What is Continuous Regression Testing?

In my previous entry I said that our business case was Continuous Regression Testing. I will make a few entries on the subject.

Our definition was quite clear on what level of testing we wanted to achieve.

We define it at the System Test level, meaning regression testing of our end to end functionality as specified by our Use Cases. Important here is end to end and driven by Use Cases.

As an example say that our system would have a "Register User" Use Case which states that the user should be able to register through a portal and receive an confirmation email. In this case our end to end delivery is from our web front to our mail server. We would be creating an automated Test Case which registers the users through the portal and then the test evaluates that there is a mail sent to the email address of the registered user.

Me and others in our company (not on the team at the time) at the time have had quite a bit of experience working with automated System Tests. Yet we ran into quite a few mines (AGAIN!!) when building our test suites. I will cover these later.

So we had a clear understanding on how we wanted to test and we knew we wanted to test it often. For us testing on each code commit and so heavily relying on test automation was way out of our comfort zone. But our business case was strong and we knew why we wanted to go down that road.

Early on we made a few key decisions. One was to test everything on every commit. Second was to only build each artifact once. We discussed this and came to the conclusion that we would need to retest after a rebuild. We felt our process would be simpler and more correct by doing this.

Our architecture was based on a few well defined components (publishing rest services) that together formed our deliverable.

This is a simplistic view of something similar. Some components where internal and some where externally integrating.

We where aware that we most likely would need to test the components in a continuous way but we ignored it on purpose. Our focus had to be on our delivery as well and we all know too well that we need to prioritize well when working on deadlines. Its easy to get carried away on theoretical models and a desire to solve everything but its not the way we work. We always want something deliverable and then iterate it.

So our initial pipe ended up looking something like this.  Each of our components where built, unit tested and released on commit. The our deployment assembly was updated one component at the time. Each time the assembly was updated we deployed it and tested it on our test server.

It was a very simplistic start to our Continuous journey but it was very adequate for us and most importantly it solved our primary concern Continuous Regression of the customer application.

Its important to remember that the application was new and our team was small at the time.