The problem is intrinsic to floats, not just due to bad implementations.

Some values like 0.3 simply cannot be represented precisely with floats.