Implicit conversion from float -1.2532637075718043 to int loses precision