Implicit conversion from float -12.324649298597194 to int loses precision