First, listening and/or connecting often fails.
Second, the test should use try/finally to release all resources even in case of failure.

This is breaking most buildbots currently: http://www.python.org/dev/buildbot/trunk/


[snip]
test_listen_config_1_ok (test.test_logging.ConfigDictTest) ... Exception in thread Thread-1044:
Traceback (most recent call last):
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/threading.py", line 526, in __bootstrap_inner
    self.run()
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/threading.py", line 479, in run
    self.__target(*self.__args, **self.__kwargs)
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/logging/config.py", line 847, in serve
    server = rcvr(port=port, handler=hdlr)
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/logging/config.py", line 827, in __init__
    ThreadingTCPServer.__init__(self, (host, port), handler)
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/SocketServer.py", line 406, in __init__
    self.server_bind()
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/SocketServer.py", line 417, in server_bind
    self.socket.bind(self.server_address)
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/socket.py", line 222, in meth
    return getattr(self._sock,name)(*args)
error: [Errno 98] Address already in use

ok

======================================================================
ERROR: test_listen_config_10_ok (test.test_logging.ConfigDictTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/test/test_logging.py", line 1597, in test_listen_config_10_ok
    self.setup_via_listener(json.dumps(self.config10))
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/test/test_logging.py", line 1581, in setup_via_listener
    sock.connect(('localhost', PORT))
  File "/home/buildslave/python-trunk/trunk.norwitz-x86/build/Lib/socket.py", line 222, in meth
    return getattr(self._sock,name)(*args)
error: [Errno 111] Connection refused

It could take benefit of "test_support.find_unused_port".

Fix checked into trunk with following changes:
  - use of test_support.find_unused_port()
  - if connection is refused, a finally clause shuts down the server thread.

Marked as pending awaiting testing of these changes on buildbots.

Further update done to disable listening tests for now (r77986). Will try to investigate these - the trouble is, they work on my system :-(

You may use the decorator @unittest.skip("see issue #7857") for these tests, until it's fixed.
(or @unittest.expectedFailure, if you want to run them anyway)

http://docs.python.org/dev/library/unittest.html#unittest.skip

Tests now seem to be fine. Thanks, Benjamin Peterson, for the fix.

There are still random failures:

test test_logging failed -- Traceback (most recent call last):
  File "/home2/buildbot/slave/trunk.loewis-sun/build/Lib/test/test_logging.py", line 1612, in test_listen_config_10_ok
    ], stream=output)
  File "/home2/buildbot/slave/trunk.loewis-sun/build/Lib/test/test_logging.py", line 116, in assert_log_lines
    self.assertEquals(len(actual_lines), len(expected_values))
AssertionError: 0 != 2

See http://www.python.org/dev/buildbot/trunk/builders/sparc%20solaris10%20gcc%20trunk/builds/348

----- Original Message ----

> From: Antoine Pitrou <report@bugs.python..org>
> 
> There are still random failures:
> 

Ouch. I'm not sure that I can debug this quickly, because the tests work on all platforms I've got access to (Windows XP, Linux Jaunty) and even where they fail, the verbose re-test often succeeds. Perhaps I should just re-enable skipping of those tests for now to avoid problems on the buildbots, and see if I can get a reproducible error on the systems I have access to.

Any suggestions gratefully received :-)