How far can you go with GUI automated tests?

[Rachel Swailes]

Hi there

I've done a lot of reading about testing swing gui's with tools like JUnit and Abbot, and I am trying really hard to get into the mindset of using these tools, but I am a bit sceptical about how far they can really go.

We have one guy at work that we generally give all the GUI's to, to test manually. He clicks everywhere in the whatever order and generally tries to be our dumbest user. And in this way we get to iron out a lot of bugs.

There's no way to really predict how he tests stuff and that is really how user's work too. There's no way to predict what bizarre things they will try to do next.

Can automated gui tools really test as thouroughly as he can?

Any opinions would be very helpful.

Many kind regards,
Rachel

[Ilja Preuss]

Automated GUI testing doesn't *replace* exploratory testing, it does complement it.

Say your manual tests highlight a bug in the system. How do you make sure that that bug never again gets introduced, and that similar bugs aren't in the system too? You can't rely on exploratory testing to *always* find this bug, simply because it doesn't always test the same things. So what you do is write an automated test that makes sure that the bug is fixed and doesn't reapear, and you reflect on what other tests to write to prevent similar bugs.

Then your tester can concentrate on finding new kinds of bugs you don't yet have tests for, instead of wasting his energy on the same bugs again and again.

Does that sound reasonable?

[Rachel Swailes]

Aah, that makes a lot of sense. So basically, our guy find the bugs and our automated testing makes sure that they don't show up again.

Thanks!

Rachel