Markus Stange [:mstange] Assignee
	Description • 10 years ago

On Mac, once an overlay scrollbars has been hovered, it shows the scrollbar track in its background. Async panning applies an async transform to the whole scrollbar, so the scrollbar track can shift during an async pan.
The scrollbar track should always stay in place and only the scrollbar thumb should be moved. This probably depends on bug 1009679.

	Markus Stange [:mstange] Assignee
	Comment 1 • 10 years ago

This creates an nsDisplayOwnLayer for the scrollbar thumb and sets the scroll target ID and the scrollbar type on that instead of on the outer nsDisplayOwnLayer of the scrollbar.

	Markus Stange [:mstange] Assignee
	Comment 2 • 10 years ago

The annotated container layer can now be nested inside another container layer (the one for the scrollbar itself), and for non-overlay scrollbars it can even be below the scrolled layer in z-order.
This traversal looks expensive, but I'm not that worried about performance because layer trees are usually of a manageable size, and usually the scrolled layer is somewhere near the scrollbar layer, so the search should terminate quickly in almost all cases.

	Markus Stange [:mstange] Assignee
	Comment 3 • 10 years ago

fix build after bug 508665

	Kartikaya Gupta (email:kats@mozilla.staktrace.com)
	Comment 4 • 10 years ago

Comment on attachment 8429490 [details] [diff] [review]
part 2: find the scroll target in the layer tree

Review of attachment 8429490 [details] [diff] [review]:
-----------------------------------------------------------------

::: gfx/layers/composite/AsyncCompositionManager.cpp
@@ +684,5 @@
>  
> +static Layer*
> +FindScrolledLayerForScrollbar(ContainerLayer* aLayer, bool* aOutIsAncestor)
> +{
> +  // Search all siblings of aLayer and its ancestors.

nit: s/and its/and of its/ just to make it less ambiguous

	Markus Stange [:mstange] Assignee
	Comment 5 • 10 years ago

https://hg.mozilla.org/integration/mozilla-inbound/rev/827fe00eecc7
https://hg.mozilla.org/integration/mozilla-inbound/rev/f0b1908ce71a

	Ed Morley [:emorley]
	Comment 6 • 10 years ago

Backed out for reftest failures on Windows:
https://tbpl.mozilla.org/php/getParsedLog.php?id=40555141&tree=Mozilla-Inbound
https://tbpl.mozilla.org/?tree=Mozilla-Inbound&rev=e6ab1b91f61c&jobname=win.*%20reftest%24

remote:   https://hg.mozilla.org/integration/mozilla-inbound/rev/184c660d5224
remote:   https://hg.mozilla.org/integration/mozilla-inbound/rev/575c14c47626

	Markus Stange [:mstange] Assignee
	Comment 7 • 10 years ago

The reftest fails because the two bottom scrollboxes in the left half of layout/reftests/text-overflow/xulscroll.html get less intense subpixel antialiasing. I don't yet understand how that can happen - the scrollbar thumb should already have had its own layer due to bug 1009679, I'm not sure why an additional ContainerLayer could break something.

	Markus Stange [:mstange] Assignee
	Comment 8 • 10 years ago

updated to apply on top of bug 1018309

	Markus Stange [:mstange] Assignee
	Comment 9 • 10 years ago

With bug 1019737 the test will not fail any more, because the scroll frames in that test aren't actually scrolled, so we will no longer layerize the scrollbar thumbs, and no component alpha layers will be created.

	Markus Stange [:mstange] Assignee
	Comment 10 • 10 years ago

https://hg.mozilla.org/integration/mozilla-inbound/rev/28bb18a1306b
https://hg.mozilla.org/integration/mozilla-inbound/rev/6b973f2ae664

	Markus Stange [:mstange] Assignee
	Comment 11 • 10 years ago

I forgot to initialize the fields I added to nsDisplayListBuilder. I pushed https://hg.mozilla.org/integration/mozilla-inbound/rev/0044cb69854a to fix that.

	Carsten Book [:Tomcat]
	Comment 12 • 10 years ago

https://hg.mozilla.org/mozilla-central/rev/28bb18a1306b
https://hg.mozilla.org/mozilla-central/rev/6b973f2ae664
https://hg.mozilla.org/mozilla-central/rev/0044cb69854a