[GTK][AC] Adding contentsLayer for image and video

[Joone Hur]

We need to add m_contentsLayer to GraphicsLayerClutter in order to render an image or video on a different ClutterActor. The patch is coming soon.

[Joone Hur]

Created attachment 196551 [details]
Patch

[ChangSeok Oh]

I leave some notes as an informal review. Please just refer to those. :)

> Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.cpp:452
> +        graphicsLayerActorInvalidateRectangle(m_contentsLayer.get(), FloatRect(FloatPoint(0, 0), size()));

I believe we don't need to call this here. Because we set ContentsNeedsDisplay below so above lines will be called in updateContentsNeedsDisplay().

> Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.cpp:701
> +        updateContentsRect();

We would be better to follow macport's order for calling above lines to avoid conflicts that might occur in the future.

> Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.cpp:716
> +        updateContentsNeedsDisplay();

ditto.

[Joone Hur]

(In reply to comment #2)
> I leave some notes as an informal review. Please just refer to those. :)
> 
> > Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.cpp:452
> > +        graphicsLayerActorInvalidateRectangle(m_contentsLayer.get(), FloatRect(FloatPoint(0, 0), size()));
> 
> I believe we don't need to call this here. Because we set ContentsNeedsDisplay below so above lines will be called in updateContentsNeedsDisplay().

Yes, right. graphicsLayerActorInvalidateRectangle is called twice. I will fix it.
 
> > Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.cpp:701
> > +        updateContentsRect();
> 
> We would be better to follow macport's order for calling above lines to avoid conflicts that might occur in the future.
> 
> > Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.cpp:716
> > +        updateContentsNeedsDisplay();
> 
> ditto.

Yes, it seems more safe to follow the Mac port's layer change order.

[Joone Hur]

Created attachment 196648 [details]
Patch

[Gustavo Noronha (kov)]

Comment on attachment 196648 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=196648&action=review

> Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.cpp:743
> +#ifndef NDEBUG
> +            clutter_actor_set_name(CLUTTER_ACTOR(m_contentsLayer.get()), "Image Layer");
> +#endif

I think we should set the name regardless of debug mode, it's not like this adds any overhead, right?

> Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.h:220
> +    // FIXME: Need to use RetainPtr

This FIXME makes no sense.

[Joone Hur]

Comment on attachment 196648 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=196648&action=review

>> Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.cpp:743
>> +#endif
> 
> I think we should set the name regardless of debug mode, it's not like this adds any overhead, right?

But, ClutterActor name is only useful to print out a PlatformLayer tree for debugging.
Is there any case where this is useful in release build?

>> Source/WebCore/platform/graphics/clutter/GraphicsLayerClutter.h:220
>> +    // FIXME: Need to use RetainPtr
> 
> This FIXME makes no sense.

Yes, RetainPtr is a specific smart pointer class of the Mac port. I will remove this comment.

[Joone Hur]

Created attachment 197186 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 197186 [details]
Patch

Clearing flags on attachment: 197186

Committed r148069: <http://trac.webkit.org/changeset/148069>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Gustavo Noronha (kov)]

(In reply to comment #6)
> But, ClutterActor name is only useful to print out a PlatformLayer tree for debugging.
> Is there any case where this is useful in release build?

Yeah, it's useful for running with CLUTTER_PAINT=paint-volumes