.Net 2.0: String.IsNullOrEmpty causes runtime exception

via Bill McCarthy.

I was sceptic at first but then I tested the sample and it is true - I got a dirty runtime exception in the release assembly. Man, this is scary.

So I compared debug vs. release assemblies via reflector and I get the following:

// VS code
static void test(string x)
{
    for (int j = 0; j < 10; j++)
    {
        if (String.IsNullOrEmpty(x))
        {       
            // TODO:
        }
    }
}


// DEBUG assembly
private static void test(string x)
{
    for (int num1 = 0; num1 < 10; num1++)
    {
        if (string.IsNullOrEmpty(x))
        {
        }
    }
}


// RELEASE assembly - you get EXCEPTION here !
private static void test(string x)
{
    for (int num1 = 0; num1 < 10; num1++)
    {
        string.IsNullOrEmpty(x);
    }
}

Related Posts:

• Enterprise Library 2 Logging: Rolling file trace listener
• Atlas, Atlas, Atlas
• Enterprise Library: Bug with decimal in DataAccess block
• Updates to Reflector: Reading local variable names from PDB files
• Enterprise Library 2.0 -- DataAccess block
• Enterprise Library 2.0 extensions
• .NET 2.0: List<T>.Exists and anonymous method
• Enterprise Library for .NET Framework 2.0
• This site is running on ASP.NET 2.0 (2)
• .NET 2.0: List<T> is my hero

This just makes me think that the problem is in the JIT working on the empty if. What happens if you place some trivial code inside the if instead ?

User: No, it still throws exception if I place some trivial code inside the if like: int i = j;

Yuck. I wonder how prevalent this failure would be in production code. Seems odd that no one found it until now. I did notice that it only seems to fail when the upper loop bound is > 4. Might be an unrolling issue?

What is unrolling issue ?