- From: Håkon Wium Lie <howcome@opera.com>
- Date: Tue, 8 Nov 2011 02:22:51 +0100
- To: John Jansen <John.Jansen@microsoft.com>
- Cc: "public-css-testsuite@w3.org" <public-css-testsuite@w3.org>
John Jansen wrote:
> I was just looking through the multicol test suite, and it looks to
> me like there is a problem with the reference files.
>
> For example, the test for reduce-000 [1] and the reference [2] have
> the exact same CSS in them. I would expect the ref to use something
> other than columns in order to achieve the same effect as the
> actual test file. This is also true in the basic reference file as
> well [3].
> [1] http://test.csswg.org/suites/css3-multicol/nightly-unstable/html4/multicol-reduce-000.htm
> [2] http://test.csswg.org/suites/css3-multicol/nightly-unstable/html4/multicol-reduce-000-ref.htm
> [3] http://test.csswg.org/suites/css3-multicol/nightly-unstable/html4/multicol-basic-ref.htm
Right. These tests were originally written as a one-page humanistic test:
http://lists.w3.org/Archives/Public/www-style/2010Aug/0278.html
It was suggested that the tests were converted to reftest format. I
was somewhat hesitant to do so:
http://lists.w3.org/Archives/Member/w3c-css-wg/2010JulSep/0085.html
And, as you have noticed, some of the tests are not very meaningful
when converte to reftests. I don't know of a way to render the text
without using columns in a way that guarantees exactly the same line
breaking etc.
Perhaps it's better to keep the one-page humanistic test as just that,
and to write new tests for automated testing. Here's the beginnings of
an automated test suite:
http://lists.w3.org/Archives/Public/public-css-testsuite/2011Sep/0041.html
Cheers,
-h&kon
Håkon Wium Lie CTO °þe®ª
howcome@opera.com http://people.opera.com/howcome
Received on Tuesday, 8 November 2011 01:23:33 UTC