Implicit conversion from float -8.069498069498081 to int loses precision