Incorrect result from Math.Round method

[vkvenkat]

The Math.Round method is expected to round a double-precision floating-point value to the nearest integral value based on the description in https://msdn.microsoft.com/en-us/library/wyk4d9cy(v=vs.110).aspx.
So, Math.Round(0.50000000000000011102230246251565404236316680908203) should return the nearest integer, 1. Instead it returns 0 on Windows and Ubuntu.

We end up calling the COMDouble::Round method in floatdouble.cpp:174, where the rounding happens.

Here is the repro. The issue is with rounding the 5th element (0.50000000000000011102230246251565404236316680908203) of the values array below:

using System;
using System.Runtime.InteropServices;

[StructLayout(LayoutKind.Explicit)] 
public struct SampleUnion
{
    [FieldOffset(0)] public ulong  ul;
    [FieldOffset(0)] public double d;
    public void Display()
    {
        Console.WriteLine("The stored value is: 0x{0:X}", ul);
        Console.WriteLine("Default rounding    {0} --> {1}", d, Math.Round(d));
        Console.WriteLine("Half to even        {0} --> {1}", d, Math.Round(d, 0, MidpointRounding.ToEven));
        Console.WriteLine("Half away from zero {0} --> {1}", d, Math.Round(d, 0, MidpointRounding.AwayFromZero));
    }
}

public class Example
{
   public static void Main()
   {
      double[] values = { -.5, -.50000000000000011102230246251565404236316680908203, -.50000000000000022204460492503130808472633361816406,
                          0.5, 0.50000000000000011102230246251565404236316680908203, 0.50000000000000022204460492503130808472633361816406,
                          1.5, 1.50000000000000022204460492503130808472633361816406,
                          2.5, 2.5000000000000004440892098500626161694526672363281 };
      SampleUnion su = new SampleUnion();

      foreach (double value in values) {
         su.d = value;
         su.Display();

      }
   }
}