Implicit conversion from float -1.2556053811659211 to int loses precision