W3C home > Mailing lists > Public > public-css-testsuite@w3.org > October 2012

RE: New tests submission for css3 background-clip/-origin/-size

From: Gérard Talbot <css21testsuite@gtalbot.org>
Date: Mon, 15 Oct 2012 17:54:07 -0400
Message-ID: <b8d122c6c052f2727041f0dccc0b3285.squirrel@ed-sh-cp3.entirelydigital.com>
To: "Zhang, Zhiqiang" <zhiqiang.zhang@intel.com>
Cc: "Public CSS testsuite mailing list" <public-css-testsuite@w3.org>, "Christensen, Kenneth" <kenneth.christensen@intel.com>, "Zhang, Haili" <haili.zhang@intel.com>, "Yu, Ling L" <ling.l.yu@intel.com>, "Yang, Lei A" <lei.a.yang@intel.com>

Le Dim 14 octobre 2012 23:04, Zhang, Zhiqiang a écrit :
> Gérard,
> Thank you very much on your feedbacks. I really appreciate your review
> of my tests.
>> Zhiqiang,
>> Here's a preliminary review of 1 test.
>> http://test.csswg.org/source/contributors/intel/submitted/css3-
>> background/background-334.html
>> 1-
>> a)
>> The test is obviously about background shorthand: so the title element
>> should say so.
> Updated @ https://hg.csswg.org/test/rev/4a42bfba3b60


I wonder why you unite, you join all the words in the last part of the
title with underscore characters.

Eg. background_shorthand_specified_color_image
eg. background_specified_box_two

>> b)
>> "
>> For specifications other than CSS 2.1, you can include the module name
>> somewhere before the colon, like "CSS Selectors Test:" or "CSS Test
>> (Selectors):". Do not include the module version number, since the
>> test
>> might get reused for the next version.
>> "
>> http://wiki.csswg.org/test/format#title-element
>> So,
>> <title>CSS Test (Backgrounds and Borders): (...)</title>
>> or
>> <title>CSS Backgrounds and Borders Test: (...)</title>
> Updated @ https://hg.csswg.org/test/rev/b3381a9594e0
>> are okay.
>> 2-
>> The test design and assert text suggest that background-size '100%
>> auto'
>> will imply a rescale of the background-image to fill the background
>> area.
>> 3-
>>             top: 1px;
>>             left: 1px;
>> I wondered why top and left 1px.
>>             width: 158px;
>>             height: 158px;
>>             background: red;
>>         }
>>         #test {
>>             position: absolute;
>>             top: 0px;
>>             left: 0px;
>>             width: 100px;
>>             height: 100px;
>>             padding: 25px;
>>             border: 5px dotted #000000;
>> The overlapping #test generates a 160px wide and 160px tall CSS box
>> while the red-overlapped #ref generates a 158px by 158px box; the only
>> possible reason I can imagine would be to counter/take into account
>> anti-aliasing.
> Yes. This is referred to
> http://test.csswg.org/source/contributors/ttwf/samples/ttwf-reftest-tutorial-001.html,
> and I think it's a good practice.

Well,... hmm... I think there is an issue here. The test allocates a
margin for error of 2px which is for anti-aliasing. We discussed before
about anti-aliasing: thread started with


although we did not discuss about allocating a 1px or 2px of margin of
error in the tests.

As far as I know and understand this, anti-aliasing will mostly occur
with oblique lines, not for straight vertical lines and not for straight
horizontal lines. Anti-aliasing may occur with glyphs, in particular
oblique ones like "V" or "W", "Y", etc. and curvy glyphs.

In the relativity, 2px may be a lot.

Let me give you an example of where such 2px may not be a good thing:


Firefox 16.0.1, Chrome 22.0.1229.94 and Konqueror 4.9.2: displays no red

Opera 12.02: there is a 1px vertical line of red on the right and a 1px
horizontal line of red at the bottom. Most likely because of rounding
effects: the scaling from 60px to 100px is not accurate. My initial
thought was that this should be considered as a bug. For practical
reasons, I would also believe this is a bug. Users or developers would
eventually report this to Opera BTS as a bug and complaint that
browserX, browserY and browserZ don't display red.

Now, to make things furthermore complicated, let's add
'background-color: #CCC' (which is the background-color originally
included in background-334) to that test. We have:


and now, it *appears* that Opera 12.02 passes the test but in reality it
does not... and the test would probably be reported as failed when/with
automated-comparing to its reftest.

Contributions to the CSS 2.1 test suite:

CSS 2.1 Test suite RC6, March 23rd 2011:

CSS 2.1 test suite harness:

Contributing to to CSS 2.1 test suite:
Received on Monday, 15 October 2012 21:54:40 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 15 October 2012 21:54:42 GMT