During finalize, orphaned tasklets may be killed. And for them to be killed, they have to be woken up. Something is breaking during this process. gc is very dangerous territory. I have seen this too but not been able to put my finger on it exactly. K
> -----Original Message----- > From: [email protected] [mailto:stackless- > [email protected]] On Behalf Of Jeff Senn > Sent: 11. mars 2010 13:49 > To: The Stackless Python Mailing List > Subject: Re: [Stackless] Segmentation fault when running example wsgi > app in linux/win > > I had this problem too (embedding Python in a completely different way > and on OS/X), > that is: a crash in GC during Py_Finalize. I spent just a bit of time > looking > but didn't resolve much -- my (not very satisfying) solution was *not* > to call > Py_Finalize since my program was exiting anyway. > > I seem to recall it was "worse" on a 64-bit compile... > Håkan, do you have the same problem with a 32-bit compile? > maybe that's a clue... (or a red herring...) > > Sorry, not much light shedding, let me know if you figure it out... > > On Mar 11, 2010, at 4:55 AM, Håkan Larsson wrote: > > > Hello all, > > > > I have been trying out the wsgi server with the app_sessionless > example downloaded from the google stacklessexamples page and i keep > getting a assertion and segmentation fault. I have browsed through the > mailing list archives and there seems to be a similar error posted > about a year ago. http://www.stackless.com/pipermail/stackless/2009- > February/003910.html > > > > I have the same symptoms as jtunney describes. Works great when you > do a single request in your web browser but when you stress it a bit > the server simply exits. When running this with a debug version of > python under windows i also can see a assertion before the server > exits. When stressing it even further under windows i get a > segmentation fault. > > Assertion failed: stack_pointer != NULL, file ..\Python\ceval.c, line > 914. > > > > This assertion/crash only happens when the clients are about to > disconnect. To reproduce this i used jmeter, 10 looping threads doing a > request against /. When starting the test i can see 10 "Accepted > connection" print outs and the server serves request. But when i shut > down the threads a assertion/segmentation fault occures. > > > > I have also been able to reproduce this under linux and there i get a > segmentation fault right away. > > I have attached some output from gdb, one with standard optimizations > and one with the flag turned to -O0 for more information. > > > > I am running the following setup. > > > > Windows: > > Win 7 x64, (same thing on Vista 64) > > > > Linux: > > Redhat 5 x64 > > Python 2.6.4 Stackless 3.1b3 060516 (release26-maint:78823M, Mar 11 > 2010, 10:10:31) > > [GCC 4.1.2 20080704 (Red Hat 4.1.2-46)] on linux2 > > > > python2.6: ELF 64-bit LSB executable, AMD x86-64, version 1 (SYSV), > for GNU/Linux 2.6.9, dynamically linked (uses shared libs), for > GNU/Linux 2.6.9, not stripped > > > > Perhaps someone can help shed some light on this one. > > > > /Håkan > > > > <debug 00.txt><debug > 03.txt>_______________________________________________ > > Stackless mailing list > > [email protected] > > http://www.stackless.com/mailman/listinfo/stackless > > > _______________________________________________ > Stackless mailing list > [email protected] > http://www.stackless.com/mailman/listinfo/stackless _______________________________________________ Stackless mailing list [email protected] http://www.stackless.com/mailman/listinfo/stackless
