Very easy to misuse `ThrowAsync` in a sync test

Description

The API for asserting exceptions in async code is VERY easy to misuse. Such tests would pass, but in fact should fail.

The problem is that action.Should().ThrowAsync<Exception>() should be awaited, but Roslyn doesn’t know that and doesn’t warn you. It’s very easy to forget to put it in an async test, and to await it.

Complete minimal example reproducing the issue

This test has a very subtle hard-to-detect bug. It should fail, but will actually pass.

[Fact]
public void sync_test_that_should_fail_but_will_pass()
{
    Func<Task> action = () => throw new ArgumentNullException("paramName");
    action.Should().ThrowAsync<ArgumentNullException>().WithParameterName("foo");     // "foo" is incorrect
}

This test fixes the bug. It should fail, and will fail.

[Fact]
public async Task async_test_that_should_fail_and_will_fail()
{
    Func<Task> action = () => throw new ArgumentNullException("paramName");
    await action.Should().ThrowAsync<ArgumentNullException>().WithParameterName("foo");
}

Expected behavior:

Don’t know! 😄

Firstly please tell me, are you aware of an analyser (third party maybe?) that checks for this?

If not, there are two options:

• Maybe someone someday will write an analyser that detects this problem. That would be best. It would see that you don’t await when calling a ThrowsAsync. I know nothing about Roslyn so regrettably cannot assist.
• Is it possible to detect this at runtime and log a warning to the console?

Actual behavior:

Doesn’t warn you.

Versions

• Which version of Fluent Assertions are you using? 6.2.0
• Which .NET runtime and version are you targeting? .NET Core 5

Additional Information

Maybe there’s a different way to handle this problem - if so I’d appreciate the advice!

[lonix1]

Excellent blog article Dennis!

[dennisdoomen]

Here you go… https://www.continuousimprover.com/2022/01/awaiting-the-awaitable