[ZODB-Dev] Tracking down a freeze (deadlock?)

[Tim Peters]

[Dieter]
>> I think both of them should get a warning: do not use "__del__".

[Jim]
> Agreed. Let's smeone add this to the rules of persistence.

As later msgs in this thread explained, a warning against using __del__ was
added to the ZODB/ZEO Programming Guide 14 months ago, but only to the ZODB
3.3 version of the docs.  That _is_ the version visible on zope.org, though.
Or are "the rules of persistence" documented in some place other than that
(if so, where?)?

I'm currently:   writing more words about that; backporting the __del__
verbiage to the 3.2 line (so the PDF we ship with ZODB 3.2.6 will contain
it); updating the generated HTML on zope.org with the new words.

> BTW, I ran into a similar problem when working on fixing ZClassesd for
> 2.8.  I change the way invalidations are handled in a way that will
> probably fix this problem. :)

Unclear which problem "this problem" means.  Changing the way invalidations
work can't prevent all __del__-caused problems.  We still don't have a test
case that provokes Florent's deadlock problem, BTW (and I still don't intend
to spend time trying to create one).