This patch comes from upstream: http://bugs.python.org/issue15881

It fixes a race condition in multiprocessing/util.py in which globals are

deleted during exit but still used during the exit.

diff -ur a/Lib/multiprocessing/util.py b/Lib/multiprocessing/util.py

--- a/Lib/multiprocessing/util.py	2014-11-12 08:02:39.431360950 -0800

+++ b/Lib/multiprocessing/util.py	2014-11-12 08:03:56.970704872 -0800

@@ -221,6 +221,12 @@

     Finalizers with highest priority are called first; finalizers with

     the same priority will be called in reverse order of creation.

     '''

+    if _finalizer_registry is None:

+        # This function may be called after this module's globals are

+        # destroyed.  See the _exit_function function in this module for more

+        # notes.

+        return

+    

     if minpriority is None:

         f = lambda p : p[0][0] is not None

     else:

@@ -252,21 +258,38 @@

 _exiting = False

-def _exit_function():

+def _exit_function(info=info, debug=debug, _run_finalizers=_run_finalizers,

+                   active_children=active_children,

+                   current_process=current_process):

+    # NB: we hold on to references to functions in the arglist due to the

+    # situation described below, where this function is called after this

+    # module's globals are destroyed.

+

     global _exiting

     info('process shutting down')

     debug('running all "atexit" finalizers with priority >= 0')

     _run_finalizers(0)

-    for p in active_children():

-        if p._daemonic:

-            info('calling terminate() for daemon %s', p.name)

-            p._popen.terminate()

-

-    for p in active_children():

-        info('calling join() for process %s', p.name)

-        p.join()

+    if current_process() is not None:

+        # NB: we check if the current process is None here because if

+        # it's None, any call to ``active_children()`` will throw an

+        # AttributeError (active_children winds up trying to get

+        # attributes from util._current_process).  This happens in a

+        # variety of shutdown circumstances that are not well-understood

+        # because module-scope variables are not apparently supposed to

+        # be destroyed until after this function is called.  However,

+        # they are indeed destroyed before this function is called.  See

+        # issues 9775 and 15881.  Also related: 4106, 9205, and 9207.

+

+        for p in active_children():

+            if p._daemonic:

+                info('calling terminate() for daemon %s', p.name)

+                p._popen.terminate()

+

+        for p in active_children():

+            info('calling join() for process %s', p.name)

+            p.join()

     debug('running the remaining "atexit" finalizers')

     _run_finalizers()