Re: [gnome-desktop-testing] A couple of comments

[Javier Collado <javier collado canonical com>]

Hello Ara,

The idea that I have regarding a regression is that it's a battery of 
test cases that are used to verify that functionality isn't broken for 
every new release. For this purpose, I would say that the test cases 
aren't required to have a common set of initial conditions so, with 
regard to the desktop testing framework, they need different setup and 
teardown methods so that's why I though of a regression as a set of test 
suites.

In my opinion, the concept of test suite that is already being used by 
the framework is fine and that it would be good to keep it simple.

Tis conversation makes me think that there isn't a clear difference in 
the framework about what is a test suite and what is an application. In 
fact, the test suites are being implemented as application classes. 
Maybe we could define some kind of main TestSuite class to make it 
clearer what are the attributes and methods of a test suite (i.e. setup 
and teardown) and make the test suite inherit from that main TestSuite 
class and contain one application object instead of making the test 
suites inherit  from an application class. What do you think about this 
change in the class structure?

Best regards,
    Javier

Ara Pulido wrote:
> Hello Javier,
>
> Thanks for your comments. Some replies below.
>
> > In addition to this, the exit method maybe should be renamed to close
> > because it seems its aim is to perform the opposite as open so the API
> > will look more coherent if we use antonyms for these method names.
>
> Both changes seem reasonable.
>
> > - xml file format: As some of you know, I would prefer to write the
> > whole information regarding test suites and test cases in the source
> > files instead of in separate xml files to have all the information in
> > one place. 
> Reason behind the XML file is to be able to parse different arguments to
> test case, without having to duplicate code.
>
> Also, people could contribute with XML files, if they don't know how to
> code. (with strange arguments causing failures, i.e.)
>
> >  For example, I have been
> > thinking in adding a way to define a whole regression in one xml file.
> > That change will only need the ability to add multiple suites to the
> > same xml file, that is, instead of:
> > <suite>
> > ...
> > </suite>
>
> This should be covered by the concept of suite. We might need to change
> OUR concept of suite to be able to cover several applications, as a test
> suite is just a collection of test cases.
>
> A regression is just a kind of test suite. Am I wrong?
>
> Cheers,
> Ara.