Implicit conversion from float -91.58890290037829 to int loses precision