As documented, the id() function can return the same value for distinct
objects with non-overlapping lifetimes which can result in signals being
sent to the wrong receivers if two distinct senders happen to have a
colliding id() value.
Since reproduction of the issue requires memory constrained
circumstances where the same exact id() is reused for two senders of the
same signal the test opt to simulate the collision by systematically
making the same id for Sender instances.
Note that we explicitly avoid keeping a strong reference to senders that
cannot be weakly referenced as that would unexpectedly prevent them from
being garbage collected. This means that id(sender) collisions could
still occur for such objects but Django itself doesn't make use of them.
Thanks Sjoerd Job Postmus for the reduced test case and Mariusz for the
review.
Co-authored-by: And Clover <and@doxdesk.com>
Exceptions from the (receiver, exception) tuples returned by
``send_robust()`` now have always their traceback attached as their
``__traceback__`` argument.
Make use of `weakref.finalize` and `weakref.WeakMethod` on python 3.4.
Simplified the removal of receivers, the old function looked overly
complicated.
Many thanks go to Antoine Pitrou for helping me to debug and explain all
the failures I ran into while writing that patch.
Added comments in the three empty models.py files that are still needed.
Adjusted the test runner to add applications corresponding to test
labels to INSTALLED_APPS even when they don't have a models module.
