зеркало из https://github.com/mozilla/gecko-dev.git
549 строки
21 KiB
C++
549 строки
21 KiB
C++
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
|
|
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
|
|
/* This Source Code Form is subject to the terms of the Mozilla Public
|
|
* License, v. 2.0. If a copy of the MPL was not distributed with this
|
|
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
|
|
|
|
/* rendering object for the HTML <canvas> element */
|
|
|
|
#include "nsHTMLCanvasFrame.h"
|
|
|
|
#include "nsGkAtoms.h"
|
|
#include "mozilla/Assertions.h"
|
|
#include "mozilla/PresShell.h"
|
|
#include "mozilla/dom/HTMLCanvasElement.h"
|
|
#include "mozilla/layers/WebRenderBridgeChild.h"
|
|
#include "mozilla/layers/WebRenderCanvasRenderer.h"
|
|
#include "mozilla/layers/RenderRootStateManager.h"
|
|
#include "nsDisplayList.h"
|
|
#include "nsLayoutUtils.h"
|
|
#include "nsStyleUtil.h"
|
|
#include "ImageLayers.h"
|
|
#include "Layers.h"
|
|
#include "ActiveLayerTracker.h"
|
|
|
|
#include <algorithm>
|
|
|
|
using namespace mozilla;
|
|
using namespace mozilla::dom;
|
|
using namespace mozilla::layers;
|
|
using namespace mozilla::gfx;
|
|
|
|
/* Helper for our nsIFrame::GetIntrinsicSize() impl. Takes the result of
|
|
* "GetCanvasSize()" as a parameter, which may help avoid redundant
|
|
* indirect calls to GetCanvasSize().
|
|
*
|
|
* @param aCanvasSizeInPx The canvas's size in CSS pixels, as returned
|
|
* by GetCanvasSize().
|
|
* @return The canvas's intrinsic size, as an IntrinsicSize object.
|
|
*/
|
|
static IntrinsicSize IntrinsicSizeFromCanvasSize(
|
|
const nsIntSize& aCanvasSizeInPx) {
|
|
return IntrinsicSize(
|
|
nsPresContext::CSSPixelsToAppUnits(aCanvasSizeInPx.width),
|
|
nsPresContext::CSSPixelsToAppUnits(aCanvasSizeInPx.height));
|
|
}
|
|
|
|
/* Helper for our nsIFrame::GetIntrinsicRatio() impl. Takes the result of
|
|
* "GetCanvasSize()" as a parameter, which may help avoid redundant
|
|
* indirect calls to GetCanvasSize().
|
|
*
|
|
* @param aCanvasSizeInPx The canvas's size in CSS pixels, as returned
|
|
* by GetCanvasSize().
|
|
* @return The canvas's intrinsic ratio, as a nsSize.
|
|
*/
|
|
static AspectRatio IntrinsicRatioFromCanvasSize(
|
|
const nsIntSize& aCanvasSizeInPx) {
|
|
return AspectRatio::FromSize(aCanvasSizeInPx.width, aCanvasSizeInPx.height);
|
|
}
|
|
|
|
class nsDisplayCanvas final : public nsPaintedDisplayItem {
|
|
public:
|
|
nsDisplayCanvas(nsDisplayListBuilder* aBuilder, nsIFrame* aFrame)
|
|
: nsPaintedDisplayItem(aBuilder, aFrame) {
|
|
MOZ_COUNT_CTOR(nsDisplayCanvas);
|
|
}
|
|
#ifdef NS_BUILD_REFCNT_LOGGING
|
|
virtual ~nsDisplayCanvas() { MOZ_COUNT_DTOR(nsDisplayCanvas); }
|
|
#endif
|
|
|
|
NS_DISPLAY_DECL_NAME("nsDisplayCanvas", TYPE_CANVAS)
|
|
|
|
virtual nsRegion GetOpaqueRegion(nsDisplayListBuilder* aBuilder,
|
|
bool* aSnap) const override {
|
|
*aSnap = false;
|
|
nsHTMLCanvasFrame* f = static_cast<nsHTMLCanvasFrame*>(Frame());
|
|
HTMLCanvasElement* canvas = HTMLCanvasElement::FromNode(f->GetContent());
|
|
nsRegion result;
|
|
if (canvas->GetIsOpaque()) {
|
|
// OK, the entire region painted by the canvas is opaque. But what is
|
|
// that region? It's the canvas's "dest rect" (controlled by the
|
|
// object-fit/object-position CSS properties), clipped to the container's
|
|
// content box (which is what GetBounds() returns). So, we grab those
|
|
// rects and intersect them.
