Hi!
It looks like you're leaking requests. This is often the consequence of setting an attribute on the REQUEST object which is aquisition wrapped. For instance:
REQUEST.adtmldoc = REQUEST['PARENTS'][0]
When requests leak, references to all the objects referenced within the request (acquisition wrappers, etc.) leak as well. This can be disastrous and gets worse when the site is busy (when there are more requests).
ok, that's what I experienced.. the day the site got some more traffic it also suddenly grow memorywise. I also didn't know that doing above is a bad thing to do.. (though I don't actually understand why this is a problem and why it's not freed then.. but on the other hand I am not that much into the inner workings of acquisition ;-)
It used to be the case that tracking this problem down was tediously difficult, and involved basically a binary search for leaks in all methods on the site. But now, finding the cause is likely a job for Shane's LeakFinder product: http://www.zope.org/Members/hathawsh/LeakFinder .
Ah, will try that out :) And actually I directly found some method which fetches the actual user object from a ZPatterns rack and sets this in the REQUEST at every request. So maybe that's already the problem then.. will try to remove that..
Note that this is likely caused by some sort of brainbending circular reference problem that might be helped by Python 2.X's cyclic garbage collector.
Well, will also try out python 2.x then... (btw. are all 2.x python versions working without problems with Zope? Don't want to get a new problem when resolving the old one ;-) Thanks! Christian