Mishandling of sublayers in RenderLayer::updateVisibilityStatus() can cause a layer that has visible content be marked as not having any.
Created attachment 12212 [details] test case
Created attachment 12214 [details] fixed My original visibility crawl loop is screwed up, it does not check nodes that are succeeding siblings of layered nodes.
Comment on attachment 12214 [details] fixed r=me
Committed in r18616 by antti with test case LayoutTests/fast/layers/layer-visibility-sublayer.html.
The layout test is invalid, giving the same results with and without the patch. Reopening the bug until a valid layout test is checked in.
Comment on attachment 12214 [details] fixed Clearing hyatt's r+ so this doesn't show up in commit queue.
Doesn't seem like the test was ever fixed but we likely have coverage elsewhere by now.
OK.