Bug 88906 – Tiled drawing means some elements can disappear behind the page

• Component: Layout and Rendering
• Status: RESOLVED FIXED

[Dean Jackson]

When tiled drawing (TileCache) is enabled on a page, and the page has preserve-3d, then elements can potentially disappear behind the tiles.

[Dean Jackson]

<rdar://problem/11601834>

[Radar WebKit Bug Importer]

<rdar://problem/11649569>

[Dean Jackson]

Created attachment 147146 [details]
Test case

[Dean Jackson]

Created attachment 147683 [details]
Patch

[Dean Jackson]

Created attachment 151121 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 151121 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=151121&action=review

> Source/WebCore/platform/graphics/ca/mac/TileCache.mm:67
> +    [m_tileContainerLayer.get() setName:@"TileCache Container Layer"];

We normally only set layer names in debug.

> Source/WebCore/rendering/RenderLayerBacking.cpp:145
> +        m_creatingPrimaryGraphicsLayer = true;

I think m_creatingPrimaryGraphicsLayer could just be a static.

> Source/WebCore/rendering/RenderLayerBacking.cpp:176
> +    m_graphicsLayer = createGraphicsLayer(layerName, true);

Not a big fan of the 'true' here.

> Source/WebCore/rendering/RenderLayerBacking.cpp:570
> +            if (m_containmentLayer && !m_usingTiledCacheLayer) {

I think the m_containmentLayer && !m_usingTiledCacheLayer thing everywhere is confusing.

Maybe add:
GraphicsLayer* tileCacheFlatteningLayer() const { return m_usingTiledCacheLayer ? m_containmentLayer : 0; }
GraphicsLayer* clippingLayer() const { return !m_usingTiledCacheLayer ? m_containmentLayer : 0; }

and use them like:

if (GraphicsLayer* clippingLayer = clippingLayer())
  clippingLayer->....

[Dean Jackson]

(In reply to comment #6)
> (From update of attachment 151121 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=151121&action=review
> 
> > Source/WebCore/platform/graphics/ca/mac/TileCache.mm:67
> > +    [m_tileContainerLayer.get() setName:@"TileCache Container Layer"];
> 
> We normally only set layer names in debug.

Done.

> 
> > Source/WebCore/rendering/RenderLayerBacking.cpp:145
> > +        m_creatingPrimaryGraphicsLayer = true;
> 
> I think m_creatingPrimaryGraphicsLayer could just be a static.

Done.

> 
> > Source/WebCore/rendering/RenderLayerBacking.cpp:176
> > +    m_graphicsLayer = createGraphicsLayer(layerName, true);
> 
> Not a big fan of the 'true' here.

I removed the default parameter, and had createPrimaryGraphicsLayer just wrap its call to createGraphicsLayer in the yucky setting of the static bool.


> > Source/WebCore/rendering/RenderLayerBacking.cpp:570
> > +            if (m_containmentLayer && !m_usingTiledCacheLayer) {
> 
> I think the m_containmentLayer && !m_usingTiledCacheLayer thing everywhere is confusing.
> 
> ...

I did this where I could. There are still come cases where we need to check m_usingTileCacheLayer directly, such as when we create the layers, but it's better now.

[Dean Jackson]

Created attachment 151167 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 151167 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=151167&action=review

> Source/WebCore/ChangeLog:11
> +        some 3d transforms, the elements could disappear behind the
> +        page background (which is rendered into the tile cache). Fix

... and could intersect with the tile cache tiles.

[Dean Jackson]

Committed r122152: <http://trac.webkit.org/changeset/122152>