JustinH wrote:Hi eValid team:
I'm trying to develop a random walk through of an application to detect some high level issues, such as 500 errors rather than testing specific functionality of the site.
Thanks for posting JustinH.
This makes sense...neat idea on your part!
It seems, too, that this kind of thing would scale (that is, could be applied to websites in general, fully automatically, and "in the large.")
If you can test objects on a page uniformly then, assuming that you keep on testing at a regular interval, you will eventually discover errors.
But before you jump in too deep, you have to accept that there often are a LOT of elements in a page, and they are sensitive to a LOT of different kinds of inputs.
So, first, you have to figure out which kind of input you're going to be injecting?
For example, a type-in or a mouse click. Lets' just analyze mouse click for a moment.
The difficulty is, if you engineer a test case to attempt to send a mouse-click event to a randomly picked element, almost ALL of them won't do anything.
But some of them will.
If sending a mouse click to an element that is not sensitive to it has no effect, then you can detect this by running a NAV check.
If there's no navigation, then the mouse-click was "null."
If it DID produce a navigation, and that navigation failed, then you've found a problem.
So, overall, this idea seems to have promise. It is worth trying it out using the resources eValid provides for manipulating the DOM of the current page, described here: http://www.e-valid.com/Products/Documen ... tives.html
The worry we have is, that there at so many possible actions that you will have a very low yield of defect detection.
Let us know of your results!
-- eValid Support