Implicit conversion from float -0.34802784222739547 to int loses precision