susan at mars.provis.com
Fri Mar 29 21:13:14 CET 2002
I thought I'd end the suspense and let you all know how
it turned out. Just as you no doubt suspected, it was
the 3rd party module walking on memory. We've had this
code running on four platforms for considerable time
without it causing a problem, but it definitely had
a memory error in it.
So, so sorry to cast aspersions on python, all is better.
(I hope I don't have to join a 12-step program if I
admit to using this, but the module was typhoon
in case any of you are using it... And as I mentioned,
I haven't been paying attention, so if Aaron Watters
has written a database module in python I'd love to
hear about it... ;-)
>[No one will remember me, but I'm back after a long
>absense because--you guessed it--I have a problem and
>I want some help...]
>We are attempting to port from Solaris to Linux, and
>I'm having some unpleasant malloc/free problems at
>exit (sys.exit). We have running software based
>currently on 1.5.2. We're compiling on Linux with gcc
>3.0.3 (on Solaris we were using Sun's compiler, not
>gcc, I think).
>Unfortunately I have a 3rd party module I'm linking with
>python that is rather ancient. We first got it when using
>python 1.3, we "ported" it to 1.5.2, and I can't quite
>pin down whether it is the root of the problem or not,
>perhaps it is doing something not quite kosher.
>(The idea of trying to get it running with 2.1 gives
>me the willies, but that is not really today's issue.)
>Anyone out there using Linux especially with 1.5.2
>is this ringing any bells? On exit, python is clearing
>the list of modules, each module is clearing dictionary
>entries by setting them to None, and it is the decref
>that is blowing up. It is not always in the same place,
>but it is always in the code that is setting a dictionary
>value to None and deleting whatever was there before.
>Any ideas welcome. I'm not too well versed on either the
>innards of python or the proper coding for module
More information about the Python-list