gecko-dev/layout/ipc/RenderFrameParent.cpp

878 строки
31 KiB
C++
Исходник Обычный вид История

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*-
* vim: sw=2 ts=8 et :
*/
/* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at:
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is Mozilla Code.
*
* The Initial Developer of the Original Code is
* The Mozilla Foundation
* Portions created by the Initial Developer are Copyright (C) 2010
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
* Chris Jones <jones.chris.g@gmail.com>
*
* Alternatively, the contents of this file may be used under the terms of
* either the GNU General Public License Version 2 or later (the "GPL"), or
* the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
* in which case the provisions of the GPL or the LGPL are applicable instead
* of those above. If you wish to allow use of your version of this file only
* under the terms of either the GPL or the LGPL, and not to allow others to
* use your version of this file under the terms of the MPL, indicate your
* decision by deleting the provisions above and replace them with the notice
* and other provisions required by the GPL or the LGPL. If you do not delete
* the provisions above, a recipient may use your version of this file under
* the terms of any one of the MPL, the GPL or the LGPL.
*
* ***** END LICENSE BLOCK ***** */
#include "mozilla/layers/ShadowLayersParent.h"
#include "BasicLayers.h"
#include "LayerManagerOGL.h"
#ifdef MOZ_ENABLE_D3D9_LAYER
#include "LayerManagerD3D9.h"
#endif //MOZ_ENABLE_D3D9_LAYER
#include "RenderFrameParent.h"
#include "gfx3DMatrix.h"
#include "nsFrameLoader.h"
#include "nsViewportFrame.h"
#include "nsSubDocumentFrame.h"
#include "nsIObserver.h"
typedef nsContentView::ViewConfig ViewConfig;
using namespace mozilla::layers;
namespace mozilla {
namespace layout {
typedef FrameMetrics::ViewID ViewID;
typedef RenderFrameParent::ViewMap ViewMap;
nsRefPtr<ImageContainer> sCheckerboard = nsnull;
class CheckerBoardPatternDeleter : public nsIObserver
{
public:
NS_DECL_NSIOBSERVER
NS_DECL_ISUPPORTS
};
NS_IMPL_ISUPPORTS1(CheckerBoardPatternDeleter, nsIObserver)
NS_IMETHODIMP
CheckerBoardPatternDeleter::Observe(nsISupports* aSubject,
const char* aTopic,
const PRUnichar* aData)
{
if (!strcmp(aTopic, "xpcom-shutdown")) {
sCheckerboard = nsnull;
}
return NS_OK;
}
// Represents (affine) transforms that are calculated from a content view.
struct ViewTransform {
ViewTransform(nsIntPoint aTranslation = nsIntPoint(0, 0), float aXScale = 1, float aYScale = 1)
: mTranslation(aTranslation)
, mXScale(aXScale)
, mYScale(aYScale)
{}
operator gfx3DMatrix() const
{
return
gfx3DMatrix::Scale(mXScale, mYScale, 1) *
gfx3DMatrix::Translation(mTranslation.x, mTranslation.y, 0);
}
nsIntPoint mTranslation;
float mXScale;
float mYScale;
};
// Matrix helpers
// For our simple purposes, these helpers apply to 2D affine transformations
// that can be represented by a scale and a translation. This makes the math
// much easier because we only expect the diagonals and the translation
// coordinates of the matrix to be non-zero.
static double GetXScale(const gfx3DMatrix& aTransform)
{
return aTransform._11;
}
static double GetYScale(const gfx3DMatrix& aTransform)
{
return aTransform._22;
}
static void Scale(gfx3DMatrix& aTransform, double aXScale, double aYScale)
{
aTransform._11 *= aXScale;
aTransform._22 *= aYScale;
}
static void ReverseTranslate(gfx3DMatrix& aTransform, ViewTransform& aViewTransform)
{
aTransform._41 -= aViewTransform.mTranslation.x / aViewTransform.mXScale;
aTransform._42 -= aViewTransform.mTranslation.y / aViewTransform.mYScale;
}
static void ApplyTransform(nsRect& aRect,
gfx3DMatrix& aTransform,
nscoord auPerDevPixel)
{
aRect.x = aRect.x * aTransform._11 + aTransform._41 * auPerDevPixel;
aRect.y = aRect.y * aTransform._22 + aTransform._42 * auPerDevPixel;
aRect.width = aRect.width * aTransform._11;
aRect.height = aRect.height * aTransform._22;
}
static void
AssertInTopLevelChromeDoc(ContainerLayer* aContainer,
nsIFrame* aContainedFrame)
{
NS_ASSERTION(
(aContainer->Manager()->GetBackendType() != LayerManager::LAYERS_BASIC) ||
(aContainedFrame->GetNearestWidget() ==
static_cast<BasicLayerManager*>(aContainer->Manager())->GetRetainerWidget()),
"Expected frame to be in top-level chrome document");
}
// Return view for given ID in aArray, NULL if not found.
