[mod_python] Session problem

Graham Dumpleton wrote:
> Bouncing this back to the mailing list where it belongs so everyone can
> follow along.
> 
> Jakub Labath wrote ..
> 
> Hi,
> 
>> When you say cookies, are these cookies being constructed using Django's
>> infrastructure for creating them?
> 
> Correct it's the sessionid cookie created by Django.
> 
>> Further, is the request object a mod_python
>> request object or a Django request object wrapper?
> 
> In the previously attached log the request is Django request.
> And the modpython request is the actual modpython request.
> 
> What I found weird is that the same mod_python request serviced two
> different clients at different times, but perhaps that makes sense for
> preforked apache.

Just to clarify your terminology, it is the same mod_python *instance* 
servicing 2 different client requests. And no, it is not weird, but 
rather the whole point of mod_python. Well maybe not the *whole* point, 
but a very important one. :) In the simplest case you get one python 
interpreter per apache child process, thus avoiding the overhead of 
creating a new interpreter for each request.

>> At this point it sounds like it may be a Django issue if you are using it as
>> it merely uses mod_python as a jumping off point and doesn't tightly
>> integrate with mod_python and use its cookies or sessions and provides
>> users with their own request object wrapper as far as I know.
> 
> That was my first thought as well, and I reviewed our and django
> session handling the best I could. I found nothing unusual. In fact we
> have not touched the way django handles the sessions and cookies at
> all. I tried the django users group nobody seems to have similar
> problems, I even talked to one of the django's main devs he has no
> clue either.
> 
>> That said, as a start I would strongly recommend that you upgrade to
>> mod_python 3.2.10 and see if your problems go away.
> 
> I'll try that but I already tried upgrading from debian to gentoo and
> from python2.3 to python2.4 and I had the same problem on both. I know
> that would mean that it is something in our code ... but I haven't
> changed anything ... oh the headache.

Indeed.

A little light just went off in my head. Is it always *exactly* the same 
session id being served up for *all* requests?

There was a similar thread awhile ago where somebody was getting messed 
up md5sums and bad session id's with Django. IIRC it was some sort of 
library conflict between md5.so and libmhash.so. Take a look in the mail 
archives and see if you can find some help there:

http://www.modpython.org/pipermail/mod_python/2006-June/021482.html
http://www.modpython.org/pipermail/mod_python/2006-July/021544.html

Jim