> > Somehow the holy grill of every manager is to create a meta-language in > > which the tests will be written, and the coding should be a framework > that > > reads it and execute it. > > The fact that tests wildly vary is a minor annoyance. :-) > > Maybe you need to introduce the *idea* of Fit http://fit.c2.com/ > Then they can write their test case in Excel files. >
How does it helps with tests that vary wildly? I'm trying to do permission testings in a web app, and identified 104 permit-able actions. Some of the actions are simple and have a pattern. for example, make sure that the right menu items are visible, and clicking on them won't lead you to an error page. Or make sure that a certain link exists in a screen. Of course, to be really sure that the link actually do something we need to click on it, and the result of that is harder to make a pattern of. And then, there are the completely un-patterned actions. can the user see his entries? other's entries? which can he edit? can he publish his entries? At this point every few tests have their own pattern and need their own setup/cleanup code. My opinion is that with some code abstraction you can make a test program that is clear and easy to write tests in. Shmuel.
_______________________________________________ Perl mailing list [email protected] http://mail.perl.org.il/mailman/listinfo/perl
