Well done. That did the trick. The memory leak issue is not observed anymore (applies to uWSGI 1.3 rev. 2603 on python 3.3).
Thanks. Andriy Kornatskyy ---------------------------------------- > From: [email protected] > Date: Mon, 1 Oct 2012 19:53:35 +0200 > To: [email protected] > Subject: Re: [uWSGI] Memory Leak: uWSGI 1.3 + Python3.3 > > > Il giorno 01/ott/2012, alle ore 15:17, Andriy Kornatskyy > <[email protected]> ha scritto: > >> >> Hm... again, a link to a simple wsgi application is in the original report, >> just in case here it is again. >> >> https://bitbucket.org/akorn/helloworld/src/tip/wsgi/app.py >> >> Yes, the string is converted to bytes. > > > > Looks like this leak is present in all of the uWSGI versions. > > Basically when uWSGI encodes headers to latin1, lose a reference after the > translation > > I have made an hack (honestly i see no other solutions without impacting > memory usage and performance): > > a python list (wsgi_req->gc_tracker) maintains a list of all of the > translated headers and destroy them after sending headers. > > This is obviously enabled only for python3 > > Let me know if it works > -- > Roberto De Ioris > http://unbit.it > JID: [email protected] > > _______________________________________________ > uWSGI mailing list > [email protected] > http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
