W3C home > Mailing lists > Public > public-css-testsuite@w3.org > January 2014

Re: Location of my CSS Tests changed

From: Gérard Talbot <css21testsuite@gtalbot.org>
Date: Sat, 11 Jan 2014 16:15:18 -0500
To: Gabriele Romanato <gabriele.romanato@gmail.com>
Cc: CSS Test Suite <public-css-testsuite@w3.org>
Message-ID: <9501d8a362df8887f8dd38e4d4c4279a@gtalbot.org>
Le 2014-01-11 04:19, Gabriele Romanato a écrit :
> Hi all,
> now the location of my submitted tests has changed to:
> http://gabrieleromanato.name/tests/
> Also, I'm ready to submit new tests both to the CSS 2.1 and CSS3
> suites. Just let me know what are the areas of interest.
> Best wishes for the new year!
> Gabriele Romanato


I do not know when someone (me?) is going to review and approve your CSS 
2.1 tests into the CSS 2.1 test suite. I am for reviewing and approving 
*_all of your CSS 2.1 tests_* unless they are clearly duplicates of 
already submitted tests.

Preliminary comments on your tests. Please consider these only as 
preliminary comments.


a) <?xml version="1.0" encoding="utf-8" ?>
  <meta http-equiv="Content-Type" content="text/html; charset=utf-8"/>

are not needed.

b) <meta name="flags" content="Valid"/>

By default, we assume (it's a prerequisite) that all tests are using 
valid rules, valid declarations, valid CSS code, etc. So, here <meta 
name="flags" content="Valid"/> will not recognized.


c) <style type="text/css" media="screen">
media="screen" is not needed and not suitable. Some browsers pass tests 
when viewed on the monitor screen but fail (for no good reasons) when 
printed or previewed.

d) minifying CSS code (removing blank white spaces and carriage return 
and line feeds): this does not help reading, reviewing CSS code.

e) The best introductory "pass/fail conditions" sentence for tests is 
"Test passes if (...)" and then the test itself folllows. With this 
typical sentence and with this test structure, "pass/fail conditions" 
sentence of tests do not need adjectives like "following", "next", 
"below", "under" and "after" words. And people taking the test suite are 
soon familiar and expecting the top-most part of tests to start with 
"Test passes if (...)".

Addendum: see "section 11 Removable and emboldenable words"
in my ImproveTestWritingGuidelines.txt

f) The best text to use for testing is "Text sample".
There is also a green "PASS" on white background versus yellow "FAIL" on 
red background.
The best text to use for background is "Filler text".

"blocks", "box", "line" should be avoided in introductory "pass/fail 
conditions" sentences. "sentence", "text", "word" are better because 
more universally understood by non-CSS-savvy people: we want ordinary 
people to be able to take test suites without having an elementary 
knowledge of CSS. But tests themselves, in source code, can be and 
should be for CSS-savvy people, of course.
Verbs like "appear", "look", "be visible", "see", "view" should be 
avoided too.


Ideally, tests should be all taken without the need to scroll on a 
modest screen dimensions (640x480). So, ideally, you want tests to not 
require more than 400px (or so) in height.

Test Format Guidelines (Design requirements)

b) The only difference between 1st and 2nd outer squares is that 
internal elements of 1st square are floated left: everything else uses 
the same CSS code. So, the CSS code could be made to be more compact and 
more straightforward.

c) there is also another difficulty with tests like these: no humans can 
actually see if 2 rather big squares are perfectly and exactly 
identical. So, here, I would probably resort to green-overlapping-red 


<a href="../img/reference.png">reference rendering</a>
a) I would rename that image female-gender-symbol-102x173.png or 
something like that (or Venus symbol: Venus symbol (U+2640 ♀).). Note 
that the image does not use a circle.
b) images should be in a support/ folder
c) ideally, you want tests involving non-simple shapes to include the 
image-to-be-compared-with reference. So, here, I would include the 
reference image with the test and then edit "Test passes if there are 
<strong>2 identical female gender symbols</strong>."


Ideally, you want the rendered layout of a test that is passed to be 
normal, to look standard, to be expected. To assume/presume that the 
test passes if 2, 3, 1 is seen is somewhat going against normal 
expectations, predictions. I would change this to display (in case of 
success) " 1, 2, 3" or "A B C" or even "P A S S".


I see a tiny sliver of red in several browsers most likely because of 
fractional pixel and rounding issues. So this test is not best 


As the number of test suites increases (now over 35 test suites) and as 
the number of tests increases (now, over 13000), we want to reduce the 
number of 1-to-1 test-to-reftest and we want to increase the number of 
N-to-1 test-to-reftest. We want people to use and reuse already created 
and available reftests from

Doing so will reduce work, fetching, performance, maintenance, 
complexity, etc.
So, here, with tiny adjustments, the test could be coded to use


as its reftest.


With tiny adjustments, this test could be pointing to
as its reftest.


Here's the change I would propose for this test:

<style type="text/css">
p {font-size: 4em;}

p:after {content:"PASS";}

p:after {content:"\; /*" FAIL " */"}

Web authors' contributions to CSS 2.1 test suite
CSS 2.1 Test suite RC6, March 23rd 2011
Received on Saturday, 11 January 2014 21:15:55 UTC

This archive was generated by hypermail 2.4.0 : Friday, 20 January 2023 19:58:20 UTC