W3C home > Mailing lists > Public > public-css-testsuite@w3.org > October 2010

RE: 15 rejectable XHTML *-replaced-width-* testcases from section 10.3.x

From: Gérard Talbot <css21testsuite@gtalbot.org>
Date: Mon, 4 Oct 2010 16:27:48 -0700
Message-ID: <d522f3de9fc233317fb1f581009c05a8.squirrel@cp3.shieldhost.com>
To: "Arron Eicholz" <Arron.Eicholz@microsoft.com>
Cc: "public-css-testsuite@w3.org" <public-css-testsuite@w3.org>
> On Sunday, October 03, 2010 11:13 PM Gérard Talbot wrote:
>> The following 15 RC2 testcases should be rejected for the same
>> they
>> were rejected as HTML: rendered layout is undefined when the
>> containing
>> block's width depends on the element's width (on which a percentage has
>> been set).
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/inline-replaced-width-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/block-replaced-width-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/float-replaced-width-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/absolute-replaced-
>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/inline-block-replaced-
>> I do not see how these testcases can be rehabilitated or corrected.
> I am confused how you think all these test cases are invalid.


This is my 5th email regarding these testcases. I already posted about
these testcases in september 23rd [1], then september 24th[2]. And also
at www-style mailing list.

The HTML4 version of these testcases were removed in RC2. They are still
in RC1.

If we remove a bunch of testcases in HTML4 for some reasons, then we
should do the same for the XHTML version of those testcases... unless
there is a specific [X]HTML version related reason for this.
If we keep a bunch of testcases in HTML4 for some reasons, then we
should do the same for the XHTML version of those testcases... unless
there is a specific [X]HTML version related reason for this.

> These
> tests are verifying the spec can have a case created that puts you in
> position where the sizing cannot be resolved. This is exactly what the
test is testing.

The spec clearly states that the rendering is undefined. So, it could be
a zero width, a 100px width, it could be *_any width_*. And you want to
include testcases covering such code situations leading to an undefined

If so, then we would have to start by reinstating all the testcases
which were removed because dimensions of containing block could not be
resolved and computed then according to spec. For starters:

and then
and then

And we should do this in the name of consistency.

> Since the spec has text saying something about sizing
> there needs to be tests to verify that you can create a test for the
situation. Is the rendering of such a test undefined maybe

Unfortunately, it does not say "maybe".

> , and maybe
> that is what we should say in the test.


and then imagine this:

<p>Test passes if there is no red visible on the page or if there is a
red rectangle.</p>

Would that make sense in the test suite ... or anywhere else?

> If possible maybe you and I can walk through each case an see if we
> create better cases for each scenario. Let me know and we can take
> work off the main mailing list and you and I can solve these case
issues. We need these cases it's just the cases may result in an
undefined rendering. Remember these case are not necessarily for
verifying user agents they are verifying that you can create a
> to verify the specification. Verification in any user agent is a
secondary goal of all the tests though we often promote it to a
> goal

To be transparent with you, coverage of statements indicating an
undefined layout (by the spec) does not make much sense in my mind. But
then again, if Fantasai, David Baron, Boris Zbarsky, etc. all say that
the test suite should include all these undefined layout situations,
then I am not going to oppose any of you.

Already, there are *many* Microsoft testcases (say, at least 15) that
test whether a property (width, min-width, border-collapse, etc.)
applies to elements with display: none. A waste of time, in my opinion,
otherwise definitely not very useful, helpful, meaningful IMO.

One last thing. SVG objects are embeddable into HTML4 webpages via the
HTML object element; SVG does not exclude HTML4. So the "svg" meta-flag
does not (and should not) force the "nonHTML" meta-flag.

regards, Gérard

Contributions to the CSS 2.1 test suite:

CSS 2.1 test suite (RC2; October 1st 2010):

CSS 2.1 test suite contributors:
Received on Monday, 4 October 2010 23:29:14 UTC

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