Andrew Haley wrote: >F9.4.4 requires pow (x, 0) to return 1 for any x, even NaN. > > Indeed. My point, basically, is that consistency appear to require the very same behavior for *complex* zero^zero. Paolo.