Implicit conversion from float -6.932114882506525 to int loses precision