Trapping NaNs

[Thomas Weber]

Hi, 

this is probably off-topic, but on the other hand, people here are used
to dealing with NaNs, so I ask here first. 

I'm currently trying to figure out Jordi's problem with Octaviz, as
discussed at
http://www.cae.wisc.edu/pipermail/octave-maintainers/2008-May/007236.html

This bug happens only on amd64 (x86_64) and gdb tells me that some
values become NaN there. Now, I would like to find out which variable
becomes NaN for the very first time; I don't know which variable it is
though, as these values are passed throught different functions and VTK
has a gazillion of structures. 

Is there some way of catching this, ie to tell the program to abort as
soon as whatever value is set to NaN?

Thanks
        Thomas