When the following is function called with "x = test(2147483648.0);" it will 
cause a floating point exception when trying to do the type conversion from long
to double.  

long test(double dvar)
{
    long lvar;

    lvar = (long)dvar;

    return(lvar);
}

I could be mistaken, but isn't that a very sensible reaction?
2^31 does not have a representation as a (long).

So, what do you think should happen in this situation?  And why?

Upgrade to v2.02, where FP exceptions are masked by default.  You
will get a NaN for this case.