blais at furius.ca
Tue Jan 3 16:34:37 EST 2006
On 1/3/06, Martin Blais <blais at furius.ca> wrote: > Hi Graham > > I applied the patch at the link you mention, and it does not resolve > the issue. The problem still occurs. > > I read the discussion, and I gather from it that there is an issue > with using multiple interpreters. This is my case: I have two > separate web apps using a python handler with different names. I > suppose I don't have a choice, right? I couldn't specify two > different apps in the same interpreter, if I have some globals in some > of the common imported modules? I also tried now with a single interpreter, the problem is still there, named explicitly or not. After much hacking (all day today), I added the explicit Py_EndInterpreter calls and it still hangs on destruction. Specifically, it hangs on PyImport_Cleanup() in Py_EndInterpreter(). I'm having trouble pinpointing exactly which symbol makes it hang. It seems to happen when cleaning the logging module. I saw that logging register an atexit hook, tried to remove that, does not fix the problem. Patch on the latest beta (3.2.5b) to come at some point. I need to rip out my horrible tracing code that I inserted all over the place. Question: Why do we need to create a new thread state on every handler call? Is this for MPM worker support?