Implicit conversion from float -49.554317548746496 to int loses precision