Re: 8 questionable border-style conflict resolution testcases in collapsing border model (Section 17.6.2.1)

On 09/13/2010 04:47 PM, "Gérard Talbot" wrote:
> Hello,
>
> 1-
> http://test.csswg.org/suites/css2.1/20100815/html4/border-conflict-style-045.htm
>
> "
> Border Conflict Resolution (by border style): 'hidden' vs. 'hidden' -
> border with highest priority border style should dominate
>   * The 'hidden' border collapsed with a 'hidden' border yields no border.
> "
>
> Such testcase can not fail and does not test what it is supposed to
> test: which is conflicting border styles at a cell edge ... and there is
> no border-style conflict in that test.
>
>
> 2-
> http://test.csswg.org/suites/css2.1/20100815/html4/border-conflict-style-047.htm
>
> "
> Border Conflict Resolution (by border style): 'hidden' vs. 'none' -
> border with highest priority border style should dominate
>   * A 'hidden' border collapsed with a 'none' border yields no border.
> "
>
> Such testcase can not fail (if 'border-style: none' and 'border-style:
> hidden' taken individually in 2 testcases show that they are properly
> implemented) and can not be verified to pass either. Border: none and
> border: hidden: how do you know which border-style won out of the
> conflict? How can the test indicate that 'border: hidden' actually had
> priority over 'border: none' or vice versa? It's impossible to say...
> and the test is suppose to test resolution of border conflicts.
>
> When testing a browser, let's say a medium black border appears in
> #center cell of that border-conflict-style-047.htm testcase: then you
> still can not and could not and would not know which of the 2
> border-style has not been properly implemented.
> I question the relevance of such testcase.

I think it's valuable to have such testcases, even though there technically
isn't a conflict here, for completeness. However, I agree with you that the
pass conditions should be more stringent. Also, the border that should be
ignored should be red. Lachlan's tests at least handle this aspect correctly.
The two sets should probably be merged, perhaps using Lachlan's code and
Microsoft's metadata.

In general, the pass conditions for all of the border conflict tests should
be more stringent, checking not only that the correct color shows up, but
that the correct width and border style show up. There could, for example,
be a bug in the browser's border conflict handling where the color and width
of the winning border is correctly used, but the border style is not stored
correctly and is wrong.

~fantasai

Received on Wednesday, 15 September 2010 05:21:31 UTC