[Kss-devel] Regression/BBB in client-side action names and parameters

[Balazs Ree]

Hi Martin,

On Sat, 29 Sep 2007 23:57:37 +0100, Martin Aspeli wrote:
 
> This is just a heads-up... I went to look at the KSS interfaces again to
> look up some syntax, and then realised that since I wrote the KSS
> chapter in my book, you've deprecated/renamed a bunch of functions and
> parameters. These are the perils of writing about a pre-released version
> of Plone, of course. :)
> 
> Whilst I welcome the consistency this brings, it does mean that the book
> is advising people to use the old syntax. Unfortunately, production is
> now finished, so it's too late to make changes. I am finding out how to
> publish errata, but at best this will be a note on some website. :-(

I am very sad to hear this, since I was extremely conscious about this 
issue. This was the last change made in kss.core, it was done way before 
the Plone release and getting it correctly in your book was a high 
priority for me.

> Furthermore, it seems that the BBB code is not working (and the error
> message in Firebug is not particularly helpful).

... and, by the time I made the change I really made sure the deprecation 
does work correctly. In addition I refused to make any new changes in the 
code of the kss.core after that point. That's all the precaution one can 
possibly imagine to not have things broken.

However it happened and the only thing that remains now is to fix the 
BBB, and I will do that soon.




Some words about regression testing of KSS because it is a connecting 
topic.

Unfortunately at the moment we currently have a single way to test issues 
like this, and this is with the kissbot. This tests things with Selenium 
in pure kss (kss.demo), but at the moment not in Plone. This affects 
addClass and friends, because at the moment they only work in Plone and 
not in pure kss (this will also change but it is so at the moment). What 
do we need to have this situation of testing improved?

One way is to *insist* that people write Selenium tests to cover *all* 
functionality. On the level of kss plugins we will support this with a 
paster template that gives you a skeleton with a working demo and a 
directory to place the selenium tests into, which then will run 
automatically by the kissbot. We also have to enforce selenium testing on 
the level of the Plone application, and since Sorrento I started to 
hassle people to do so as a base requirement for new features. However 
since then noone really complies to this "rule" which is currently mine 
only and there is no community consensus behind it.

The next thing is that we also need to be able to run these tests 
automatically, at least nightly. For this we need more servers to be 
donated as slaves in the kissbot pool. Godefroid will provide more 
information about this, afaik we really need a lot of computing capacity 
for this This calls for a community action and I believe we should sit 
together to discuss this issue in Naples. It would be a real benefit for 
Plone to have full automated Selenium test coverage, not only for the 
benefit of KSS. 

A last remark about the error messages: we redesigned the error handling 
to give more useful error traceback during the last kss sprint, but that 
is only part of the new kss.core version (will come in effect in Plone 
3.5). It would be possible to backport it to 3.0 but it touches a lot in 
the code and the decision was not to do it.

-- 
Balazs Ree