[meta] Implement mus-gpu to mus-ws window tree updates |
|||||||
Issue descriptionGiven the mus-ws, mus-gpu divide, it seems like we don't need nor want cc::Surfacehittest. Instead, we should ship positioning/transform data hanging off SurfaceDrawQuads back to mus-ws. This should be done over a mojo interface. As a prerequisite, we should try to find a way to avoid using type converters for CompositorFrames (this seems expensive). Right, we piggyback off the typeconverter to extract surface IDs from SurfaceDrawQuads. We should avoid walking the CompositorFrame multiple times I think (maybe?), and instead have a hook from cc/surfaces that lets us observe the walk. Maybe a SurfaceVisitor? Maybe this could be a refactor of SurfaceAggregator?
,
May 10 2016
I've already added some security checks to sanitize CompositorFrames wrt mus::Windows. A mus::Window can only refer to other mus::Window surfaces in its subtree. Thus, an arbitrary window cannot move a different arbitrary window. The window manager has the one true view of top level windows, and should not be moving them around using compositor frames, thus a malicious mus-gpu cannot move windows outside of the bounds of a top level window for the purposes of hit testing. Of course, mus-gpu can do other not so nice things, but that's no worse than the current gpu process I believe.
,
Oct 4 2016
,
Dec 3 2016
This bug refers to things that are no longer a problem but the fundamental issue still needs ot be resolved. Window Tree Updates should be pushed from Mus-gpu to Mus-ws. CompositorFrames no longer use type converters. This is very input related so I'm adding a few relevant folks.
,
Jan 22 2017
Based on our latest discussions, I don't believe we want to pursue this architecture in the short term. Closing.
,
Jun 13 2017
,
Feb 26 2018
,
Feb 26 2018
|
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by fsam...@chromium.org
, May 10 2016