W3C home > Mailing lists > Public > public-svg-wg@w3.org > April to June 2012

Re: SVG at Test The Web Forward

From: <tavmjong@free.fr>
Date: Sun, 10 Jun 2012 17:52:15 +0200 (CEST)
To: Dirk Schulze <dschulze@adobe.com>
Cc: Doug Schepers <schepers@w3.org>, SVG WG <public-svg-wg@w3.org>
Message-ID: <1192723695.131848271.1339343535022.JavaMail.root@zimbra44-e7.priv.proxad.net>

----- Mail original -----
De: "Dirk Schulze" <dschulze@adobe.com>
À: "<tavmjong@free.fr> <tavmjong@free.fr>" <tavmjong@free.fr>
Cc: "Doug Schepers" <schepers@w3.org>, "SVG WG" <public-svg-wg@w3.org>
Envoyé: Dimanche 10 Juin 2012 04:55:40
Objet: Re: SVG at Test The Web Forward

On Jun 9, 2012, at 7:10 PM, <tavmjong@free.fr> <tavmjong@free.fr> wrote:

> Hi Doug and WG,
> I have some tests in my area of the HG repository that might serve as examples. They compare CSS styling vs. using attributes for CSS3 Transforms. I think these tests are good examples as they will work in automated testing yet when viewed manually indicate what is being tested and show the expected results. The tests can be seen at:
> http://tavmjong.free.fr/SVG/svg2-tests/contributors/tavmjong/incoming/
> The first three tests are ones that I wrote, The fourth is one of the CSS tests from Adobe that I converted from an SVG in HTML test to a standalone SVG test. I prefer my tests to the Adobe tests as it is not clear from just looking at the Adobe test what it is testing. (BTW, I have a script that converts the hundreds of Adobe SVG in HTML tests to stand-alone SVG tests. I will try get them uploaded in the next few days.)
If these tests are explicitly testing a difference between CSS styling and presentation attribute styling, then they are great. Otherwise they are violating one of the main rules on ref testing. Don't test a feature with the same feature. For instance if transforms are not supported at all, all your tests would still pass. Same if just a certain transformation function is not supported, the tests would still pass - a false positive. That should be avoided as much as possible.

>Like I said, when your tests are testing the difference between CSS styling and presentation attribute they are still fine. But there are other ways to prove that without the using transforms on the ref file. We have some examples on the CSS test suite.


The tests were designed only to test the difference between CSS styling and presentation attributes. The references only use SVG 1.1 level attributes which I assumed be would be tested by the existing SVG 1.1 tests (maybe ported to the new format). Using the same mechanism avoids risk of rounding errors causing problems.


> Tav
> ----- Mail original -----
> De: "Doug Schepers" <schepers@w3.org>
> À: "SVG WG" <public-svg-wg@w3.org>
> Envoyé: Vendredi 8 Juin 2012 22:30:04
> Objet: SVG at Test The Web Forward
> Hi, SVG WG-
> As you know, I'm presenting at Adobe's Test the Web Forward event next 
> Friday.
> There are over 120 people attending this event, who will be there to 
> help make tests for W3C specs.  It's a great idea.
> I just had a call with Peter Linss and Chris Lilley about the test 
> framework and what my presentation should feature.
> We agreed that since most people aren't as familiar with SVG as with 
> CSS, the most likely tests we will get them to contribute are those that 
> are shared between SVG and CSS, such as the FX TF deliverables 
> (specifically Filters, Transforms, and Compositing, and maybe Animation 
> and Transitions), and CSS features as applied to SVG rather than HTML 
> (e.g. opacity and HSLA, and so on).
> We thought that the best outcome might be for us to provide 5-10 good 
> examples of tests we want, for use as templates, and ask them to help 
> iterate on those, along the lines of a set of features they could test. 
>  So, maybe we make 2-3 good Filters examples, and ask them to extend 
> those to other filter types, or we make a scale transform test, and 
> suggest that they do similar ones for translate, rotate, skew, and so 
> on, or variations on different values.
> Do people like this approach? If so, what tests should I use as 
> templates, and what variations should I suggest?
> Thanks-
> -Doug
Received on Sunday, 10 June 2012 15:52:45 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:20:14 UTC