Implicit conversion from float -6.976744186046517 to int loses precision