Implicit conversion from float -12.217741935483872 to int loses precision