On 14/04/2016 07:34, Gerry Boland wrote:
> Why did you rule out this possibility: having the MirSurfaceItem position update the position of the corresponding mir::Surface in Mir's model? If Mir's model was roughly correct, we wouldn't need to replace the default CoordinateTranslator implementation. That option strikes me as simpler, with less code in qtmir, and less cross-thread synchronisation to deal with.
A QQuickItem has no knowledge of whether its transformation
(translation, rotation and scaling) relative to the root item has
changed. Nor it should as that would be pretty bad performance-wise
(imagine moving a parent would generate a deluge of signals to all the
children in its subtree). Trying to go that way would be a mistake in my
opinion.
On 14/04/2016 07:34, Gerry Boland wrote: lator implementation. That option strikes me as simpler, with less code in qtmir, and less cross-thread synchronisation to deal with.
> Why did you rule out this possibility: having the MirSurfaceItem position update the position of the corresponding mir::Surface in Mir's model? If Mir's model was roughly correct, we wouldn't need to replace the default CoordinateTrans
A QQuickItem has no knowledge of whether its transformation
(translation, rotation and scaling) relative to the root item has
changed. Nor it should as that would be pretty bad performance-wise
(imagine moving a parent would generate a deluge of signals to all the
children in its subtree). Trying to go that way would be a mistake in my
opinion.