark.dickinson | 2009-11-09 19:54:51 +0000 (Mon, 09 Nov 2009) | 1 line
Add extra information to a test_signal failure message to aid diagnosis of buildbot failure.
........
................
svn+ssh://pythondev@www.python.org/python/branches/py3k
................
r75238 | mark.dickinson | 2009-10-04 19:43:54 +0100 (Sun, 04 Oct 2009) | 11 lines
Merged revisions 75236 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/trunk
........
r75236 | mark.dickinson | 2009-10-04 19:38:39 +0100 (Sun, 04 Oct 2009) | 4 lines
Issue #7042: Fix test_signal failure on OS X 10.6 64-bit builds
(and also, with luck, on the x86 FreeBSD buildbot), by making sure
that some user time is used in test_itimer_virtual.
........
................
svn+ssh://pythondev@svn.python.org/python/trunk
........
r70554 | benjamin.peterson | 2009-03-23 16:25:15 -0500 (Mon, 23 Mar 2009) | 1 line
complain when there's no last exception
........
r70588 | benjamin.peterson | 2009-03-24 17:56:32 -0500 (Tue, 24 Mar 2009) | 1 line
fix newline issue in test summary
........
r70589 | benjamin.peterson | 2009-03-24 18:07:07 -0500 (Tue, 24 Mar 2009) | 1 line
another style nit
........
r70598 | benjamin.peterson | 2009-03-25 16:24:04 -0500 (Wed, 25 Mar 2009) | 1 line
add shorthands for expected failures and unexpected success
........
r70605 | benjamin.peterson | 2009-03-26 11:32:23 -0500 (Thu, 26 Mar 2009) | 1 line
remove uneeded function
........
r70611 | benjamin.peterson | 2009-03-26 13:35:37 -0500 (Thu, 26 Mar 2009) | 1 line
add much better tests for python version information parsing
........
r70612 | benjamin.peterson | 2009-03-26 13:55:48 -0500 (Thu, 26 Mar 2009) | 1 line
more and more implementations now support sys.subversion
........
r70613 | benjamin.peterson | 2009-03-26 13:58:30 -0500 (Thu, 26 Mar 2009) | 1 line
roll old test in with new one
........
r70614 | benjamin.peterson | 2009-03-26 14:09:21 -0500 (Thu, 26 Mar 2009) | 1 line
add support for PyPy
........
r70615 | benjamin.peterson | 2009-03-26 14:58:18 -0500 (Thu, 26 Mar 2009) | 5 lines
add some useful utilities for skipping tests with unittest's new skipping ability
most significantly apply a modified portion of the patch from #4242 with
patches for skipping implementation details
........
r70616 | benjamin.peterson | 2009-03-26 15:05:50 -0500 (Thu, 26 Mar 2009) | 1 line
rename TestCase.skip() to skipTest() because it causes annoying problems with trial #5571
........
r70617 | benjamin.peterson | 2009-03-26 15:17:27 -0500 (Thu, 26 Mar 2009) | 1 line
apply the second part of #4242's patch; classify all the implementation details in test_descr
........
r70618 | benjamin.peterson | 2009-03-26 15:48:25 -0500 (Thu, 26 Mar 2009) | 1 line
remove test_support.TestSkipped and just use unittest.SkipTest
........
r70619 | benjamin.peterson | 2009-03-26 15:49:40 -0500 (Thu, 26 Mar 2009) | 1 line
fix naming
........
r70620 | benjamin.peterson | 2009-03-26 16:10:30 -0500 (Thu, 26 Mar 2009) | 1 line
fix incorrect auto-translation of TestSkipped -> unittest.SkipTest
........
r70621 | benjamin.peterson | 2009-03-26 16:11:16 -0500 (Thu, 26 Mar 2009) | 1 line
must pass argument to get expected behavior ;)
........
r70623 | benjamin.peterson | 2009-03-26 16:30:10 -0500 (Thu, 26 Mar 2009) | 1 line
add missing import
........
r70624 | benjamin.peterson | 2009-03-26 16:30:54 -0500 (Thu, 26 Mar 2009) | 1 line
** is required here
........
r70626 | benjamin.peterson | 2009-03-26 16:40:29 -0500 (Thu, 26 Mar 2009) | 1 line
update email tests to use SkipTest
........
r70627 | benjamin.peterson | 2009-03-26 16:44:43 -0500 (Thu, 26 Mar 2009) | 1 line
fix another name
........
svn+ssh://pythondev@svn.python.org/python/trunk
........
r62194 | jeffrey.yasskin | 2008-04-07 01:04:28 +0200 (Mon, 07 Apr 2008) | 7 lines
Add enough debugging information to diagnose failures where the
HandlerBException is ignored, and fix one such problem, where it was thrown
during the __del__ method of the previous Popen object.
We may want to find a better way of printing verbose information so it's not
spammy when the test passes.
........
r62197 | mark.hammond | 2008-04-07 03:53:39 +0200 (Mon, 07 Apr 2008) | 2 lines
Issue #2513: enable 64bit cross compilation on windows.
........
r62198 | mark.hammond | 2008-04-07 03:59:40 +0200 (Mon, 07 Apr 2008) | 2 lines
correct heading underline for new "Cross-compiling on Windows" section
........
r62204 | gregory.p.smith | 2008-04-07 08:33:21 +0200 (Mon, 07 Apr 2008) | 4 lines
Use the new PyFile_IncUseCount & PyFile_DecUseCount calls appropriatly
within the standard library. These modules use PyFile_AsFile and later
release the GIL while operating on the previously returned FILE*.
