> Patches should only address one issue. The visit cookie should not be > modified in the same patch as a hack to get around a flaw in SQLObject.
Yep, sorry, I just diffed in a hurry and had the visit key already hacked so the patch included both hacks. I've uploaded to the Trac two separate patches, one for the key and one for sqlobject (In the same ticket, though.. If the Trac is up I'll create a separate ticket for the key thingie when I finish writing this) I've posted the sqlobject patch at sqlobject's sourceforge tracker too. > Have you checked whether this threading problem is present without the > reloader thread? > Reloader thread?? Do you mean: a) the template reloader? (I'm too lazy/busy to check right now in the code if it's threaded or something. b) the visit-expiration-time-updater-thread? c) cherrypy's module reloader? d) none of the above. If the answer is (a), the yes as I'm running with kid.precompiled=True as to work around the "None not callable" bug on a semi-production site. If you mean (b) then no, as I don't know how to "turn it off". If you mean (c) then yes as it's on "production" environment (I think cherrypy's autoreloader thread is off in "production", am I wrong?). If the answer is d, the I don't know what you mean :) Anyway, I'm pretty sure it's an sqlobject's bug as Randall Smith described the exact same problem before at http://pythonpaste.org/archives/message/20051202.225107.c9697187.en.html (it's where I "stole" the hack from) I'm pretty sure it's not an identity bug anyway, just what made the bug arise so this probably shouldn't have spilled to your side. It's the first thing I saw on the traceback though, and I posted way before investigating... my apologies. Cheers, Alberto
