Re: Torture tests for SVG

Hi David,

> I quite like your initial test. I'd like to see it break further in browser
> though; at least Opera, as Opera almost passes it already,

Easier said than done - Opera and Batik are really the industry standard imo :)

Actually I think it's ok for browsers to pass most of the first test -
the state of the big three SVG browsers has improved dramatically in
three years.  But you might be surprised at how wrong Renesis (for
instance) gets my initial test and it will be good to see how far SVG
Web gets, as another example.  Also, if any other browsers come along
(ahem), this would be a good way to make sure a certain set of basic
things are supported.

My initial thoughts were to break the tests along SVG Tiny, Basic,
Full lines but not so sure now.  I think SMIL deserves its own test
(and that's part of SVG Tiny).  DOM definitely deserves its own test
because it's so large.


> so not much of a
> challenge if we just need to fix the iframe issue (is this certainly a bug,
> as FF does the same. I don't know too much about iFrames, but regular HTML
> pages usually have this behaviour, but I guess the SVG spec could override
> that)

I'm not sure about the iframe - my intention is to go through the SVG
Integration and HTML5 specs and see if the box where an iframe lives
is intended to be essentially the same as the box for an 'object'.  Is
this a 'browsing context' and should there be any difference between
these two in terms of size negotiation for referencing an SVG file.
If anyone wants to start this research and post their results on the
wiki that would be awesome.

Also, I think you're right about the hover interactivity for <img>, we
need to check out the SVG Integration spec and make sure.  More
research...


> Though a SVG Marilyn Manroe/Andy
> Warhol style image with filters could be pretty cool.

Neat idea :)


> Should we edit the wiki directly or send you comments?

Yes - please update the wiki directly, I definitely don't want to be
the sole driver or it just won't get done :)

Once we're all reasonably happy with the breakdown of the list we can
take bits and pieces of my initial test as a starting point and start
adding code to the googlecode svn repository.


> For test 2, I think it is good to have HTML5 there, as the spec isn't quite
> ready yet, and waiting for browsers to pass the first test will give us and
> them time to finalise that.

Ok, SGTM

> * For SVG integration we should also include list-style-image
> * Opera had some nasty break with foreignObject so that should be included
> * use element referencing external element  as well as internal
> * I use external CSS quite extensively in my SVG via CSS bg image, so I hope
> it should be allowed ;)

* list-style-image - good one.  Not sure how in the context of the
4-poster SVG image but let's figure out something.
* I was thinking of foreignObject for the 2nd test as it didn't seem
'basic' enough.  But I'm willing to be voted down :)
* Not sure about external use for the first test.  Also plugins might
be restricted on that - any thoughts?
* Yes, I'd like to see background-image in the first test for sure.


> I'd like to see filters combines with a transform, as I know Opera broke
> this recently, as we fast blitted it too aggressively for performance and
> made the edges jaggy on a shape that was rotated.

Filters might be fodder for the 2nd or 3rd test.

Regards,
Jeff

Received on Monday, 12 October 2009 19:39:50 UTC