static nsContentView*
FindViewForId(const ViewMap& aMap, ViewID aId)
{
ViewMap::const_iterator iter = aMap.find(aId);
return iter != aMap.end() ? iter->second : NULL;
}
static const FrameMetrics*
GetFrameMetrics(Layer* aLayer)
{
ContainerLayer* container = aLayer->AsContainerLayer();
return container ? &container->GetFrameMetrics() : NULL;
}
static nsIntPoint
GetRootFrameOffset(nsIFrame* aContainerFrame, nsDisplayListBuilder* aBuilder)
{
nscoord auPerDevPixel = aContainerFrame->PresContext()->AppUnitsPerDevPixel();
// Offset to the content rect in case we have borders or padding
nsPoint frameOffset =
(aBuilder->ToReferenceFrame(aContainerFrame->GetParent()) +
aContainerFrame->GetContentRect().TopLeft());
return frameOffset.ToNearestPixels(auPerDevPixel);
}
// Compute the transform of the shadow tree contained by
// |aContainerFrame| to widget space. We transform because the
// subprocess layer manager renders to a different top-left than where
// the shadow tree is drawn here and because a scale can be set on the
// shadow tree.
static ViewTransform
ComputeShadowTreeTransform(nsIFrame* aContainerFrame,
nsFrameLoader* aRootFrameLoader,
const FrameMetrics* aMetrics,
const ViewConfig& aConfig,
float aInverseScaleX,
float aInverseScaleY)
{
// |aMetrics->mViewportScrollOffset| The frame's scroll offset when it was
// painted, in content document pixels.
// |aConfig.mScrollOffset| What our user expects, or wants, the
// frame scroll offset to be in chrome
// document app units.
//
// So we set a compensating translation that moves the content document
// pixels to where the user wants them to be.
//
nscoord auPerDevPixel = aContainerFrame->PresContext()->AppUnitsPerDevPixel();
nsIntPoint scrollOffset =
aConfig.mScrollOffset.ToNearestPixels(auPerDevPixel);
// metricsScrollOffset is in layer coordinates.
nsIntPoint metricsScrollOffset = aMetrics->mViewportScrollOffset;
if (aRootFrameLoader->AsyncScrollEnabled() && !aMetrics->mDisplayPort.IsEmpty()) {
// Only use asynchronous scrolling if it is enabled and there is a
// displayport defined. It is useful to have a scroll layer that is
// synchronously scrolled for identifying a scroll area before it is
// being actively scrolled.
nsIntPoint scrollCompensation(
scrollOffset.x * aInverseScaleX - metricsScrollOffset.x * aConfig.mXScale,
scrollOffset.y * aInverseScaleY - metricsScrollOffset.y * aConfig.mYScale);
return ViewTransform(-scrollCompensation, aConfig.mXScale, aConfig.mYScale);
} else {
return ViewTransform(nsIntPoint(0, 0), 1, 1);
}
}
// Use shadow layer tree to build display list for the browser's frame.
static void
BuildListForLayer(Layer* aLayer,
nsFrameLoader* aRootFrameLoader,
const gfx3DMatrix& aTransform,
nsDisplayListBuilder* aBuilder,
nsDisplayList& aShadowTree,
nsIFrame* aSubdocFrame)
{
const FrameMetrics* metrics = GetFrameMetrics(aLayer);
gfx3DMatrix transform;
if (metrics && metrics->IsScrollable()) {
const ViewID scrollId = metrics->mScrollId;
// We need to figure out the bounds of the scrollable region using the
// shadow layer tree from the remote process. The metrics viewport is
// defined based on all the transformations of its parent layers and
// the scale of the current layer.
