Negative testing, pairwise testing, orthogonal array testing, and combinatorial testing

Impossible-to-test-for scenarios involve combinations of test inputs that will NEVER appear together in the real world. A good example would be using an Apple computer's Operating System and trying to launch Internet Explorer. It cannot be done under typical circumstances. There is no point in trying to test for it, because Internet Explorer has not been available on Mac computers for years. The appropriate way to handle these scenarios is to simply prevent them from appearing in your tests using the Invalid Pair or Married Pair feature.

Negative Tests are different. You want to include certain scenarios in order to confirm, for example, that certain types of users will not be able to perform certain actions. If you had different types of users for an airplane reservation system, for example, you might want to confirm that no role other than a Super-Admin User would be able to modify the ticket price. It is important to test for these kinds of scenarios.

It might be confusing at first to understand how to address negative tests in the context of generating Hexawise tests, but the effort you make to clearly understand your negative testing options will be well worth it because this topic comes up frequently in most projects.

The easiest way to handle negative tests is to simply keep them separate from your Hexawise-generated tests. Many teams using Hexawise find it easiest just to use Hexawise to generate their positive tests. Then, outside of Hexawise, they will document negative tests in the same way prior to Hexawise.

There is nothing wrong with this approach and it has the advantage of being "clean" and easy to explain. Depending upon your teams tools and processes, and perhaps the quantity and nature of the particular negative tests you have in mind, it might be an appropriate solution for you.

If you decide to use this approach, consider using the "Notes" feature within Hexawise; some teams use the Notes feature to document their ideas for negative tests to make sure they don't get lost.

Hexawise sets of tests include scenarios that generally have the same number of steps.  It is important that every test actually gets executed from start to finish. Why? Because of interactive coverage measurement. Hexawise tests ensure that you will achieve coverage of all of the interactions you have selected (2-way interactions, 3-way, etc.)  If some of your tests stop part-way through, you will not achieve your desired coverage goal. The following example demonstrates this important point.

Imagine that you have a test that says:

What is the problem with this kind of scenario that would risk "Killing the Script" in the 5th step?  Steps 6 through 9 would never get executed.  The Hexawise coverage algorithm has specifically selected each of the values that appear in each of the 9 steps to achieve your coverage objective. But if this test were to get killed after step 5, all of the following interactions that should be tested would never actually get executed:

From India & Vegetarian

From India & XYZ Discount type

From India & FF Miles

From India & Snail Mail

To France & Vegetarian

To France & XYZ Discount type

To France & FF Miles

To France & Snail Mail

Depart Tomorrow & Vegetarian

Depart Tomorrow& XYZ Discount type

Depart Tomorrow & FF Miles

Depart Tomorrow & Snail Mail

First Class Travel & Vegetarian

First Class Travel & XYZ Discount type

First Class Travel & FF Miles

First Class Travel & Snail Mail

Vegetarian & XYZ Discount Type

Vegetarian & FF Miles

Vegetarian & Snail Mail

XYZ Discount Type & FF Miles

XYZ Discount Type & Snail Mail

FF Miles & Snail Mail

Sometimes it is simple to avoid the problem described above.  Sometimes you can just change the way you're describing the Value that causes an error message to appear in a way that makes it possible to achieve both of these objectives:

In the flight reservation example above, here's how we could change the description of the Value we enter into Hexawise to accomplish both of those goals:

Problem solved.

The example above assumes that a tester would be able to trigger an error message with an invalid entry, and then fix the problem that caused it and continue onwards to execute the entire test script. Sometimes, though, a tester might not be able to fix the problem nor continue executing steps. The following kind of test for an ATM is an example of a "script-killing" test:

"Confirm that when an ATM user enters their PIN incorrectly 5 times, then the machine should physically destroy the user's card."

If you had a set of tests that involved multiple steps and were supposed to happen after the user entered their PIN code, such as withdraw a certain amount of money from a certain type of account, you would run into the problem described immediately above. Important combinations of test inputs would not be tested. In this example, there is nothing that a tester would be able to do after the card was physically destroyed; it would not work to say, "now fix the problem and continue executing the test." It would be impossible to resurrect the card once it had been physically destroyed.

If you have a script-killing test that you want to run, do not include it in your Hexawise tests. Instead, create a test for that scenario outside of Hexawise.
​