Geoff Bache

Automated testing via the GUI has got itself a bad reputation - and with good reason. It is associated with bloated, expensive tools, horrible proprietary languages and maintenance nightmares. The prevailing wisdom has become that the best thing to do is to minimize testing through the GUI and instead go directly into the code via an API. But does it have to be like that? The heart of the problem is one of coupling. Tests that go through the GUI end up too tightly coupled to it, and hence break when it changes. What if we had an abstraction layer in between, a "GUI interpreter" that mapped the possible actions in the current GUI to carefully chosen domain-language statements? We could have high-level, readable tests that describe user intentions rather than GUI mechanics. PyUseCase is a tool that enables you to work in exactly this way. I've been using it for several years to testing rich client applications written with Python and GTK (similar tools exist for Java and .Net). In this talk I hope to demonstrate how productive this approach to GUI testing can be.

Testing a program by monitoring changes in it's plain text log files is an old idea, but one which has fallen out of fashion lately. This talk aims to show you that with a tool like TextTest, it can not only work, but work well. Well enough to enable agile development, especially in situations where for example xUnit breaks down. This is a technique and a tool that I have developed over the past decade at Jeppesen, for automating maintainable black-box tests that work entirely outside of the code. In this talk I will demonstrate how the tool works in practice, and examine the advantages and disadvantages of this approach in general.