// Calculate transform for this layer.
nsContentView* view =
aRootFrameLoader->GetCurrentRemoteFrame()->GetContentView(scrollId);
// XXX why don't we include aLayer->GetTransform() in the inverse-scale here?
// This seems wrong, but it doesn't seem to cause bugs!
gfx3DMatrix applyTransform = ComputeShadowTreeTransform(
aSubdocFrame, aRootFrameLoader, metrics, view->GetViewConfig(),
1 / GetXScale(aTransform), 1 / GetYScale(aTransform));
transform = applyTransform * aLayer->GetTransform() * aTransform;
// As mentioned above, bounds calculation also depends on the scale
// of this layer.
gfx3DMatrix tmpTransform = aTransform;
Scale(tmpTransform, GetXScale(applyTransform), GetYScale(applyTransform));
// Calculate rect for this layer based on aTransform.
nsRect bounds;
{
nscoord auPerDevPixel = aSubdocFrame->PresContext()->AppUnitsPerDevPixel();
bounds = metrics->mViewport.ToAppUnits(auPerDevPixel);
ApplyTransform(bounds, tmpTransform, auPerDevPixel);
}
aShadowTree.AppendToTop(
new (aBuilder) nsDisplayRemoteShadow(aBuilder, aSubdocFrame, bounds, scrollId));
} else {
transform = aLayer->GetTransform() * aTransform;
}
for (Layer* child = aLayer->GetFirstChild(); child;
child = child->GetNextSibling()) {
BuildListForLayer(child, aRootFrameLoader, transform,
aBuilder, aShadowTree, aSubdocFrame);
}
}
// Go down shadow layer tree and apply transformations for scrollable layers.
static void
TransformShadowTree(nsDisplayListBuilder* aBuilder, nsFrameLoader* aFrameLoader,
nsIFrame* aFrame, Layer* aLayer,
const ViewTransform& aTransform)
{
ShadowLayer* shadow = aLayer->AsShadowLayer();
shadow->SetShadowClipRect(aLayer->GetClipRect());
shadow->SetShadowVisibleRegion(aLayer->GetVisibleRegion());
const FrameMetrics* metrics = GetFrameMetrics(aLayer);
gfx3DMatrix shadowTransform;
ViewTransform layerTransform = aTransform;
if (metrics && metrics->IsScrollable()) {
const ViewID scrollId = metrics->mScrollId;
const nsContentView* view =
aFrameLoader->GetCurrentRemoteFrame()->GetContentView(scrollId);
NS_ABORT_IF_FALSE(view, "Array of views should be consistent with layer tree");
const gfx3DMatrix& currentTransform = aLayer->GetTransform();
ViewTransform viewTransform = ComputeShadowTreeTransform(
aFrame, aFrameLoader, metrics, view->GetViewConfig(),
1 / (GetXScale(currentTransform)*layerTransform.mXScale),
1 / (GetYScale(currentTransform)*layerTransform.mYScale)
);
// Apply the layer's own transform *before* the view transform
shadowTransform = gfx3DMatrix(viewTransform) * currentTransform;
if (metrics->IsRootScrollable()) {
layerTransform.mTranslation = viewTransform.mTranslation;
// Apply the root frame translation *before* we do the rest of the transforms.
