How to use resetState method of org.mockitousage.plugins.stacktrace.PluginStackTraceFilteringTest class

Nutshell: Can't always use when to stub spies; use doReturn.

Assuming static imports of spy and doReturn (both PowerMockito):

@RunWith(PowerMockRunner.class)
@PrepareForTest(B.class)
public class BTest {
    @Test public void testClass() throws Exception {
        B b = spy(new B());
        doReturn(42).when(b, "m");
        b.asd();
    }
}

You could also @PrepareForTest(A.class) and set up the doReturn on when(a, "m"). Which makes more sense depends on the actual test.

Like you said, it fails because the arguments are different. Take a look at the test below and you'll see that the second test method will fail because the status in your MyClass instance is different from SomeStatus that you passed in the matcher.

public class MatcherTest {

    class MyClass{
        private String status;

        MyClass(String status) {
            this.status = status;
        }

        public String getStatus(){
            return status;
        }
    }

    class MyDao {
        public void update(MyClass myClass){}
    }

    class StatusMatcher extends ArgumentMatcher<MyClass> {
        private String status;
        public StatusMatcher(String hs) {
            status = hs;
        }

        @Override
        public boolean matches(Object argument) {
            return status.equals(((MyClass)argument).getStatus());
        }
    }

    @Test
    public void shouldMatchStatus(){
        MyDao mock = mock(MyDao.class);
        mock.update(new MyClass("expectedStatus"));
        verify(mock, times(1)).update(argThat(new StatusMatcher("expectedStatus")));
    }

    @Test
    public void shouldNotMatchStatus(){
        MyDao mock = mock(MyDao.class);
        mock.update(new MyClass("unexpectedStatus"));
        /* THE BELLOW WILL FAIL BECAUSE ARGUMENTS ARE DIFFERENT */
        verify(mock, times(1)).update(argThat(new StatusMatcher("expectedStatus")));
    }
}

I could take a wild guess that you could be reusing variables, or have a static field, etc. But without seeing your test code, no one can tell.

It sounds like you may want to be using the .expects(1) type methods that mock frameworks usually provide.

Using mockito, if you were testing a List and wanted to verify that clear was called 3 times and add was called at least once with these parameters you do the following:

List mock = mock(List.class);        
someCodeThatInteractsWithMock();                 

verify(mock, times(3)).clear();
verify(mock, atLeastOnce()).add(anyObject());      

Source - MockitoVsEasyMock

You can write your own Answer:

when(mock.testMethod(anyString())).thenAnswer(new Answer<String>() {
    @Override
    public String answer(InvocationOnMock invocation) {
        return invocation.getArgumentAt(0, String.class);
    }
});

Or:

when(mock.testMethod(anyString()))
    .thenAnswer(AdditionalAnswers.<String>returnsFirstArg());

The answer is yes, you can, and there are basically two levels of doing this, based on the need of your test.

If you merely want to test the interaction with the mocked object, you can simply use the verify() method, to verify that the void method was called.

If your test genuinely needs the mocked object to modify parameters passed to it, you will need to implement an Answer:

EDITED to show proper form of using Answer with void method

doAnswer(new Answer() {
    @Override
    Object answer(InvocationOnMock invocation) {
        Object[] args = invocation.getArguments();
        ((MyClass)args[0]).myClassSetMyField(NEW_VALUE);
        return null; // void method, so return null
    }
}).when(mock).someMethod();

In Java 8+, the above is simplified with a lambda:

doAnswer(invocation-> {
    Object[] args = invocation.getArguments();
    ((MyClass)args[0]).myClassSetMyField(NEW_VALUE);
    return null;
}).when(mock).someMethod();

It's counting your setup as invocations since deeps stubs is not supported in the verification API, and complains on the second call which is:

when(mockSandBox.getSand().doB()).thenReturn(1);

I would skip using RETURNS_DEEP_STUBS and just use another mock:

...
@Mock
SandBox mockSandBox;

@Mock
Sand sand;

@Test
public void should(){
    when(mockSandBox.getSand()).thenReturn(sand);
    when(sand.doA()).thenReturn(1);
    when(sand.doB()).thenReturn(1);
    when(sand.doC()).thenReturn(1);
...

You need to inject the mock into the class which you are testing. You do not need access to Subscriber. The way mockito and other mocking frameworks help is that you do not need access to objects which you are interacting with. You do however need a way to get mock objects into the class you are testing.

public class WebAdaptor {

    public WebAdaptor(Subscriber subscriber) { /* Added a new constructor */
       this.subscriber = subscriber;
    }

    private Subscriber subscriber;

    public void run() {
        subscriber.init();
    }
}

Now you can verify your interactions on the mock, rather than on the real object.

@Test
public void runShouldInvokeSubscriberInit() {

    // Given
    Subscriber mockSubscriber = mock(Subscriber.class);
    WebAdaptor adaptor = new WebAdaptor(mockSubscriber);  // Use the new constructor

    // When
    adaptor.run();

    // Then
    verify(mockSubscriber).init();
}

If adding the Subscriber to the constructor is not the correct approach, you could also consider using a factory to allow WebAdaptor to instantiate new Subscriber objects from a factory which you control. You could then mock the factory to provider mock Subscribers.

I would avoid it. Generally speaking, you should structure things as best you can to avoid needing to test private variables and stick with the interactions that the rest of the application will be exposed to. This is typically public methods/properties and protected methods by using fakes/doubles.

The purpose of the unit tests are to verify how the object under test is interacted with, not necessarily it's implementation details. Basically, with input x, I should get output/effect y. If you get hung up on testing exactly how x produces y, you run the risk of making your classes inflexible to change. Every refactor of your class is likely to break your tests, even if the output and end result is the same.

The following suggestion let's you test abstract classes without creating a "real" subclass - the Mock is the subclass.

use Mockito.mock(My.class, Mockito.CALLS_REAL_METHODS), then mock any abstract methods that are invoked.

Example:

public abstract class My {
  public Result methodUnderTest() { ... }
  protected abstract void methodIDontCareAbout();
}

public class MyTest {
    @Test
    public void shouldFailOnNullIdentifiers() {
        My my = Mockito.mock(My.class, Mockito.CALLS_REAL_METHODS);
        Assert.assertSomething(my.methodUnderTest());
    }
}

Note: The beauty of this solution is that you do not have to implement the abstract methods, as long as they are never invoked.

In my honest opinion, this is neater than using a spy, since a spy requires an instance, which means you have to create an instantiatable subclass of your abstract class.