The following example gives a float rounding error on 1.1 (and 1.1.1B)
but not on previous versions.

It IS only a rounding error, but Licensee (IBM) is concerned
that anything as fundamental as arithmetic should apparently have
regressed.

import java.io.*;
import java.lang.*;

public class IOTest
{
    public static void main (String[] args)
    {
        if (args.length == 0)
        {
            System.out.println ("There is [" +
                args.length + "] arguments:");
            System.exit (1);
        }

        for (int i= 0; i < args.length; i++)
        {
            System.out.println ("Input String: <" +
                args[i] + ">");
        }

        float output = new Float(args[0]).floatValue();
        System.out.println ("Before addition: <" + output + ">");
        output += 1.011;
        System.out.println ("After adding 1.011: <" + output + ">");
    }
}

Results from JDK1.0.2:

java IOTest 445.348
Input String: <445.348>
Before addition: <445.348>
After adding 1.011: <446.359>

Results from JDK1.1:

java IOTest 445.348
Input String: <445.348>
Before addition: <445.348>
After adding 1.011: <446.35898>