Xamarin中的NUnit异步测试不起作用

Question

I'm new to the C# world, trying to give Xamarin an evaluation. I've been making a small web api library, and it has been going smoothly so far, but now that I'm trying to test networking components, I'm having some trouble.

OS:           OS X 10.9.1
Xamarin:      4.2.2 (build 2)
Library:      PCL targeting 4.0 (Profile158)
Nunit target: Mono / .NET 4.5

The class / method:

using System;
using System.Threading;
using System.Threading.Tasks;
using System.Net.Http;

namespace LibTest
{
    public class AsyncTest
    {
        public async Task<string> DoTest()
        {
            return await new HttpClient().GetStringAsync( "http://www.reddit.com/.json?limit=1" );
        }
    }
}

The NUnit test class / case:

using System;
using NUnit.Framework;
using LibTest;

namespace LibTestTests
{
    [TestFixture]
    public class LibTestTest
    {
        [Test]
        public async void TempTest()
        {
            string data = await new AsyncTest().DoTest();

            Assert.IsNotNull( data );
            Assert.IsNull( data );
        }
    }
}

All of my other tests seem to be working (parsing, etc), but this one just goes right by, acting as if it passed, but I don't see how it could have, with the contradictory assertions.

Is this not supported currently, or am I incorrect in my implementation?

If I make it synchronous with ContinueWith() / Wait() , it works as expected, but obviously I don't want to have duplicate implementations in my library, just to pass the tests.

Answer 1

NUnit does support async Task unit tests; unfortunately, Xamarin is using an older version of NUnit that does not. However, I do not recommend synchronously blocking on the task directly because that will complicate your error path testing (exceptions are wrapped in an AggregateException ).

I have an AsyncContext type in my AsyncEx library that can be used as such:

[Test]
public void TempTest()
{
    AsyncContext.Run(async () =>
    {
        string data = await new AsyncTest().DoTest();

        Assert.IsNotNull( data );
        Assert.IsNull( data );
    });
}

Answer 2

You have an async void method. This should be a big red flag whenever you use it.

As soon as TempTest hits the very first await it will return to the caller. The rest of the code runs , but from the point of view of the caller the method has completed, and the caller has no way of knowing when the method ends, any exceptions it throws, any asserts that fail, etc.

When you're synchronously waiting you are of course ensuring that the exception is thrown by the thread that calls this method, rather than in some other thread in some other place where the testing framework can't observe it.

To properly test an asynchronous method the testing framework really needs to have been designed to support it, really. It should be able to handle methods that return a Task , and act based on the results of that Task . If that's not an option, you'll need to do this work manually. Ideally, to maintain proper asynchrony, instead of waiting on any of the tasks, you should set up your own message pump (see here for an example) which will be able to pump the messages until the top level operation finishes, thus allowing you to run asynchronous code in a synchronous method.

Answer 3

The sequence of operations here is:

1. NUnit runs TempTest()
2. TempTest calls new AsyncTest().DoTest(), which returns a Task
3. The rest of TempTest (the asserts) is scheduled as a continuation on the task
4. The TempTest method returns (void), and NUnit moves onto the next test
5. The continuation executes the asserts at some point in the future
6. Since nothing is observing the continuation, the assert failure exception is unhandled

What you probably want is for your test to be synchronous, and block on the result of the asynchronous thing you're testing:

[Test]
public void TempTest()
{
    string data = new AsyncTest().DoTest().Result;

    Assert.IsNotNull( data );
    Assert.IsNull( data );
}