Last weekend, the next Code Camp was
held in Silicon Valley - a large-scale event for developers. Some sessions were devoted to software testing, and I would like to briefly tell you about one of them.
The session was announced as “Next Generation of Testing for HTML5”. In reality, Kevin
briefly described the innovations in html5 using this presentation
, and then outlined his vision of which web application testing tools are most interesting today. As an illustration of what will be discussed, he attached a screenshot of the signature from the Google testing blog: “Life is too short for manual testing”.
As a developer, Kevin could not circumvent unit testing in the report. Quote: “ Many say that it is necessary to cover 80% or even 90% of the code with unit tests. I think that 10% is enough for testing the main functionality. Further testing will be picked up by a QA specialist. However, if he returns to you with a bug - it's time to explore the area in which you made a mistake, go deeper, and start writing regression tests
The unit-testing tool he recommends is QUnit, which was already well covered in Habré
. However, it is not suitable for integration testing, since it cannot verify the functionality of the site depending on the page refresh (log in, for example). And so Kevin’s next recommendation is Mr. Hudson
Hudson is what large companies call “testing lab” (and write for internal needs on their own): it monitors the tasks they perform, collects builds and sends error messages. In different companies, this tool is used by different specialists, but if it is the tester who is instructed to build builds, then at least you should familiarize yourself with Hudson.
"The speaker grumbled.
Easy and simple alternative - Canoo Webtest
. The latest build is September 29, 2010, so the project is alive and improving, despite the simple site.
The product is more serious, and, perhaps, even more revolutionary - Test Swarm
. The innovativeness of the idea is crowdsourcing: it is not necessary to knock out a dozen test machines for yourself, if you just need to check the work of the site in browsers of different versions. You can borrow colleagues' machines for this purpose. The web server analyzes which browsers are available on which of the connected machines and uses them all in testing. Visual presentation on Vimeo
about how it works. The alpha version of the project was released in August last year, I saw the latest commits from March of this year. Kevin said that somehow his colleagues presented him with a build that accurately passed all the tests in one of the versions of FF installed on his computer. “Hmm, let's run it through Test Swarm,” he said, and in all other versions of browsers one or another failure occurred.
PS Perhaps, the mentioned tools are known to most of the subscribers of the community. In this case, congratulations, you are on the front line in the struggle for the quality of web applications)