Bug 66995
| Summary: | [chromium] Eradicate unneded RefPtr usage in compositor | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Nat Duca <nduca> |
| Component: | WebCore Misc. | Assignee: | Nobody <webkit-unassigned> |
| Status: | RESOLVED WONTFIX | ||
| Severity: | Normal | CC: | enne, jamesr, nduca, schenney |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Bug Depends on: | 67440, 68121, 68135 | ||
| Bug Blocks: | |||
Nat Duca
CCLayerTreeHost does not need to be refcounted.
Nor does LayerRendererChromium.
CCLayerImpl, I don't know?
RenderSurfaces? Not if we go to a RenderPass abstraction and let the LayerRenderer manage the RenderPass list. Might be less fragile, anyway.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
James Robinson
I'm going to take a crack at CCLayerTreeHost/LayerRendererChromium.
James Robinson
https://bugs.webkit.org/show_bug.cgi?id=68121 fixes LayerRendererChromium to not be RefCounted.
CCLayerImpl is a bit tricky since currently the TreeSynchronizer relies on the refcounting. We can probably do this another way but it's not trivial.
RenderSurfaceChromium and CCRenderSurface are single ownership (and need a rewrite anyway).
CCLayerTreeHost is definitely possible, we just need to be careful to clear back pointers since the GraphicsLayers and LayerChromiums can very easily outlive the CCLayerTreeHost. I'll take a crack at this one next.
After that I think we're pretty much done.
Stephen Chenney
https://code.google.com/p/chromium/issues/detail?id=230616