Re: Re: corruption

Jeff Hodges wrote:
> Its a small unknown distribution.  It's a src based one.  If you need
> any other versions look here:
> http://www.lunar-linux.org/pages.php?node=02/03/28/8425753

Jeff, let's check my theory out for certain, to see if I'm right or 
wrong ;-)

Open your "metakit.py" file in your Python lib directory (on my system, 
/usr/lib/python2.3/site-packages/metakit.py) and check that the version 
matches the binary of the installed version of the metakit library 
itself. The first few (non-documentation) lines of my metakit.py file are:

_oldname = __name__
__name__ = "metakit"
__version__ = "2.4.9.3"
__description__ = "Python bindings to the Metakit database library"
__author__ = "Gordon McMillan / Jean-Claude Wippler"
__email__ = "[email protected]"
__url__ = "http://www.equi4.com/metakit/python.html"
__license__ = "X/MIT style, see: http://www.equi4.com/mklicense.html"
from Mk4py import *
import string


Unfortunately, since we are *both* on source distributions, it is extra 
useless to compare MD5 hashes of the metakit libraries. (The metakit.py 
file md5sum is 7f31ae8b110d8a2e210f61c5716fd91a, but that's not likely 
to help even if it's different for you.)

Generally, for a source distribution (I'm on Gentoo) when you have 
problems like this, it can be because of over-aggresive optimizations by 
the system. Since GCC 3.x is so good, I find it is now not worth it to 
be aggressive, and just use "-march=X -O2 -pipe"; basically, "I have an 
X computer, and I want fast-but-safe binaries, use your judgement." If 
your settings are more aggresive then that, try turning them down and 
recompiling metakit, the Python metakit interface, and potentially 
anything metakit depends on.

I have, on occasion, had to recompile *everything* with less aggresive 
optimizations to make things like Mozilla work. Fortunately, -O2 is 
pretty good on its own, now.