|
|
nsRect constraintRect = GetBounds(aBuilder, aSnap);
|
|
|
|
// Need intrinsic size & ratio, for ComputeObjectDestRect:
|
|
nsIntSize canvasSize = f->GetCanvasSize();
|
|
IntrinsicSize intrinsicSize = IntrinsicSizeFromCanvasSize(canvasSize);
|
|
AspectRatio intrinsicRatio = IntrinsicRatioFromCanvasSize(canvasSize);
|
|
|
|
const nsRect destRect = nsLayoutUtils::ComputeObjectDestRect(
|
|
constraintRect, intrinsicSize, intrinsicRatio, f->StylePosition());
|
|
return nsRegion(destRect.Intersect(constraintRect));
|
|
}
|
|
return result;
|
|
}
|
|
|
|
virtual nsRect GetBounds(nsDisplayListBuilder* aBuilder,
|
|
bool* aSnap) const override {
|
|
*aSnap = true;
|
|
nsHTMLCanvasFrame* f = static_cast<nsHTMLCanvasFrame*>(Frame());
|
|
return f->GetInnerArea() + ToReferenceFrame();
|
|
}
|
|
|
|
virtual already_AddRefed<Layer> BuildLayer(
|
|
nsDisplayListBuilder* aBuilder, LayerManager* aManager,
|
|
const ContainerLayerParameters& aContainerParameters) override {
|
|
return static_cast<nsHTMLCanvasFrame*>(mFrame)->BuildLayer(
|
|
aBuilder, aManager, this, aContainerParameters);
|
|
}
|
|
|
|
virtual bool CreateWebRenderCommands(
|
|
mozilla::wr::DisplayListBuilder& aBuilder,
|
|
wr::IpcResourceUpdateQueue& aResources, const StackingContextHelper& aSc,
|
|
mozilla::layers::RenderRootStateManager* aManager,
|
|
nsDisplayListBuilder* aDisplayListBuilder) override {
|
|
HTMLCanvasElement* element =
|
|
static_cast<HTMLCanvasElement*>(mFrame->GetContent());
|
|
element->HandlePrintCallback(mFrame->PresContext()->Type());
|
|
|
|
switch (element->GetCurrentContextType()) {
|
|
case CanvasContextType::Canvas2D:
|
|
case CanvasContextType::WebGL1:
|
|
case CanvasContextType::WebGL2: {
|
|
bool isRecycled;
|
|
RefPtr<WebRenderCanvasData> canvasData =
|
|
aManager->CommandBuilder()
|
|
.CreateOrRecycleWebRenderUserData<WebRenderCanvasData>(
|
|
this, aBuilder.GetRenderRoot(), &isRecycled);
|
|
nsHTMLCanvasFrame* canvasFrame =
|
|
static_cast<nsHTMLCanvasFrame*>(mFrame);
|
|
if (!canvasFrame->UpdateWebRenderCanvasData(aDisplayListBuilder,
|
|
canvasData)) {
|
|
return true;
|
|
}
|
|
WebRenderCanvasRendererAsync* data =
|
|
static_cast<WebRenderCanvasRendererAsync*>(
|
|
canvasData->GetCanvasRenderer());
|
|
MOZ_ASSERT(data);
|
|
data->UpdateCompositableClient(aBuilder.GetRenderRoot());
|
|
|
|
// Push IFrame for async image pipeline.
|
|
// XXX Remove this once partial display list update is supported.