nsIntPoint rootFrameOffset = GetRootFrameOffset(aFrame, aBuilder);
shadowTransform = shadowTransform *
gfx3DMatrix::Translation(float(rootFrameOffset.x), float(rootFrameOffset.y), 0.0);
}
} else {
shadowTransform = aLayer->GetTransform();
}
if (aLayer->GetIsFixedPosition() &&
!aLayer->GetParent()->GetIsFixedPosition()) {
ReverseTranslate(shadowTransform, layerTransform);
const nsIntRect* clipRect = shadow->GetShadowClipRect();
if (clipRect) {
nsIntRect transformedClipRect(*clipRect);
transformedClipRect.MoveBy(shadowTransform._41, shadowTransform._42);
shadow->SetShadowClipRect(&transformedClipRect);
}
}
shadow->SetShadowTransform(shadowTransform);
layerTransform.mXScale *= GetXScale(shadowTransform);
layerTransform.mYScale *= GetYScale(shadowTransform);
for (Layer* child = aLayer->GetFirstChild();
child; child = child->GetNextSibling()) {
TransformShadowTree(aBuilder, aFrameLoader, aFrame, child, layerTransform);
}
}
static void
ClearContainer(ContainerLayer* aContainer)
{
while (Layer* layer = aContainer->GetFirstChild()) {
aContainer->RemoveChild(layer);
}
}
// Return true iff |aManager| is a "temporary layer manager". They're
// used for small software rendering tasks, like drawWindow. That's
// currently implemented by a BasicLayerManager without a backing
// widget, and hence in non-retained mode.
static PRBool
IsTempLayerManager(LayerManager* aManager)
{
return (LayerManager::LAYERS_BASIC == aManager->GetBackendType() &&
!static_cast<BasicLayerManager*>(aManager)->IsRetained());
}
// Recursively create a new array of scrollables, preserving any scrollables
// that are still in the layer tree.
//
// aXScale and aYScale are used to calculate any values that need to be in
// chrome-document CSS pixels and aren't part of the rendering loop, such as
// the initial scroll offset for a new view.
static void
BuildViewMap(ViewMap& oldContentViews, ViewMap& newContentViews,
nsFrameLoader* aFrameLoader, Layer* aLayer,
float aXScale = 1, float aYScale = 1)
{
ContainerLayer* container = aLayer->AsContainerLayer();
if (!container)
return;
const FrameMetrics metrics = container->GetFrameMetrics();
const ViewID scrollId = metrics.mScrollId;
const gfx3DMatrix transform = aLayer->GetTransform();
aXScale *= GetXScale(transform);
aYScale *= GetYScale(transform);
if (metrics.IsScrollable()) {
nscoord auPerDevPixel = aFrameLoader->GetPrimaryFrameOfOwningContent()
->PresContext()->AppUnitsPerDevPixel();
nsContentView* view = FindViewForId(oldContentViews, scrollId);
if (view) {
// View already exists. Be sure to propagate scales for any values
// that need to be calculated something in chrome-doc CSS pixels.
ViewConfig config = view->GetViewConfig();
aXScale *= config.mXScale;
aYScale *= config.mYScale;
view->mFrameLoader = aFrameLoader;
} else {
// View doesn't exist, so generate one. We start the view scroll offset at
// the same position as the framemetric's scroll offset from the layer.
// The default scale is 1, so no need to propagate scale down.
ViewConfig config;
config.mScrollOffset = nsPoint(
NSIntPixelsToAppUnits(metrics.mViewportScrollOffset.x, auPerDevPixel) * aXScale,
NSIntPixelsToAppUnits(metrics.mViewportScrollOffset.y, auPerDevPixel) * aYScale);
view = new nsContentView(aFrameLoader, scrollId, config);
}
view->mViewportSize = nsSize(
NSIntPixelsToAppUnits(metrics.mViewport.width, auPerDevPixel) * aXScale,
NSIntPixelsToAppUnits(metrics.mViewport.height, auPerDevPixel) * aYScale);
view->mContentSize = nsSize(
NSIntPixelsToAppUnits(metrics.mContentSize.width, auPerDevPixel) * aXScale,
NSIntPixelsToAppUnits(metrics.mContentSize.height, auPerDevPixel) * aYScale);
newContentViews[scrollId] = view;
}
for (Layer* child = aLayer->GetFirstChild();
child; child = child->GetNextSibling()) {
BuildViewMap(oldContentViews, newContentViews, aFrameLoader, child,
aXScale, aYScale);
}
}
#define BOARDSIZE 32
#define CHECKERSIZE 16
already_AddRefed<gfxASurface>
GetBackgroundImage()
{
static unsigned int data[BOARDSIZE * BOARDSIZE];
static bool initialized = false;
if (!initialized) {
initialized = true;
for (unsigned int y = 0; y < BOARDSIZE; y++) {
for (unsigned int x = 0; x < BOARDSIZE; x++) {
bool col_odd = (x / CHECKERSIZE) & 1;
bool row_odd = (y / CHECKERSIZE) & 1;
if (col_odd ^ row_odd) { // xor
data[y * BOARDSIZE + x] = 0xFFFFFFFF;
}
else {
data[y * BOARDSIZE + x] = 0xFFDDDDDD;
}
}
}
}
nsRefPtr<gfxASurface> s =
new gfxImageSurface((unsigned char*) data,
gfxIntSize(BOARDSIZE, BOARDSIZE),
BOARDSIZE * sizeof(unsigned int),
gfxASurface::ImageFormatARGB32);
return s.forget();
}
static void
BuildBackgroundPatternFor(ContainerLayer* aContainer,
ContainerLayer* aShadowRoot,
const FrameMetrics& aMetrics,
const ViewConfig& aConfig,
LayerManager* aManager,
nsIFrame* aFrame,
nsDisplayListBuilder* aBuilder)
{
// We tile a visible region that is the frame's area \setminus the
// rect in our frame onto which valid pixels from remote content
// will be drawn. It's just a waste of CPU cycles to draw a
// checkerboard behind that content.
//
// We want to give the background the illusion of moving while the
// user pans, so we nudge the tiling area a bit based on the
// "desired" scroll offset.
//
// The background-image layer is added to the layer tree "behind"
// the shadow tree. It doesn't matter in theory which is behind/in
// front, except that having the background in front of content
// means we have to be more careful about snapping boundaries,
// whereas having it behind allows us to trade off simplicity for
// "wasted" drawing of a few extra pixels.
ShadowLayer* shadowRoot = aShadowRoot->AsShadowLayer();
gfxMatrix t;
if (!shadowRoot->GetShadowTransform().Is2D(&t)) {
return;
}
// Get the rect bounding the shadow content, transformed into the
// same space as |aFrame|
nsIntRect contentBounds = shadowRoot->GetShadowVisibleRegion().GetBounds();
gfxRect contentVis(contentBounds.x, contentBounds.y,
contentBounds.width, contentBounds.height);
gfxRect localContentVis(t.Transform(contentVis));
// Round *in* here because this area is punched out of the background
localContentVis.RoundIn();
nsIntRect localIntContentVis(localContentVis.X(), localContentVis.Y(),
localContentVis.Width(), localContentVis.Height());
// Get the frame's rect
nscoord auPerDevPixel = aFrame->PresContext()->AppUnitsPerDevPixel();
nsIntRect frameRect = aFrame->GetRect().ToOutsidePixels(auPerDevPixel);
// If the shadow tree covers the frame rect, don't bother building
// the background, it wouldn't be visible
if (localIntContentVis.Contains(frameRect)) {
return;
}
nsRefPtr<gfxASurface> bgImage = GetBackgroundImage();
gfxIntSize bgImageSize = bgImage->GetSize();
// Set up goop needed to get a cairo image into its own layer
if (!sCheckerboard) {
sCheckerboard = aManager->CreateImageContainer().get();
const Image::Format fmts[] = { Image::CAIRO_SURFACE };
nsRefPtr<Image> img = sCheckerboard->CreateImage(fmts, 1);
CairoImage::Data data = { bgImage.get(), bgImageSize };
static_cast<CairoImage*>(img.get())->SetData(data);
sCheckerboard->SetCurrentImage(img);
nsCOMPtr<nsIObserverService> observerService =
mozilla::services::GetObserverService();
if (!observerService) {
return;
}
nsresult rv = observerService->AddObserver(new CheckerBoardPatternDeleter, "xpcom-shutdown", PR_FALSE);
if (NS_FAILED(rv)) {
return;
}
}
nsRefPtr<ImageLayer> layer = aManager->CreateImageLayer();
layer->SetContainer(sCheckerboard);
// The tile source is the entire background image
nsIntRect tileSource(0, 0, bgImageSize.width, bgImageSize.height);
layer->SetTileSourceRect(&tileSource);
// The origin of the tiling plane, top-left of the tile source rect,
// is at layer-space point <0,0>. Set up a translation from that
// origin to the frame top-left, with the little nudge included.
