[Zope] AssertionError after update to Zope 2.8.1

[Tim Peters]

[Ricardo Newbery]
> Oh so close...
>
> Migrated from Zope 2.7.2 to 2.8.1
> Updated all the Catalog instances according to the instructions.
> Fixed a few outdated products.
>
> Then just for kicks, I tried to update the Python scripts by visiting
> the /manage_addProduct/PythonScripts/recompile url.  After about a
> minute, it threw up an error page.  This is what shows up in the
> log...
>
>
> 2005-08-21T01:56:45 ERROR txn.170046464 Error in abort() on manager
> <Connection at 0956724c>
> Traceback (most recent call last):
>   File "/Zope-2.8/lib/python/transaction/_transaction.py", line 456,
> in _cleanup
>   File "/Zope-2.8/lib/python/ZODB/Connection.py", line 348, in abort
>   File "/Zope-2.8/lib/python/ZODB/Connection.py", line 360, in _abort
> AssertionError
> ------
> 2005-08-21T01:56:45 ERROR Zope.SiteErrorLog
> http://someurl.com/manage_addProduct/PythonScripts/recompile
> Traceback (most recent call last):
>   File "/Zope-2.8/lib/python/ZPublisher/Publish.py", line 119, in publish
>   File "/Zope-2.8/lib/python/Zope2/App/startup.py", line 215, in commit
>   File "/Zope-2.8/lib/python/transaction/_manager.py", line 84, in commit
>   File "/Zope-2.8/lib/python/transaction/_transaction.py", line 381, in commit
>   File "/Zope-2.8/lib/python/transaction/_transaction.py", line 379, in commit
>   File "/Zope-2.8/lib/python/transaction/_transaction.py", line 424,
> in _commitResources
>   File "/Zope-2.8/lib/python/ZODB/Connection.py", line 462, in commit
>   File "/Zope-2.8/lib/python/ZODB/Connection.py", line 483, in _commit
> AssertionError

They both look like "the same" error:  in both cases, a Connection is
marching over its list of registered (modified) objects (to abort the
current transaction in the first case, and to commit the current
transaction in the second case), and in both cases it finds an object
in the list that has never been assigned an oid (object id). 
Concretely, in the first case, this is the assert that's failing:

        for obj in self._registered_objects:
            oid = obj._p_oid
            assert oid is not None  # HERE

The second case is similar.  Of course asserts should never trigger. 
It suggests that something, somewhere, is "cleverly" creating new
objects in a flawed away, and they're getting caught.

> Nothing appears to be broken yet but I'm wondering whether I should
> roll back to my backup.  Anyone have any ideas about this error?  I'm
> tempted to just pretend that I didn't see it.

That's up to you, but wishing away assert failures isn't recommended
practice ;-)

If you can find a finite way to reproduce it, please open a Zope bug report.