|
|
|
|
nsIntSize canvasSizeInPx = data->GetSize();
|
|
IntrinsicSize intrinsicSize =
|
|
IntrinsicSizeFromCanvasSize(canvasSizeInPx);
|
|
AspectRatio intrinsicRatio =
|
|
IntrinsicRatioFromCanvasSize(canvasSizeInPx);
|
|
|
|
nsRect area =
|
|
mFrame->GetContentRectRelativeToSelf() + ToReferenceFrame();
|
|
nsRect dest = nsLayoutUtils::ComputeObjectDestRect(
|
|
area, intrinsicSize, intrinsicRatio, mFrame->StylePosition());
|
|
|
|
LayoutDeviceRect bounds = LayoutDeviceRect::FromAppUnits(
|
|
dest, mFrame->PresContext()->AppUnitsPerDevPixel());
|
|
|
|
// We don't push a stacking context for this async image pipeline here.
|
|
// Instead, we do it inside the iframe that hosts the image. As a
|
|
// result, a bunch of the calculations normally done as part of that
|
|
// stacking context need to be done manually and pushed over to the
|
|
// parent side, where it will be done when we build the display list for
|
|
// the iframe. That happens in WebRenderCompositableHolder.
|
|
|
|
wr::LayoutRect r = wr::ToLayoutRect(bounds);
|
|
aBuilder.PushIFrame(r, !BackfaceIsHidden(), data->GetPipelineId().ref(),
|
|
/*ignoreMissingPipelines*/ false);
|
|
|
|
gfx::Matrix4x4 scTransform;
|
|
gfxRect destGFXRect = mFrame->PresContext()->AppUnitsToGfxUnits(dest);
|
|
scTransform.PreScale(destGFXRect.Width() / canvasSizeInPx.width,
|
|
destGFXRect.Height() / canvasSizeInPx.height,
|
|
1.0f);
|
|
|
|
MaybeIntSize scaleToSize;
|
|
LayoutDeviceRect scBounds(LayoutDevicePoint(0, 0), bounds.Size());
|
|
wr::ImageRendering filter = wr::ToImageRendering(
|
|
nsLayoutUtils::GetSamplingFilterForFrame(mFrame));
|
|
wr::MixBlendMode mixBlendMode = wr::MixBlendMode::Normal;
|
|
aManager->WrBridge()->AddWebRenderParentCommand(
|
|
OpUpdateAsyncImagePipeline(data->GetPipelineId().value(), scBounds,
|
|
scTransform, scaleToSize, filter,
|
|
mixBlendMode),
|
|
aBuilder.GetRenderRoot());
|
|
break;
|
|
}
|
|
case CanvasContextType::ImageBitmap: {
|
|
nsHTMLCanvasFrame* canvasFrame =
|
|
static_cast<nsHTMLCanvasFrame*>(mFrame);
|
|
nsIntSize canvasSizeInPx = canvasFrame->GetCanvasSize();
|
|
if (canvasSizeInPx.width <= 0 || canvasSizeInPx.height <= 0) {
|
|
return true;
|
|
}
|
|
bool isRecycled;
|
|
RefPtr<WebRenderCanvasData> canvasData =
|
|
aManager->CommandBuilder()
|
|
.CreateOrRecycleWebRenderUserData<WebRenderCanvasData>(
|
|
this, aBuilder.GetRenderRoot(), &isRecycled);
|
|
if (!canvasFrame->UpdateWebRenderCanvasData(aDisplayListBuilder,
|
|
canvasData)) {
|
|
canvasData->ClearImageContainer();
|
|
return true;
|
|
}
|
|
|
|
IntrinsicSize intrinsicSize =
|
|
IntrinsicSizeFromCanvasSize(canvasSizeInPx);
|
|
AspectRatio intrinsicRatio =
|
|
IntrinsicRatioFromCanvasSize(canvasSizeInPx);
|
|
|
|
nsRect area =
|
|
mFrame->GetContentRectRelativeToSelf() + ToReferenceFrame();
|
|
nsRect dest = nsLayoutUtils::ComputeObjectDestRect(
|
|
area, intrinsicSize, intrinsicRatio, mFrame->StylePosition());
