http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57749
--- Comment #9 from Vittorio Zecca <zeccav at gmail dot com> --- Yes, I agree that there is a bug, and IMO it is in cpow/cpowf/cpowl. With -ffpe-trap=invalid,zero, as I wrote earlier, complex zero**I where I is integer equal to one, does not raise any exception and delivers the expected (by me) result, zero. Complex zero**X where X is real equal to one raises exception. Complex zero**C where C is complex equal to (1e0,0e0) raises exception. This is because glibc computes x**y as exp(y*log(x)) so it fails when x=0, but IMO it should not compute log(x) if x=0 and y>0 (y real), it should take a shortcut instead and deliver complex zero.