nsIntPoint translation = frameRect.TopLeft();
nsIntPoint panNudge = aConfig.mScrollOffset.ToNearestPixels(auPerDevPixel);
// This offset must be positive to ensure that the tiling rect
// contains the frame's visible rect. The "desired" scroll offset
// is allowed to be negative, however, so we fix that up here.
panNudge.x = (panNudge.x % bgImageSize.width);
if (panNudge.x < 0) panNudge.x += bgImageSize.width;
panNudge.y = (panNudge.y % bgImageSize.height);
if (panNudge.y < 0) panNudge.y += bgImageSize.height;
translation -= panNudge;
layer->SetTransform(gfx3DMatrix::Translation(translation.x, translation.y, 0));
// The visible area of the background is the frame's area minus the
// content area
nsIntRegion bgRgn(frameRect);
bgRgn.Sub(bgRgn, localIntContentVis);
bgRgn.MoveBy(-translation);
layer->SetVisibleRegion(bgRgn);
aContainer->InsertAfter(layer, nsnull);
}
RenderFrameParent::RenderFrameParent(nsFrameLoader* aFrameLoader)
: mFrameLoader(aFrameLoader)
{
if (aFrameLoader) {
mContentViews[FrameMetrics::ROOT_SCROLL_ID] =
new nsContentView(aFrameLoader, FrameMetrics::ROOT_SCROLL_ID);
}
}
RenderFrameParent::~RenderFrameParent()
{}
void
RenderFrameParent::Destroy()
{
size_t numChildren = ManagedPLayersParent().Length();
NS_ABORT_IF_FALSE(0 == numChildren || 1 == numChildren,
"render frame must only have 0 or 1 layer manager");
if (numChildren) {
ShadowLayersParent* layers =
static_cast<ShadowLayersParent*>(ManagedPLayersParent()[0]);
layers->Destroy();
}
}
nsContentView*
RenderFrameParent::GetContentView(ViewID aId)
{
return FindViewForId(mContentViews, aId);
}
void
RenderFrameParent::ContentViewScaleChanged(nsContentView* aView)
{
// Since the scale has changed for a view, it and its descendents need their
// shadow-space attributes updated. It's easiest to rebuild the view map.
BuildViewMap();
}
void
RenderFrameParent::ShadowLayersUpdated()
{
mFrameLoader->SetCurrentRemoteFrame(this);
// View map must only contain views that are associated with the current
// shadow layer tree. We must always update the map when shadow layers
// are updated.
BuildViewMap();
nsIFrame* docFrame = mFrameLoader->GetPrimaryFrameOfOwningContent();
if (!docFrame) {
// Bad, but nothing we can do about it (XXX/cjones: or is there?
// maybe bug 589337?). When the new frame is created, we'll
// probably still be the current render frame and will get to draw
// our content then. Or, we're shutting down and this update goes
// to /dev/null.
return;
}
// FIXME/cjones: we should collect the rects/regions updated for
// Painted*Layer() calls and pass that region to here, then only
// invalidate that rect
//
// We pass INVALIDATE_NO_THEBES_LAYERS here because we're
// invalidating the <browser> on behalf of its counterpart in the
// content process. Not only do we not need to invalidate the
// shadow layers, things would just break if we did --- we have no
// way to repaint shadow layers from this process.
nsRect rect = nsRect(nsPoint(0, 0), docFrame->GetRect().Size());
docFrame->InvalidateWithFlags(rect, nsIFrame::INVALIDATE_NO_THEBES_LAYERS);
}
already_AddRefed<Layer>
RenderFrameParent::BuildLayer(nsDisplayListBuilder* aBuilder,
nsIFrame* aFrame,
LayerManager* aManager,
const nsIntRect& aVisibleRect)
{
NS_ABORT_IF_FALSE(aFrame,
"makes no sense to have a shadow tree without a frame");
NS_ABORT_IF_FALSE(!mContainer ||
IsTempLayerManager(aManager) ||
mContainer->Manager() == aManager,
"retaining manager changed out from under us ... HELP!");
if (mContainer && mContainer->Manager() != aManager) {
// This can happen if aManager is a "temporary" manager, or if the
// widget's layer manager changed out from under us. We need to
// FIXME handle the former case somehow, probably with an API to
// draw a manager's subtree. The latter is bad bad bad, but the
// the NS_ABORT_IF_FALSE() above will flag it. Returning NULL
// here will just cause the shadow subtree not to be rendered.
