Attached is a patch to remove the last sleeps from test_mailbox. I believe this makes the test suite deterministic. It also shaves 4 seconds of fixed overhead off the test run time.

Nice ping pong you play..
I, buildbot get:

Using random seed 2215045
[1/1] test_mailbox
test test_mailbox failed -- multiple errors occurred; run in verbose mode for details
test test_mailbox failed -- Traceback (most recent call last):
  File "/Users/steffen/usr/opt/py3k/lib/python3.3/test/test_mailbox.py", line 1033, in test_lock_conflict
    ipc.signal('done')
  File "/Users/steffen/usr/opt/py3k/lib/python3.3/test/test_mailbox.py", line 1007, in __exit__
    self.c_sock.shutdown(socket.SHUT_RDWR)
socket.error: [Errno 57] Socket is not connected

test test_mailbox failed -- Traceback (most recent call last):
  File "/Users/steffen/usr/opt/py3k/lib/python3.3/test/test_mailbox.py", line 1033, in test_lock_conflict
    ipc.signal('done')
  File "/Users/steffen/usr/opt/py3k/lib/python3.3/test/test_mailbox.py", line 1007, in __exit__
    self.c_sock.shutdown(socket.SHUT_RDWR)
socket.error: [Errno 57] Socket is not connected

/Users/steffen/usr/opt/py3k/lib/python3.3/test/regrtest.py:1082: ResourceWarning: unclosed <socket.socket object, fd=5, family=2, type=1, proto=0>
gc.collect()
1 test failed:
test_mailbox
Re-running failed tests in verbose mode
Re-running test 'test_mailbox' in verbose mode

...

======================================================================
ERROR: test_lock_conflict (test.test_mailbox.TestMbox)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/Users/steffen/usr/opt/py3k/lib/python3.3/test/test_mailbox.py", line 1033, in test_lock_conflict
    ipc.signal('done')
  File "/Users/steffen/usr/opt/py3k/lib/python3.3/test/test_mailbox.py", line 1007, in __exit__
    self.c_sock.shutdown(socket.SHUT_RDWR)
socket.error: [Errno 57] Socket is not connected

======================================================================
ERROR: test_lock_conflict (test.test_mailbox.TestMMDF)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/Users/steffen/usr/opt/py3k/lib/python3.3/test/test_mailbox.py", line 1033, in test_lock_conflict
    ipc.signal('done')
  File "/Users/steffen/usr/opt/py3k/lib/python3.3/test/test_mailbox.py", line 1007, in __exit__
    self.c_sock.shutdown(socket.SHUT_RDWR)
socket.error: [Errno 57] Socket is not connected

Ran 332 tests in 55.487s

FAILED (errors=2)
test test_mailbox failed -- multiple errors occurred
ok

... repeats yet another two times (very long output though)

----------------------------------------------------------------------
Ran 332 tests in 151.484s

OK
[118424 refs]

Thanks for testing this. I was afraid something like that would happen, since socket implementations are different on different platforms. I presume you ran it on OSX.

Now I have to decide if I want to fix it, or if I should just switch to using threads.

Short glance into it, and after
commenting out
self.c_sock_close = self.c_sock_shutdown = True
in the parent process it ends up as
(maybe i can spend more time this evening):

15:36 ~/tmp $ python3 -E -Wd -m test -r -w -uall test_mailbox
Using random seed 9754656
[1/1] test_mailbox
/Users/steffen/usr/opt/py3k/lib/python3.3/mailbox.py:723: ResourceWarning: unclosed <socket.socket object, fd=5, family=2, type=1, proto=0>
return self._toc[key]
/Users/steffen/usr/opt/py3k/lib/python3.3/mailbox.py:77: ResourceWarning: unclosed <socket.socket object, fd=5, family=2, type=1, proto=0>
return self.get_message(key)
1 test OK.
[118589 refs]

I think the fix is to either put a try/except around the socket shutdown call, or to remove it entirely (I think things will still work right on linux without it). If you leave the self.c_sock_close = True in, it should take care of the resource warning.

'Was not allowed to look yesterday.
If the child only closes and not
self.c_sock_shutdown = True
(shutdown is an ugly word for a child anyway) then
(and i hope Apple did not modify the BSD network stack):

12:59 ~/tmp $ python3 -E -Wd -m test -r -w test_mailbox
Using random seed 1985762
[1/1] test_mailbox
1 test OK.

There was a recent buildbot failure on test_lock_conflict() because of a race.
Looking at your patch, I must be missing something, but why not simply use a multiprocessing condition to signal when the parent process has acquired the lock?
Otherwise you could simply use a pipe or a socketpair, it would be much simpler.
Also, there's a race: if the parent process calls connect() before the child calls listen(), he'll get ECONNREFUSED.

Probably because I'm a threading/multiprocessing neophyte :) (Though I just learned a bunch about programming with threads recently...)

Probably because I'm a threading/multiprocessing neophyte :)

That's a very good reason :-)

Here's a version using two multiprocessing events. Note that I use
timeouts for wait() just to avoid being stuck if something goes wrong:
the test now runs in a dozen ms.

By the way, next time you need a duplex communication between two
processes, you can use socketpair(), which returns a pair of connected
sockets.

Charles-François's patch looks good to me.

New changeset c6d41dd60d2d by Charles-François Natali in branch '2.7':
Issue bpo-11867: Make test_mailbox.test_lock_conflict deterministic (and fix a
http://hg.python.org/cpython/rev/c6d41dd60d2d

New changeset 0053b7c68a02 by Charles-François Natali in branch '3.2':
Issue bpo-11867: Make test_mailbox.test_lock_conflict deterministic (and fix a
http://hg.python.org/cpython/rev/0053b7c68a02

New changeset 020260ec44f2 by Charles-François Natali in branch 'default':
Issue bpo-11867: Make test_mailbox.test_lock_conflict deterministic (and fix a
http://hg.python.org/cpython/rev/020260ec44f2

Should be fixed now, thanks!

New changeset c9facd251725 by Charles-François Natali in branch '2.7':
Followup to issue bpo-11867: Use socketpair(), since FreeBSD < 8 doesn't really
http://hg.python.org/cpython/rev/c9facd251725

New changeset 9dee8a095faf by Charles-François Natali in branch '3.2':
Followup to issue bpo-11867: Use socketpair(), since FreeBSD < 8 doesn't really
http://hg.python.org/cpython/rev/9dee8a095faf

New changeset 9014e0cc5589 by Charles-François Natali in branch 'default':
Followup to issue bpo-11867: Use socketpair(), since FreeBSD < 8 doesn't really
http://hg.python.org/cpython/rev/9014e0cc5589