/* log1p(x) = log(1+x). The log1p function is designed to avoid the significant loss of precision that arises from direct evaluation when x is small. Use the substitute from _math.h on all platforms: it includes workarounds for buggy handling of zeros. */ static double _Py_log1p(double x) { … } #define m_log1p …