Should we make MultipleAssertException.TestResult maybe-null or obsolete two constructors?

[jnm2]

@nunit/framework-team

Option 1

namespace NUnit.Framework
{
    public class MultipleAssertException : ResultStateException
    {
        // Throws if testResult is null
        public MultipleAssertException(ITestResult testResult);
        
        // Leaves TestResult null
        public MultipleAssertException(string? message);
        
        // Leaves TestResult null
        public MultipleAssertException(string? message, Exception? inner);

        protected MultipleAssertException(System.Runtime.Serialization.SerializationInfo info, System.Runtime.Serialization.StreamingContext context);

        public override ResultState ResultState { get; }

        // Maybe null
        public ITestResult? TestResult { get; }
    }
}

Option 2

namespace NUnit.Framework
{
    public class MultipleAssertException : ResultStateException
    {
        // Throws if testResult is null
        public MultipleAssertException(ITestResult testResult);
        
        // Leaves TestResult null
+       [Obsolete("The MultipleAssertException(ITestResult) constructor must be used.")]
        public MultipleAssertException(string? message);
        
        // Leaves TestResult null
+       [Obsolete("The MultipleAssertException(ITestResult) constructor must be used.")]
        public MultipleAssertException(string? message, Exception? inner);

        protected MultipleAssertException(System.Runtime.Serialization.SerializationInfo info, System.Runtime.Serialization.StreamingContext context);

        public override ResultState ResultState { get; }

-       // Maybe null
-       public ITestResult? TestResult { get; }
+       // Not null
+       public ITestResult TestResult { get; }
    }
}

master...jnm2:annotate_exceptions

[rprouse]

We only ever use the first constructor and it was a mistake to include the other two, so I'd vote for obsoleting the constructors or even removing them. This is a public class, but it is an internal implementation. What is the likelihood of someone extending NUnit and using this exception and those constructors?

[jnm2]

It seems very unlikely to me, but I don't have a good way to reality-test that. We could also take the stance that code that calls those methods is already in error and we are aren't making it that much worse by exchanging the exception-with-null-TestResult for a MissingMethodException (if running a library that calls these constructors) or a compile-time error (if calling these constructors directly)?