Unexpected GreaterThan(NaN) behavior

[shimat]

Hi,

even though 0.1 > double.NaN returns False, ClassicAssert.GreaterThan(0.1, double.NaN) passes. Of course, the behavior is similar for any value other than 0.1. Is this the intended behavior?

NUnit 4.1.0 / .NET 8

public class Tests
{
    // Fail
    [Test]
    public void DirectGreaterThan() => ClassicAssert.IsTrue(0.1 > double.NaN);

    // Fail
    [Test]
    public void DirectLessThan() => ClassicAssert.IsTrue(0.1 < double.NaN);

    // Pass (???)
    [Test]
    public void NUnitGreaterThan()
    {
        ClassicAssert.Greater(0.1, double.NaN);
        Assert.That(0.1, Is.GreaterThan(double.NaN));
    }

    // Fail
    [Test]
    public void NUnitLessThanTest() => ClassicAssert.LessOrEqual(0.1, double.NaN);
}

[manfred-brands]

@shimat I can confirm this issue, but what is correct is not that straight forward:

NUnit will use Double.CompareTo

This contains the following table:

[Test]
public void CompareWithNaN()
{
    const double actual = 0.1;

    Assert.Multiple(() =>
    {
        Assert.That(actual.CompareTo(double.NaN), Is.Positive);
        Assert.That(double.NaN.CompareTo(actual), Is.Negative);
        Assert.That(double.NaN.CompareTo(double.NaN), Is.Zero);

        Assert.That(actual, Is.GreaterThan(double.NaN));
        Assert.That(double.NaN, Is.LessThan(actual));
        Assert.That(double.NaN, Is.EqualTo(double.NaN));

        Assert.That(actual > double.NaN, Is.True);
        Assert.That(double.NaN < actual, Is.True);
        Assert.That(double.NaN == double.NaN, Is.True);
    });
}

Which matches the current NUnit behaviour, but not the C# expression as only the last 3 tests fail.

[shimat]

Thank you, I understand.
I still think GreaterThan when NaN is in the argument is misleading behavior, but I understand the reason and this issue can be closed.