Skip to main content

The Madman's Approach to Testing Software

Posted by rah003 on May 26, 2014 at 6:14 AM PDT

I've been developing webapps pretty much since the start of the web in the second half of the 90s, and in Java ever since the first edition of Java EE. During that time, I've met many people and have seen many different approaches to both the development and the testing of the code.

One of the funnier parts of web app testing was a testing method used by one of the people I worked with. Between me and other colleagues, we secretly called it the Madman's test. The test consisted of starting up a webapp. Once it was up, or at least once the start screen was present, he started to click around the web app like crazy. It didn't matter to him what he clicked or what he did, all that mattered was whether he managed to make the application produce an error message. If there was a single error, the app would fail the test. It was part funny, part annoying.

Only lately have I come to appreciate this method of testing. It can certainly help with discovering some of the issues that regular users might run into, for example by double clicking on some action by accident. And as you know, any issue that can be found and fixed in development will make user appreciate the app just a little more. Especially since some of these issues can be fixed with very little effort, once you know the exact cause and right place for the fix. Just like the fix provided in MGNLUI-2701: in the end, the fix was just a single line of code per type of dialog. Kudos to the team for coming up with the UI architecture of Magnolia 5 - their hard work and focus on clean code without duplicates made fixing such issues easy.

So I encourage you to try the Madman's test in your next software testing session...Or, if you have experience with it, please share valuable findings in the comments.

Image source: