import changes from `dev' branch of rmottola/Arctic-Fox:

- Bug 1196462 - Add a pref to highlight checkerboarded areas for debugging. r=kats (ffd1d24515)
- Bug 1207944 (part 1) - Pass a gfx::Color& instead of a gfxRGBA& to SetDeviceColor(). r=jwatt. (117317c610)
- Bug 1207944 (part 2) - Pass a gfx::Color& instead of a gfxRGBA& to GetDeviceColor(). r=jwatt. (82739fb35a)
- Bug 1207944 (part 3) - Add a gfxContext::SetColor method that takes a gfx::Color&. r=jwatt. (65d081203d)
- Bug 1207944 (part 4) - Use SetColor(const Color&) when setting from an nscolor. r=jwatt. (4e3ca8e36b)
- Bug 1207944 (part 5) - Use SetColor(const Color&) when setting from a constructed gfxRGBA. r=jwatt. (d4583dbee1)
- Bug 1207944 (part 6) - Use gfx::Color instead of gfxRGBA in ColorStop. r=jwatt. (74ad3407a2)
- Bug 1207944 (part 7) - Use gfx::Color instead of gfxRGBA in BlurCache. r=jwatt. (1402a3f334)
- Bug 1207944 (part 8) - Change nsDisplayBackgroundColor::mColor from gfxRGBA to gfx::Color. r=jwatt. (42259ef66b)
- Bug 1207944 (part 9) - Remove SetColor(const gfxRGBA&). r=jwatt. (064879c237)
- Bug 1208283 (part 1) - Change gfxPattern's single-arg constructor to take a gfx::Color& instead of a gfxRGBA&. r=jwatt. (1aa348d966)
- Bug 1208283 (part 2) - Change ColorLayer::mColor, ColorLayerProperties::mColor, ReadbackLayer::mBackgroundColor from gfxRBGA to gfx::Color. r=jwatt. (859d5c4a45)
- Bug 1208283 (part 3) - Remove the ToDeviceColor() that takes a gfxRGBA. r=jwatt. (525cd78338)
- Bug 1208283 (part 4) - Change FrameMetrics::mBackgroundColor from gfxRGBA to gfx::Color. r=jwatt. (49e3c39013)
- Bug 1208283 (part 5) - Pass a gfx::Color instead of a gfxRGBA to PaintRectToSurface(). r=jwatt. (04997ce877)
- Bug 1157669 - Assert aGroupTarget is not null in PaintSelfOrChildren method. r=nical (7f3cc9c792)
- Bug 1208300 (part 4) - Remove gfxRGBA and some related things. r=jwatt. (c158d04b63)
- Bug 1207931 - Remove no-longer-needed SetBackgroundColor IPC message. r=mattwoodrow. (a54d91bf53)
- Bug 1206904 - Work around APZ crashes when dealing with nested content processes. r=botond (3252333fb4)
- Bug 1189611 - Improve the APZ minimap position and ignore trivial APZ. r=kats (12c5ff5215)
- Bug 1196412 - Correctly scale canvas layers that have a transform applied to fill the screen, r=vlad (5036fd706c)
- Bug 1189473 - Don't paint the tiles in the APZ minimap. r=mstange (fc3db414f7)
- Bug 1189924 - Add a background color to the APZ minimap. r=kats (dd347db883)
- add some dead stuff back in (de8fe1d1a0)
- Treat mix-blend-mode layers as transparent for occlusion culling. (bug 1207041, r=mstange) (c3c3844d7b)

layout/ipc/RenderFrameParent.cpp

@@ -290,7 +290,6 @@ RenderFrameParent::RenderFrameParent(nsFrameLoader* aFrameLoader,
   : mLayersId(0)
   , mFrameLoader(aFrameLoader)
   , mFrameLoaderDestroyed(false)
-  , mBackgroundColor(gfxRGBA(1, 1, 1))
   , mAsyncPanZoomEnabled(false)
 {
   *aId = 0;
@@ -339,7 +338,11 @@ RenderFrameParent::GetApzcTreeManager()
   // created and the static getter knows which CompositorParent is
   // instantiated with this layers ID. That's why try to fetch it when
   // we first need it and cache the result.
-  if (!mApzcTreeManager && mAsyncPanZoomEnabled) {
+  // Note: the IsParentProcess check is to deal with nested content process
+  // scenarios, since in those cases we can have RenderFrameParent instances
+  // in a child process, but the APZC machinery is not in that process. Bug
+  // 1020199 should fix this more comprehensively.
+  if (!mApzcTreeManager && mAsyncPanZoomEnabled && XRE_IsParentProcess()) {
     mApzcTreeManager = CompositorParent::GetAPZCTreeManager(mLayersId);
   }
   return mApzcTreeManager.get();