Implicit conversion from float -67.52895752895753 to int loses precision