Implicit conversion from float -4.103773584905667 to int loses precision