complex pow fails with boost::float128_t

[gpeterhoff]

Hi,
std::pow(std::complex<boost::float128_t>, boost::float128_t>) works not with negative real and zero imag, eg:

template <typename Type>
void show_complex_pow(const Type r, const Type i, const Type e)
{
 std::complex<Type> c(r, i);
 c = std::pow(c, e);
 std::cout << c << std::endl;
}
const double
 r = -0.7,
 i = 0.0,
 e = 2.3;
show_complex_pow<double>(r, i, e);
show_complex_pow<boost::float128_t>(r, i, e);

(+0.258788,+0.356191)
(-nan,+0)

[mborland]

Are you using GCC on x86-64 so boost::float128_t is a typedef of __float128? Does it work if you call cpowq instead of pow?

[gpeterhoff]

yes + yes

__complex128 c = r + i;
 c= cpowq(c, e);
 std::cout << '(' <<crealq(c) << ','<< cimagq(c) << ")\n";

(+0.258788,+0.356191)
The problem is the overload in boost/math/cstdfloat/cstdfloat_complex_std.hpp