In both of those cases the problem is not problem with floating point arithmetic but the parsing of floating point literals, or rather how systems take input and then substitute it with something different quietly behind the scenes. None of 0.1, 0.2, 0.3 are true (binary) fp values. Nobody would be surprised that

    0.1000000000000000056 + 0.2000000000000000111 != 0.2999999999999999889