Unit testing function that calls other function

Question

Say I have the following two functions:

add_five (number) -> number + 2

add_six (number) -> add_five(number) + 1

As you can see, add_five has a bug.

If I now test add_six it would fail because the result is incorrect, but the code is correct.

Imagine you have a large tree of functions calling each other, it would be hard to find out which function contains the bug, because all the functions will fail (and not only the one with the bug).

So my question is: should unit tests fail because of incorrect behaviour (wrong results) or because of incorrect code (bugs).

Possible duplicate of [How to test a class that has private methods, fields or inner classes?](http://stackoverflow.com/questions/34571/how-to-test-a-class-that-has-private-methods-fields-or-inner-classes) — Raedwald, Aug 20 '16 at 11:11
No test can perform the magic to detect incorrect code. All it can do is detect incorrect behaviour. — Raedwald, Aug 20 '16 at 11:12
Is it right to say that in that case unit tests don't really help to find bugs? It tells you that there is invalid behaviour, but doesn't directly tell you where the bug is. — Jan-Paul Kleemans, Aug 20 '16 at 11:25
Incorrect behaviour indicates the presence of a bug, not the location. — Raedwald, Aug 20 '16 at 17:34

score 3 · Accepted Answer · answered Aug 20 '16 at 11:48

should unit tests fail because of incorrect behaviour (wrong results) or because of incorrect code (bugs)?

Unit tests usually fail because of wrong results. That's what you write in assertions: you call a method and you define the expected result.

Unit tests cannot identify incorrect code. If the operation is return number+5 and your CPU or your RAM has a hardware problem and return something different, then the test will fail as well, even if the code is correct.

Also consider:

public int add_five(int number)
{
    Thread.Sleep(5000);
    return number+5;
}

How shall the unit test know whether the Sleep is intended or not?

So, if any unit test fails, it's your job to look at it, find out why it fails and if it fails in a different method, write a new unit test for that method so you can exclude that method next time.

score 0 · Answer 2 · answered Aug 21 '16 at 18:44

Presumably you have a test for add_five/1 and a test for add_six/1. In this case, the test for add_six/1 will fail alongside the test for add_five/1.

Let's assume you decide to check add_six/1 first. You see that it depends on add_five/1, which is also failing. You can immediately assume that the bug in add_five/1 is cascading up.

Your module dependencies form a directed (hopefully acyclic) graph. If a dependency of you function or module is broken, that should be what you target for debugging first.

Another option is to mock out the add_five function when testing your add_sixfunction, but this quickly creates a lot of extra typing and logic duplication. If you change the spec of add_five, you have to change every place you reimplemented it as a mock.

If you use a quickcheck-style testing library, you can test for certain logic errors based on the properties of what you're testing. These bugs are detected using randomly generated cases that produce incorrect results, but all you as a tester write are library-specific definitions of the properties you're testing for. However, this will also suffer from dependency breakages unless you've mocked out dependent modules/functions.

Unit testing function that calls other function

2 Answers2