[Zope-Coders] Python 2.2 and Zope 2.6

[Guido van Rossum]

> > But after Matt Behrens is done with his new setup.py we won't need
> > Makefile.pre.in any more right?  Already (with the feeble setup.py
> > we've got now) I can build everything for 2.2.
> 
> Hold the phone, has anyone solved collector #200
> <http://collector.zope.org/Zope/200> yet?  On Solaris and OpenBSD,
> Zope 2.5.0 was crashing repeatedly on startup with Python 2.2, but
> not Python 2.1.  I believe the consensus was that there was some
> EC issues with 2.2.  If Shane has a branch that solves this, PLEASE
> point me in that direction...

Have you tried this on Linux?  If it doesn't occur on Linux, but does
occur on Solaris and *BSD, I strongly expect it's writing a malloc'ed
block out of bounds -- I know BSD's malloc is much more picky.  Could
you try with 2.2.1?  It fixes some bugs.

More work, but perhaps leading to the cause, could you try with "2.3"
(== current CVS), pymalloc enabled (which it is by default in 2.3), in
debug mode, and with the symbol PYMALLOC_DEBUG defined manually?  That
enables heavy malloc debugging which often points to the culprit.

I'm cc'ing Tim Peters because pymalloc-debug is his baby and he can
provide much better info on how to enable.  Alternatively, the
comments in obmalloc.c may lead the way.

(It may not work in debug mode, but you can use PYMALLOC_DEBUG even
outside debug mode.  I vaguely remember that I couldn't get 3rd party
extension linkage working in debug mode, but I'm not sure about this.)

--Guido van Rossum (home page: http://www.python.org/~guido/)