Implicit conversion from float -6.370656370656377 to int loses precision