[Webkit-unassigned] [Bug 66083] [chromium] Split LayerTextureUpdater into main thread / compositor thread versions

Thu Aug 11 14:47:56 PDT 2011

https://bugs.webkit.org/show_bug.cgi?id=66083

Adrienne Walker <enne at google.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |alokp at chromium.org

--- Comment #3 from Adrienne Walker <enne at google.com>  2011-08-11 14:47:56 PST ---
I think this is shaping up in the right direction.  :)

I'm not quite sure that I'm comfortable with m_client pointers across threads.  I'm guessing you're doing this to safely push properties via virtual dispatch rather than static_cast, but it also means that if a LayerChromium gets deleted from the tree, there's a CCLayer with a bogus pointer that will (hopefully) get cleaned up on the next tree sync.

If you're worried about type safety in pushing from one layer to another, I'd rather see some sort of virtual dispatch where you don't hold a member pointer to something on another thread and instead do it on a function arg somewhere in the tree synchronizing callstack. (That's just my thought on how to handle that.)

Also, I think the CCLayerTextureUpdaterImpl will eventually need to live on the CCLayerImpl instead of also on the LayerChromium, but I think might be better done as a part of bug 66074, given which layer tree currently gets walked to do the actual upload.

alokp: Are there any accelerated drawing tests that we can run to make sure refactorings here and in the tiler haven't broken things for you?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.