Re: [xwiki-devs] [Brainstorming] Should we support only one skin in our functional tests or not?

  On 21 Aug 2014 at 11:58:01, Denis Gervalle (dgl@softec.lu(mailto:dgl@softec.lu)) wrote: You’re mixing a bit everything in the same bag here :) There are 2 aspects: - It’s specifically because we don’t have that much manpower that we must have our CI automatically test on various browsers/databases/OSes. This works well because the tests don’t need to be modified (in theory at least but mostly true) to be executed on various browsers/databases/OSes. - for the various skins it’s different because you need to write several test frameworks (one per skin) so it’s indeed more work. It depends if we’re able to fulfil a large majority of use cases with only 1 skin or not. The fact that we’ve been able to support only one skin shows how complex writing an XWiki skin is. If you check on extensions.xwiki.org you’ll see we have very few skins, especially compared to other tools like wordpress, drupal, etc. We’ve been saying this for years but we haven’t made progress. Quite the opposite, over the years, we’ve made writing skins more complex (for example right now if you want to write a new skin you need to make sure to implement all Extension Points in your skin if you want your skin to behave properly with extensions!). Now one solution could be to have only 1 (or a few) base skins and improve Skin Themes so that users can do more and more in a Theme, thus allowing just to modify Themes and not a full skin. We started with Color Themes, added the ability to change the logo, and we are now moving to Themes with the ability to also change fonts. This is a step in the right direction I believe. In some future, we could maybe also support more Theme features. At some point we might need to fold default Panels layout (1 column, 2 columns, 3 columns, etc) and Panels selection into Themes too. Etc. If we push this to a level where having 1 skin and using various Themes allow XWiki users to implement the UI they need for their project then we won’t need to support more than 1 skin. But I fear this is just pushing the testing problem to Themes (but maybe to a lesser degree, depending on the freedom we give to Themes) ;) And if we wish to support several Themes we’ll need to test them… Anyway, for 6.2 I agree we don’t have the time to make our UI test framework multiskin-enabled. However, I believe we’ll need to refactor our test framework to allow supporting testing several UI. That will be useful when we need to do that but also for our users writing new skins and needing tests for them. I don’t think it’s that complicated to do. Thanks -Vincent