Implicit conversion from float -12.493261455525612 to int loses precision