Implicit conversion from float -253.51145038167942 to int loses precision