-focus handling (rather large change in qt-qml to this recently)
-key handling
-tiling modes in borderimage
-anywhere there is a XXX
-there have been numerous fixes in qt-qml (+ some API changes) that need to be integrated

- Remove the unused nodes in src/scenegraph/convenience, like SolidRectnode...
- Move QxParticles into its own plugin as it is not part of the QML core API
- Remove QxPaintedItem and friends from sourcertree..

- New directory structure perhaps?
  declarative/
              canvas/
                     view
                     animation driver
              items/
                    all the qxitems
  scenegraph/
             coreapi/
                     node
                     material
                     renderer

             materials/
                      TextureMaterial
                      FlatColorMaterial
                      VertexColorMaterial

             nodes/
                   VertexColorRectNode
                   TextureColorRectNode

             qmladaptation/
                           adaptation interfaces
                           default implementations
                           qmlrenderer

- Make QxImage push normalized source coords into TextureNode so we don't have to renormalize for
  every geometry change. It also makes a lot more sense, now that TextureNode doesn't know the
  size of the original pixmap anymore...

- Fix any reference to "// ### gunnar: port properly". Some of these will be pointless to port
  as the upgrade to new QML items is pretty close, so we'll have to evaluate those individually..

- Threaded texture uploads on Mac OS X seems pretty pointless. No matter what I try it seems
  like I get a hit of 25 ms in the main thread even though shark tells me that a lot of effort
  has been spent in the background thread too. From the shark traces it looks like the driver
  does a texture memcpy for the first glDrawElements() that uses the texture in the UI thread.
  I guess that only leaves PBO uploads, but that won't remedy the problem, I recon...
  EDIT: CLIENT_STORAGE + TEXTURE_STORAGE as SHARED + seems to kill ALL performance problems