Implicit conversion from float -32.34468937875752 to int loses precision