After some experimenting it seems as if it is not possible to use the
"a == a" trick to detect NaN. So I found no other way than to use
a < 3.0e+36 in a NaNtoInvalid macro. -Inf will not be caught and
if there already are invalid positions in the field unexpected things
are probably going to happen. Interestingly enough it seems as if
the constant oin the expression above needs to be a float and can not be 
a double.

The macro is available as a net at 
http://www.pdc.kth.se/%7esmeds/DX/NaNtoInvalid.net

/Nils

Reply via email to