there’s two fields in java.lang.Math that are not final, but they really should be:

    // Use raw bit-wise conversions on guaranteed non-NaN arguments.
    private static long negativeZeroFloatBits = Float.floatToRawIntBits(-0.0f);
    private static long negativeZeroDoubleBits = Double.doubleToRawLongBits(-0.0d);

since they’re used in Math.min/max, these non-final fields do have performance implications.