Implicit conversion from float -41.19658119658121 to int loses precision