PyApache memory leak?

Oleg Broytmann phd at phd.russ.ru
Thu Apr 27 14:56:04 CEST 2000


Hello!

On Thu, 27 Apr 2000, Gregory Trubetskoy wrote:
> On Fri, 21 Apr 2000, Oleg Broytmann wrote:
> 
> >    This don't help saving memory. Other way round - PyApache destroys
> > subinterpreter after processing a request, thus saving memory; but httpdapy
> > keeps the interpreter with all (unneccessary and superfluos) imported
> > modules, thus leaking memory.
> 
> I wouldn't call this "leaking". "Leaking" means forgetting to free memory
> where you should be freeing it (like overlooking a Py_DECREF), loosing a
> pointer to memory, etc. To the best of my knowledge, httpdapy 1.7b does
> not leak memory. I've done a lot of work to make sure of this.
> 
> Another note to consider - I am not 100% sure that Py_EndInterpreter()
> actually frees all memory. In fact I'm almost sure it doesn't, and you
> still need to watch your reference counts if you create object via the
> C-API. If that is the case, then PyApache has a higher likelyhood of
> actually leaking memory.

   May be...

> BTW, if I remember it correctly, the decision by Lele (the author of
> PyApache) to recreate the interpreter was based not on memory usage, but
> because the CGI wouldn't work as expected otherwise and for security
> reasons.

   I am pretty satisfied with PyApache. (Note: I didn't tried httpdapy yet,
so I cannot say bad words about it; I belive it's good product anyway).

Oleg.            (All opinions are mine and not of my employer)
---- 
    Oleg Broytmann      Foundation for Effective Policies      phd at phd.russ.ru
           Programmers don't die, they just GOSUB without RETURN.





More information about the Python-list mailing list