Implicit conversion from float -9.902912621359235 to int loses precision