|
|
|
|
LayoutDeviceRect bounds = LayoutDeviceRect::FromAppUnits(
|
|
dest, mFrame->PresContext()->AppUnitsPerDevPixel());
|
|
|
|
aManager->CommandBuilder().PushImage(
|
|
this, canvasData->GetImageContainer(), aBuilder, aResources, aSc,
|
|
bounds, bounds);
|
|
break;
|
|
}
|
|
case CanvasContextType::NoContext:
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
virtual LayerState GetLayerState(
|
|
nsDisplayListBuilder* aBuilder, LayerManager* aManager,
|
|
const ContainerLayerParameters& aParameters) override {
|
|
if (HTMLCanvasElement::FromNode(mFrame->GetContent())
|
|
->ShouldForceInactiveLayer(aManager))
|
|
return LayerState::LAYER_INACTIVE;
|
|
|
|
// If compositing is cheap, just do that
|
|
if (aManager->IsCompositingCheap() ||
|
|
ActiveLayerTracker::IsContentActive(mFrame))
|
|
return mozilla::LayerState::LAYER_ACTIVE;
|
|
|
|
return LayerState::LAYER_INACTIVE;
|
|
}
|
|
|
|
// FirstContentfulPaint is supposed to ignore "white" canvases. We use
|
|
// MaybeModified (if GetContext() was called on the canvas) as a standin for
|
|
// "white"
|
|
virtual bool IsContentful() const override {
|
|
nsHTMLCanvasFrame* f = static_cast<nsHTMLCanvasFrame*>(Frame());
|
|
HTMLCanvasElement* canvas = HTMLCanvasElement::FromNode(f->GetContent());
|
|
return canvas->MaybeModified();
|
|
}
|
|
};
|
|
|
|
nsIFrame* NS_NewHTMLCanvasFrame(PresShell* aPresShell, ComputedStyle* aStyle) {
|
|
return new (aPresShell)
|
|
nsHTMLCanvasFrame(aStyle, aPresShell->GetPresContext());
|
|
}
|
|
|
|
NS_QUERYFRAME_HEAD(nsHTMLCanvasFrame)
|
|
NS_QUERYFRAME_ENTRY(nsHTMLCanvasFrame)
|
|
NS_QUERYFRAME_TAIL_INHERITING(nsContainerFrame)
|
|
|
|
NS_IMPL_FRAMEARENA_HELPERS(nsHTMLCanvasFrame)
|
|
|
|
void nsHTMLCanvasFrame::Init(nsIContent* aContent, nsContainerFrame* aParent,
|
|
nsIFrame* aPrevInFlow) {
|
|
nsContainerFrame::Init(aContent, aParent, aPrevInFlow);
|
|
|
|
// We can fill in the canvas before the canvas frame is created, in
|
|
// which case we never get around to marking the content as active. Therefore,
|
|
// we mark it active here when we create the frame.
|
|
ActiveLayerTracker::NotifyContentChange(this);
|
|
}
|
|
|
|
nsHTMLCanvasFrame::~nsHTMLCanvasFrame() {}
|
|
|
|
nsIntSize nsHTMLCanvasFrame::GetCanvasSize() {
|
|
nsIntSize size(0, 0);
|
|
HTMLCanvasElement* canvas = HTMLCanvasElement::FromNodeOrNull(GetContent());
|
|
if (canvas) {
|
|
size = canvas->GetSize();
|
|
MOZ_ASSERT(size.width >= 0 && size.height >= 0,
|
|
"we should've required <canvas> width/height attrs to be "
|
|
"unsigned (non-negative) values");
|
|
} else {
|
|
MOZ_ASSERT_UNREACHABLE("couldn't get canvas size");
|
|
}
|
|
|
|
return size;
|
|
}
|
|
|
|
/* virtual */
|
|
nscoord nsHTMLCanvasFrame::GetMinISize(gfxContext* aRenderingContext) {
|
|
// XXX The caller doesn't account for constraints of the height,
|
|
// min-height, and max-height properties.
