Implicit conversion from float -12.519685039370074 to int loses precision