W3C home > Mailing lists > Public > public-css-testsuite@w3.org > February 2009

RE: Suggestions and corrections to the MS-supplied CSS 2.1 tests

From: Arron Eicholz <Arron.Eicholz@microsoft.com>
Date: Wed, 18 Feb 2009 11:02:02 -0800
To: James Hopkins <james@idreamincode.co.uk>, "public-css-testsuite@w3.org" <public-css-testsuite@w3.org>
Message-ID: <66D5CB5D6AB0694592FAF5487C50368B29B53862AE@NA-EXMSG-C111.redmond.corp.microsoft.com>
James Hopkins wrote:
> Regarding their :first-letter tests, I am by no means an implementor
> but I would have thought punctuation characters are implemented in
> Unicode character sets, so am wondering whether there is any need for
> creating an individual test case for each one of these characters
> (there are currently around 411 test cases for this purpose in the MS
> Conformance Suite). If however there is a need, I would suggest
> aggregating all 411 into one test case where all punctuation
> characters each precede and succeed the first letter - I believe this
> would make it far quicker for implementors to see whether all
> characters are indeed supported. Certainly from my own bug-testing
> point of view, I think this and similar features could be tested on an
> adhoc basis, similar to what the CSS2.1 Test Case Authoring Guidelines
> (http://www.w3.org/Style/CSS/Test/guidelines.html#methodical) state.

There is a need for individual tests, let me try and explain. We created individual tests so that if the developer does have an error in the code there is a simple repro case to look at and they can take quick action for the targeted issue. If you have a large page that contains all the characters and there is an error on it you don't know what the actual error is until you reduce it down to its simplest form. You can guess that it's a character selection problem in this case but it may not be it may be a combination of the characters above or below that cause the issue. While I do agree that a single test containing all the characters would be a quick check that the characters are working how can you prove that the other characters on the page are not affecting your results? Rule of thumb for most of our tests simple is better and faster for resolving and fixing issues. Also I know that there are very few things influencing the specific test I am running.

> I agree with Johannes that there should be a clear pass or fail
> condition. For example, there are many cases where the black text
> color, combined with the text "Test passes if there is no red visible
> on the page", constitutes a pass. Even though the text is self
> explanatory, I suggest the black text become green, constituting a
> pass- this further iterates, and emphasizes a clear pass condition.

There may be particular cases that can fall into that category and if you identify a few of them I may come up with a general solution for all of our cases. But there is one inherent problem with doing this for all cases. What about display: none? Those cases can't alter the text just because they are not visible on the page. And making the pass condition text green in this scenario could be confusing if the test failed. Not to mention that green text is harder to read for some people.

Arron Eicholz
Received on Wednesday, 18 February 2009 19:02:55 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:13:18 UTC