return nsnull;
}
if (mContainer) {
ClearContainer(mContainer);
}
ContainerLayer* shadowRoot = GetRootLayer();
if (!shadowRoot) {
mContainer = nsnull;
return nsnull;
}
NS_ABORT_IF_FALSE(!shadowRoot || shadowRoot->Manager() == aManager,
"retaining manager changed out from under us ... HELP!");
// Wrap the shadow layer tree in mContainer.
if (!mContainer) {
mContainer = aManager->CreateContainerLayer();
}
NS_ABORT_IF_FALSE(!mContainer->GetFirstChild(),
"container of shadow tree shouldn't have a 'root' here");
mContainer->InsertAfter(shadowRoot, nsnull);
AssertInTopLevelChromeDoc(mContainer, aFrame);
ViewTransform transform;
TransformShadowTree(aBuilder, mFrameLoader, aFrame, shadowRoot, transform);
mContainer->SetClipRect(nsnull);
if (mFrameLoader->AsyncScrollEnabled()) {
const nsContentView* view = GetContentView(FrameMetrics::ROOT_SCROLL_ID);
BuildBackgroundPatternFor(mContainer,
shadowRoot,
shadowRoot->GetFrameMetrics(),
view->GetViewConfig(),
aManager, aFrame, aBuilder);
}
mContainer->SetVisibleRegion(aVisibleRect);
return nsRefPtr<Layer>(mContainer).forget();
}
void
RenderFrameParent::OwnerContentChanged(nsIContent* aContent)
{
NS_ABORT_IF_FALSE(mFrameLoader->GetOwnerContent() == aContent,
"Don't build new map if owner is same!");
BuildViewMap();
}
void
RenderFrameParent::ActorDestroy(ActorDestroyReason why)
{
if (mFrameLoader && mFrameLoader->GetCurrentRemoteFrame() == this) {
// XXX this might cause some weird issues ... we'll just not
// redraw the part of the window covered by this until the "next"
// remote frame has a layer-tree transaction. For
// why==NormalShutdown, we'll definitely want to do something
// better, especially as nothing guarantees another Update() from
// the "next" remote layer tree.
mFrameLoader->SetCurrentRemoteFrame(nsnull);
}
mFrameLoader = nsnull;
}
PLayersParent*
RenderFrameParent::AllocPLayers(LayerManager::LayersBackend* aBackendType)
{
if (!mFrameLoader) {
*aBackendType = LayerManager::LAYERS_NONE;
return nsnull;
}
2011-07-15 22:35:55 +04:00
LayerManager* lm = GetLayerManager();
switch (*aBackendType = lm->GetBackendType()) {
2011-07-15 22:35:55 +04:00
case LayerManager::LAYERS_BASIC: {
BasicShadowLayerManager* bslm = static_cast<BasicShadowLayerManager*>(lm);
return new ShadowLayersParent(bslm);
}
case LayerManager::LAYERS_OPENGL: {
LayerManagerOGL* lmo = static_cast<LayerManagerOGL*>(lm);
return new ShadowLayersParent(lmo);
}
#ifdef MOZ_ENABLE_D3D9_LAYER
case LayerManager::LAYERS_D3D9: {
LayerManagerD3D9* lmd3d9 = static_cast<LayerManagerD3D9*>(lm);
return new ShadowLayersParent(lmd3d9);
}
#endif //MOZ_ENABLE_D3D9_LAYER
default: {
NS_WARNING("shadow layers no sprechen D3D backend yet");
*aBackendType = LayerManager::LAYERS_NONE;
return nsnull;
}
}
}
bool
RenderFrameParent::DeallocPLayers(PLayersParent* aLayers)
{
delete aLayers;
return true;
}
void
RenderFrameParent::BuildViewMap()
{
ViewMap newContentViews;
// BuildViewMap assumes we have a primary frame, which may not be the case.
if (GetRootLayer() && mFrameLoader->GetPrimaryFrameOfOwningContent()) {
// Some of the content views in our hash map may no longer be active. To
// tag them as inactive and to remove any chance of them using a dangling
// pointer, we set mContentView to NULL.
