homepage

multiprocessing.dummy crashes when attempting to create a ThreadPool from a Thread.
The following code will crush on 2.7.3:
 import multiprocessing.dummy
 import threading
 
 class Worker(threading.Thread):
 def __init__(self):
 threading.Thread.__init__(self)
 
 def run(self):
 poll = multiprocessing.dummy.Pool(5)
 print str(poll)
 
 w = Worker()
 w.start()
will crush with the following traceback:
 poll = multiprocessing.dummy.Pool(5)
 File "C:\Python27\lib\multiprocessing\dummy\__init__.py", line 150, in Pool
 return ThreadPool(processes, initializer, initargs)
 File "C:\Python27\lib\multiprocessing\pool.py", line 685, in __init__
 Pool.__init__(self, processes, initializer, initargs)
 File "C:\Python27\lib\multiprocessing\pool.py", line 136, in __init__
 self._repopulate_pool()
 File "C:\Python27\lib\multiprocessing\pool.py", line 199, in _repopulate_pool
 w.start()
 File "C:\Python27\lib\multiprocessing\dummy\__init__.py", line 73, in start
 self._parent._children[self] = None
AttributeError: 'Worker' object has no attribute '_children'
If you have access to the thread itself, you can set the _children attribute youself (w._children = weakref.WeakKeyDictionary()), but it is not possible with threads different from threading.thread (Such as PyQt4's QThread thread, which is essential for GUI programming).
The fix that I found is to edit the Python27\Lib\multiprocessing\dummy\__init__.py file. The crashing code is line number 73.
This line should be nested in an if block:
 if hasattr(self._parent, '_children'):
 self._parent._children[self] = None
That way the code is fixed.
Thanks!

New changeset 1f5d2642929a by Richard Oudkerk in branch '2.7':
Issue #14881: Allow normal non-main thread to spawn a dummy process
http://hg.python.org/cpython/rev/1f5d2642929a
New changeset 0528ec18e230 by Richard Oudkerk in branch '3.2':
Issue #14881: Allow normal non-main thread to spawn a dummy process
http://hg.python.org/cpython/rev/0528ec18e230 

3.3 commit is http://hg.python.org/cpython/rev/9373ca8c6c55
Richard, it did not record here because you just said 'Merge' rather than "#14881 merge" ;-).
Great to see a crasher fixed. Ready to close?

I'll, remember that in future;-)
Closing.

Is this a duplicate for http://bugs.python.org/issue10015
#10015 ?

Since this is already fixed and closed, the question is more relevant to #10015 and whether it should be closed. The answer seems to be yes.

By the way, thanks for noticing, so it *can* be closed.