Re: Floating point exception associated with python

[Bob Kimble <bob@iplicity.com>]

On Friday 04 July 2003 14:30, Kelledin wrote:
> On Friday 04 July 2003 11:31 am, Bob Kimble wrote:
> > I'm running unstable on an AlphaPC and I'm having problems
> > installing and upgrading anything related to python. All the
> > packages now break during configuration, reporting a floating
> > point exception. Here is an example message from when I tried
> > to reinstall python2.2:
> >
> > Setting up python2.2 (2.2.3-2.1) ...
> > Compiling python modules in /usr/lib/python2.2 ...
> > /var/lib/dpkg/info/python2.2.postinst: line 49: 20390 Floating
> > point exception/usr/bin/python2.2
> > /usr/lib/python2.2/compileall.py -q $i dpkg: error processing
> > python2.2 (--configure):
> >  subprocess post-installation script returned error exit
> > status 136
>
> I'm guessing the SIGFPE is from something that wasn't compiled
> with -mieee (and probably ought to be, or maybe needs a source
> patch).
>
> > I even tried reinstalling python2.2 from source with the same
> > result. My guess is that the problem is not with python per
> > se, but with some library somewhere. However, I'm pretty much
> > clueless on how to proceed. Any suggestions?
>
> You could try compiling python from source with -mieee in your
> CFLAGS.  That won't fix it if the SIGFPE comes from an external
> library though.  If it doesn't fix it, you can use gdb to get a
> backtrace.  Even with all debug symbols stripped away, gdb
> should at least be able to tell you what library the SIGFPE
> occurred in.

What I ended up doing was uninstalling python, which forced the removal of a 
whole bunch of other packages. I then reinstalled python and most of the 
other stuff and now everything is working just fine. There were updates to 
unstable while I was doing all this, so it's possible that something in there 
got fixed. Thanks for the suggestion.

> This SIGFPE/-mieee thing has actually been a running debate for a
> while now...I don't know what Debian's current plans are for
> handling it.
>
> --
> Kelledin
> "If a server crashes in a server farm and no one pings it, does
> it still cost four figures to fix?"