win32 Zope 2.5.0 vs python2.1&win32all

[Robin Becker]

In article <3C6BB8B5.8020903 at mxm.dk>, Max M <maxm at mxm.dk> writes
>Robin Becker wrote:
>
>> I am having bad problems with zope not starting up properly. I think
>> this is caused by having Python-2.1.2 + win32all installed. It seems
>> that paths set up in the registry by win32all are getting into sys.path
>> when zope runs. This is causing massive confusion between modules
>> installed in the base python and those I wish to test with zope.
>
>I have several versions of Zope and Python + win32all installed and has 
>never had any problems.
>

well I think you just have to look at the path printed out to see that something is wrong. There should
logically be no path elements starting without the prefix c:\Python\ZOPE_2_5_0 as that is my Zope home. The
other path elements creep in because of the win32all registry paths.

I'm not doing anything here except running the python.exe that came with zope. If the path is wrong in
zopexxx\bin\python.exe there's not much I can do about it except to try and eliminate path elements in z2.py
which I'm not yet doing.

When I run zope it's via the standard service mechanism. Using procexp I can see mixed pyds if I don't
eliminate the registry path.

Certainly in the past I had different versions of python from zope and installed and they coexisted.

It seems to me that python should allow a registry free startup somehow and then the pythonservice could
pass that flag to the service.  


My feeling is that in z2.py there ought to be something like

sys.path = filter(lambda p,swhome=swhome: not p or swhome.lower()==p[0:len(swhome)].lower(),sys.path)
but it seems hard to guarantee that we get the right sys unless we mess with imp et al.

>Are you starting zope via your own script or via z2.py ??
>
....
>So that version runs on another port.
>
>All pretty painless.
>
>regards Max M
>

-- 
Robin Becker