........
r62205 | mark.summerfield | 2008-04-07 09:39:23 +0200 (Mon, 07 Apr 2008) | 4 lines
changed "2500 components" to "several thousand" since the number keeps
growning:-)
........
r62214 | georg.brandl | 2008-04-07 20:51:59 +0200 (Mon, 07 Apr 2008) | 2 lines
#2525: update timezone info examples in the docs.
........
r62219 | andrew.kuchling | 2008-04-08 01:57:07 +0200 (Tue, 08 Apr 2008) | 1 line
Write PEP 3127 section; add items
........
r62220 | andrew.kuchling | 2008-04-08 01:57:21 +0200 (Tue, 08 Apr 2008) | 1 line
Typo fix
........
r62221 | andrew.kuchling | 2008-04-08 03:33:10 +0200 (Tue, 08 Apr 2008) | 1 line
Typographical fix: 32bit -> 32-bit, 64bit -> 64-bit
........
r62227 | andrew.kuchling | 2008-04-08 23:22:53 +0200 (Tue, 08 Apr 2008) | 1 line
Add items
........
r62229 | amaury.forgeotdarc | 2008-04-08 23:27:42 +0200 (Tue, 08 Apr 2008) | 7 lines
Issue2564: Prevent a hang in "import test.autotest", which runs the entire test
suite as a side-effect of importing the module.
- in test_capi, a thread tried to import other modules
- re.compile() imported sre_parse again on every call.
........
r62230 | amaury.forgeotdarc | 2008-04-08 23:51:57 +0200 (Tue, 08 Apr 2008) | 2 lines
Prevent an error when inspect.isabstract() is called with something else than a new-style class.
........
r62231 | amaury.forgeotdarc | 2008-04-09 00:07:05 +0200 (Wed, 09 Apr 2008) | 8 lines
Issue 2408: remove the _types module
It was only used as a helper in types.py to access types (GetSetDescriptorType and MemberDescriptorType),
when they can easily be obtained with python code.
These expressions even work with Jython.
I don't know what the future of the types module is; (cf. discussion in http://bugs.python.org/issue1605 )
at least this change makes it simpler.
........
r62233 | amaury.forgeotdarc | 2008-04-09 01:10:07 +0200 (Wed, 09 Apr 2008) | 2 lines
Add a NEWS entry for previous checkin
........
r62234 | trent.nelson | 2008-04-09 01:47:30 +0200 (Wed, 09 Apr 2008) | 37 lines
- Issue #2550: The approach used by client/server code for obtaining ports
to listen on in network-oriented tests has been refined in an effort to
facilitate running multiple instances of the entire regression test suite
in parallel without issue. test_support.bind_port() has been fixed such
that it will always return a unique port -- which wasn't always the case
with the previous implementation, especially if socket options had been
set that affected address reuse (i.e. SO_REUSEADDR, SO_REUSEPORT). The
new implementation of bind_port() will actually raise an exception if it
is passed an AF_INET/SOCK_STREAM socket with either the SO_REUSEADDR or
SO_REUSEPORT socket option set. Furthermore, if available, bind_port()
will set the SO_EXCLUSIVEADDRUSE option on the socket it's been passed.
This currently only applies to Windows. This option prevents any other
sockets from binding to the host/port we've bound to, thus removing the
possibility of the 'non-deterministic' behaviour, as Microsoft puts it,
that occurs when a second SOCK_STREAM socket binds and accepts to a
host/port that's already been bound by another socket. The optional
preferred port parameter to bind_port() has been removed. Under no
circumstances should tests be hard coding ports!
test_support.find_unused_port() has also been introduced, which will pass
a temporary socket object to bind_port() in order to obtain an unused port.
The temporary socket object is then closed and deleted, and the port is
returned. This method should only be used for obtaining an unused port
in order to pass to an external program (i.e. the -accept [port] argument
to openssl's s_server mode) or as a parameter to a server-oriented class
that doesn't give you direct access to the underlying socket used.
Finally, test_support.HOST has been introduced, which should be used for
the host argument of any relevant socket calls (i.e. bind and connect).
The following tests were updated to following the new conventions:
test_socket, test_smtplib, test_asyncore, test_ssl, test_httplib,
test_poplib, test_ftplib, test_telnetlib, test_socketserver,
test_asynchat and test_socket_ssl.
It is now possible for multiple instances of the regression test suite to
run in parallel without issue.
........
r62235 | gregory.p.smith | 2008-04-09 02:25:17 +0200 (Wed, 09 Apr 2008) | 3 lines
Fix zlib crash from zlib.decompressobj().flush(val) when val was not positive.
It tried to allocate negative or zero memory. That fails.
........
r62237 | trent.nelson | 2008-04-09 02:34:53 +0200 (Wed, 09 Apr 2008) | 1 line
Fix typo with regards to self.PORT shadowing class variables with the same name.
........
r62238 | andrew.kuchling | 2008-04-09 03:08:32 +0200 (Wed, 09 Apr 2008) | 1 line
Add items
........
r62239 | jerry.seutter | 2008-04-09 07:07:58 +0200 (Wed, 09 Apr 2008) | 1 line
Changed test so it no longer runs as a side effect of importing.
........
Antoine Pitrou