Implicit conversion from float -35.976562499999986 to int loses precision