It's on the FreeBSD 10.0 buildbot:
http://buildbot.python.org/all/builders/AMD64%20FreeBSD%2010.0%203.x/builds/1169

======================================================================
ERROR: test_getsockaddrarg (test.test_socket.GeneralModuleTests)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/home/buildbot/koobs-freebsd10/3.x.koobs-freebsd10/build/Lib/test/test_socket.py", line 1159, in test_getsockaddrarg
    sock.bind((host, port))
OSError: [Errno 48] Address already in use

This test is inherently subject to a race condition:
"""
        port = support.find_unused_port()
        [...]
        try:
            self.assertRaises(OverflowError, sock.bind, (host, big_port))
            self.assertRaises(OverflowError, sock.bind, (host, neg_port))
            sock.bind((host, port))
        finally:
            sock.close()
"""

The buildbot being set up ton run the test suite with "-j4" is
probably the most important factor.

A possibility would be to try the find_unused_port() + bind() in a
loop a couple times.

For reference:

CPython code comments showing that this may be an anticipated problem:

http://hg.python.org/cpython/file/0830670a9d9d/Lib/test/support/__init__.py#l562

An example of another project that seems to have tackled this problem in the way neologix suggests:

http://openvswitch.org/pipermail/dev/2013-April/026430.html

I am inclined to frame this issue as a choice between LBYL (try to predict, subject to the race, whether a port will be available) and EAFP (blindly try a few random high ports before inferring failure).

Clearly the long term solution is to fix the problems in the cpython code referenced in msg205227, but in the short term is it worth attempting a work around as suggested in msg205131 ?

Here's a patch.

New changeset 897c9e6ddb1a by Charles-François Natali in branch '3.4':
Issue #19875: Fix random test_getsockaddrarg() failure.
http://hg.python.org/cpython/rev/897c9e6ddb1a

New changeset 619feea86ce4 by Charles-François Natali in branch 'default':
Issue #19875: Fix random test_getsockaddrarg() failure.
http://hg.python.org/cpython/rev/619feea86ce4

Should be fixed now, thanks!

Thank you for taking care of this Charles-François :) Requesting backport to 3.3 and 2.7 too please, both are open for fixes.

New changeset 57e3c4ae37ea by Charles-François Natali in branch '2.7':
Issue #19875: Fix random test_getsockaddrarg() failure.
http://hg.python.org/cpython/rev/57e3c4ae37ea

Backported to 2.7 (don't know how Iforgot it).
3.3 is only open for security issues, so not backporting.

Updating versions to reflect branch changes. Will come in handy for those tracking for manual packaging backports