Implicit conversion from float -7.816901408450718 to int loses precision