__builtin_cpow((0,0),(0,0))

[Paolo Carlini]

Hi everyone,

why the result is (0,0) instead of (1,0)?!? It seems to me that only the 
latter is consistent with the C99 requirements for real power (F.9.4.4) 
- that is 1.0 - and, on the other hand, according to G.6.4.1 and note 
318, cpow behavior in "special cases" is rather implementation dependent...

I'm asking because, in the C++ library we are trying to make consistent 
use of both pow and __builtin_cpow in the implementation of 
complex::pow, and as soon as you deal with "special cases", issues like 
the above surface quickly.

Thanks,
Paolo.