(j3.2006) (SC22WG5.5301) [ukfortran] Did we intend to prohibit this?

[Malcolm Cohen]

>  integer, parameter :: RK = kind(1.0e0)
>  real(rk), parameter :: SNaN = merge( IEEE_Value(1.0_rk,IEEE_Signaling_NaN), &
>                                     & 0.0_rk, IEEE_Support_DataType(1.0_rk) )

This would in any case not be conforming when either IEEE_SUPPORT_DATATYPE is 
false or IEEE_SUPPORT_NAN is false, since MERGE is not "short-cut" evaluation.

There are many traps and pitfalls associated with trying to use signalling NaN 
in user code, especially in code that might be evaluated at compile-time.  I 
would not recommend embarking on such a journey.

Cheers,
-- 
................................Malcolm Cohen, Nihon NAG, Tokyo.