[Zope-Coders] Linux Test Results - HEAD/Py2.2 - Not OK :-(

[Tim Peters]

[Fred]
> ...
> There will be a 2.3.3 release, likely in December, but that's only
> because Jim decided to tickle a very tedious corner case in Zope 3.

Thomas Heller stumbled into a distinct segfault-with-weakref bug, which
would have been enough to force 2.3.3 on its own.  I'm amazed more people
haven't stumbled into that one.  Like the other bug Jim stumbled into, it's
been there as long weakrefs and new-style classes have been in Python (i.e.,
both these bugs are in Python 2.2.n too -- nothing unique to 2.3 about
them).  But Thomas's bug is much easier to provoke; Jim probably wouldn't
have seen his problem at all if Py_Finalize in 2.3 had not also been changed
to try collecting cyclic trash after modules got torn down.  That change to
Py_Finalize has been a catalyst in provoking other symptoms too (like the
long chain of information-free error msgs we used to see after running the
Zope3 test suite under 2.3).

> The cost of maintaining Zope 2.x for more than one version of Python
> for each branch is excessive, and does more to contribute to bugginess
> and cost for Zope Corporation than just fixing problems that are
> found.

Fully agreed there, and anyone who wants free support for any Python before
2.3 now is flat out of luck (for example, nobody has volunteered to backport
fixes for the two critical weakref segfault bugs beyond the 2.3 maintenance
branch, and I don't expect that anyone will).