//
// BuildViewMap will restore mFrameLoader if the content view is still
// in our hash table.
for (ViewMap::const_iterator iter = mContentViews.begin();
iter != mContentViews.end();
++iter) {
iter->second->mFrameLoader = NULL;
}
mozilla::layout::BuildViewMap(mContentViews, newContentViews, mFrameLoader, GetRootLayer());
}
// Here, we guarantee that *only* the root view is preserved in
// case we couldn't build a new view map above. This is important because
// the content view map should only contain the root view and content
// views that are present in the layer tree.
if (newContentViews.empty()) {
newContentViews[FrameMetrics::ROOT_SCROLL_ID] =
FindViewForId(mContentViews, FrameMetrics::ROOT_SCROLL_ID);
}
mContentViews = newContentViews;
}
LayerManager*
RenderFrameParent::GetLayerManager() const
{
nsIDocument* doc = mFrameLoader->GetOwnerDoc();
return doc->GetShell()->GetLayerManager();
}
ShadowLayersParent*
RenderFrameParent::GetShadowLayers() const
{
const nsTArray<PLayersParent*>& shadowParents = ManagedPLayersParent();
NS_ABORT_IF_FALSE(shadowParents.Length() <= 1,
"can only support at most 1 ShadowLayersParent");
return (shadowParents.Length() == 1) ?
static_cast<ShadowLayersParent*>(shadowParents[0]) : nsnull;
}
ContainerLayer*
RenderFrameParent::GetRootLayer() const
{
ShadowLayersParent* shadowLayers = GetShadowLayers();
return shadowLayers ? shadowLayers->GetRoot() : nsnull;
}
NS_IMETHODIMP
RenderFrameParent::BuildDisplayList(nsDisplayListBuilder* aBuilder,
nsSubDocumentFrame* aFrame,
const nsRect& aDirtyRect,
const nsDisplayListSet& aLists)
{
// We're the subdoc for <browser remote="true"> and it has
// painted content. Display its shadow layer tree.
nsDisplayList shadowTree;
if (aBuilder->IsForEventDelivery()) {
nsRect bounds = aFrame->EnsureInnerView()->GetBounds();
ViewTransform offset =
ViewTransform(GetRootFrameOffset(aFrame, aBuilder), 1, 1);
BuildListForLayer(GetRootLayer(), mFrameLoader, offset,
aBuilder, shadowTree, aFrame);
} else {
shadowTree.AppendToTop(
new (aBuilder) nsDisplayRemote(aBuilder, aFrame, this));
}
// Clip the shadow layers to subdoc bounds
nsPoint offset = aFrame->GetOffsetToCrossDoc(aBuilder->ReferenceFrame());
nsRect bounds = aFrame->EnsureInnerView()->GetBounds() + offset;
return aLists.Content()->AppendNewToTop(
new (aBuilder) nsDisplayClip(aBuilder, aFrame, &shadowTree,
bounds));
}
} // namespace layout
} // namespace mozilla
already_AddRefed<Layer>
nsDisplayRemote::BuildLayer(nsDisplayListBuilder* aBuilder,
LayerManager* aManager,
const ContainerParameters& aContainerParameters)
{
PRInt32 appUnitsPerDevPixel = mFrame->PresContext()->AppUnitsPerDevPixel();
nsIntRect visibleRect = GetVisibleRect().ToNearestPixels(appUnitsPerDevPixel);
nsRefPtr<Layer> layer = mRemoteFrame->BuildLayer(aBuilder, mFrame, aManager, visibleRect);
return layer.forget();
}
void
nsDisplayRemoteShadow::HitTest(nsDisplayListBuilder* aBuilder, const nsRect& aRect,
HitTestState* aState, nsTArray<nsIFrame*> *aOutFrames)
{
// If we are here, then rects have intersected.
//
// XXX I think iframes and divs can be rounded like anything else but we don't
// cover that case here.
//
if (aState->mShadows) {
aState->mShadows->AppendElement(mId);
}
}