Please let us know. It may or may not be related. On Wednesday, 10 October 2012 19:29:01 UTC-5, rochacbruno wrote: > > Great! Maybe this will solve the problems I reported eith uWSGI. I will > checkout and test tonight! > > Thanks again! > > > On Wed, Oct 10, 2012 at 2:34 PM, Massimo Di Pierro > <massimo....@gmail.com<javascript:> > > wrote: > >> Once in a while we have had reports of memory leaks. I and others could >> never reproduce it. Today I managed to reproduce it. >> >> On a Mac Pro I was able to reproduce a memory leak of 2000Kb/request with >> python 2.5, 2.7 and all previous version of web2py. >> >> What is really odd is that some of us have run the same tests on other >> platforms (including other macs) and have not observed the problem. >> I have run the same tests myself before and I have not seen it before. >> I was able to determine that the problem was caused by cache.disk which >> is initialized whether you want it or not and had an overly complicated >> logic mechanism. >> I do not believe the memory leak was caused by the lock itself but, to >> the best of what I could determine, by a circular reference in the shelve >> object itself. >> >> Anyway. THIS IS NOW FIXED FOR GOOD! :-) >> >> First of all the circular reference has been removed and the logic >> simplified. The problem disappeared. >> Moreover the creating of cache.ram and cache.disk is now lazy so they are >> only initialized when used. This resulted in a 14% speedup in a typical >> hello world app. >> >> If you are using cache extensively you may want to check out trunk and/or >> nightly built. Please report any problem. >> >> Massimo >> >> >> >> >> >> >> >> -- >> >> >> >> > > >
--
