Implicit conversion from float -0.41782729805012764 to int loses precision