[libavg-devel] Div Node Rotations Patch

[Ulrich von Zadow]

Hi,

I've tried out your patch here - and while I haven't found the Words 
node error, there are one or two other things...

First of all, since this is more work than I'd initially anticipated, 
I've created a branch for the div node rotation stuff and checked in 
your current patch on to that branch:

   https://www.libavg.de/svn/branches/divrotations.

If you switch your local copy over to that branch, you can send patches 
that just contain the differences to what you've already done, which 
should make our work a bit easier. I updated the patch to work with 
current svn before checking it in, so don't be surprised if some thing 
have changed. DFBDisplayEngine is missing entirely, for instance :-). 
I've also added the rotation test to Test.py in the branch - and added a 
second rotation test which shows another clipping bug that's still in 
there somewhere. The text in rotate2.avg is displayed at the right place 
and angle, but it's clipped when it shouldn't be.

 From the short amount of time I've looked at the code, 
Node::getVisibleRect() still returns an intersection of clip rects up 
the node hierarchy, which won't work if there are rotations in between. 
For that matter, Node::calcAbsViewport() calculates the absolute 
position of a node more or less by just adding up the top left corners 
of the divs in between, and that's certainly broken now.

Hm... maybe you want to build a few more tests that exercise nested 
rotations, translations and clipping, just to get a grip on what's 
really happening?

Anyway, I really have to go back to what I was doing before, which was 
preparing the next libavg release ;-).

Regards,

   Uli

Nick Hebner schrieb:
> I have determined that the problem somehow surrounds the testWords test. 
> If this is disabled, all tests pass. Also if this test is placed first, 
> it causes a few other tests to fail in addition to testAnim. Any ideas 
> why this might be? Does the  words node change the state of the 
> rendering system in any way that might affect stencil operations?
> 
> On 8/22/07, *Nick Hebner* <hebnern at gmail.com <mailto:hebnern at gmail.com>> 
> wrote:
> 
>     Well after a few days at it, I have gotten it to the point that I am
>     only failing testAnim. I am not sure why right now as avg.avg runs
>     fine in other tests, and by itself, but not in testAnim. I will
>     continue to investigate, but have attached my current version. I
>     have moved clip rect application to DivNode::render as this is
>     currently the only place that it is applicable. I really dont see
>     much reason to push every node's visible rectangle in the
>     DisplayEngine, we just need the relavent ones (DivNodes). If anyone
>     sees the problem before I do, please let me know.
> 
>     Nick
> 
> 
> 


-- 

Ulrich von Zadow | +49-172-7872715 | cocacoder at jabber.berlin.ccc.de