[issue3270] test_multiprocessing: test_listener_client flakiness

[Jesse Noller]

Jesse Noller <jnoller at gmail.com> added the comment:

Trent/Antoine - I'm stuck on the fence about this. Per trent's own 
suggestion - removing the allowance for the 0.0.0.0 style address means 
that the self._address is always a connectable end-point: this provides 
clarity to the API.

However - to Antoine's point - using 0.0.0.0 to listen on "all 
addresses" is actually pretty common, especially when working with 
bigger servers (which generally have multiple cores). Using the 0.0.0.0 
address makes it easy to say "listen everywhere" - where, in the case of 
a Manager is actually very useful (you could have processes connecting 
to a Manager on a machine from different networks).

Attached is a cleaned up diff of the removal of the fqdn call - this 
should resolve the original problem while leaving the door open for the 
ability to connect to the 0.0.0.0 "address".

I need someone with a windows machine (Hi Trent!) that exposed the 
original problem to test the patch. Currently I'm favoring this rather 
than locking out the 0.0.0.0 option.

Added file: http://bugs.python.org/file11088/connection_v2.patch

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue3270>
_______________________________________