|
|
bool vertical = GetWritingMode().IsVertical();
|
|
nscoord result;
|
|
if (StyleDisplay()->IsContainSize()) {
|
|
result = 0;
|
|
} else {
|
|
result = nsPresContext::CSSPixelsToAppUnits(
|
|
vertical ? GetCanvasSize().height : GetCanvasSize().width);
|
|
}
|
|
DISPLAY_MIN_INLINE_SIZE(this, result);
|
|
return result;
|
|
}
|
|
|
|
/* virtual */
|
|
nscoord nsHTMLCanvasFrame::GetPrefISize(gfxContext* aRenderingContext) {
|
|
// XXX The caller doesn't account for constraints of the height,
|
|
// min-height, and max-height properties.
|
|
bool vertical = GetWritingMode().IsVertical();
|
|
nscoord result;
|
|
if (StyleDisplay()->IsContainSize()) {
|
|
result = 0;
|
|
} else {
|
|
result = nsPresContext::CSSPixelsToAppUnits(
|
|
vertical ? GetCanvasSize().height : GetCanvasSize().width);
|
|
}
|
|
DISPLAY_PREF_INLINE_SIZE(this, result);
|
|
return result;
|
|
}
|
|
|
|
/* virtual */
|
|
IntrinsicSize nsHTMLCanvasFrame::GetIntrinsicSize() {
|
|
if (StyleDisplay()->IsContainSize()) {
|
|
return IntrinsicSize(0, 0);
|
|
}
|
|
return IntrinsicSizeFromCanvasSize(GetCanvasSize());
|
|
}
|
|
|
|
/* virtual */
|
|
AspectRatio nsHTMLCanvasFrame::GetIntrinsicRatio() {
|
|
if (StyleDisplay()->IsContainSize()) {
|
|
return AspectRatio();
|
|
}
|
|
return IntrinsicRatioFromCanvasSize(GetCanvasSize());
|
|
}
|
|
|
|
/* virtual */
|
|
LogicalSize nsHTMLCanvasFrame::ComputeSize(
|
|
gfxContext* aRenderingContext, WritingMode aWM, const LogicalSize& aCBSize,
|
|
nscoord aAvailableISize, const LogicalSize& aMargin,
|
|
const LogicalSize& aBorder, const LogicalSize& aPadding,
|
|
ComputeSizeFlags aFlags) {
|
|
IntrinsicSize intrinsicSize;
|
|
AspectRatio intrinsicRatio;
|
|
if (StyleDisplay()->IsContainSize()) {
|
|
intrinsicSize = IntrinsicSize(0, 0);
|
|
// intrinsicRatio is already implicitly zero via default ctor.
|
|
} else {
|
|
nsIntSize canvasSizeInPx = GetCanvasSize();
|
|
intrinsicSize = IntrinsicSizeFromCanvasSize(canvasSizeInPx);
|
|
intrinsicRatio = IntrinsicRatioFromCanvasSize(canvasSizeInPx);
|
|
}
|
|
|
|
return ComputeSizeWithIntrinsicDimensions(
|
|
aRenderingContext, aWM, intrinsicSize, intrinsicRatio, aCBSize, aMargin,
|
|
aBorder, aPadding, aFlags);
|
|
}
|
|
|
|
void nsHTMLCanvasFrame::Reflow(nsPresContext* aPresContext,
|
|
ReflowOutput& aMetrics,
|
|
const ReflowInput& aReflowInput,
|
|
nsReflowStatus& aStatus) {
|
|
MarkInReflow();
|
|
DO_GLOBAL_REFLOW_COUNT("nsHTMLCanvasFrame");
|
|
DISPLAY_REFLOW(aPresContext, this, aReflowInput, aMetrics, aStatus);
|
|
MOZ_ASSERT(aStatus.IsEmpty(), "Caller should pass a fresh reflow status!");
|
|
NS_FRAME_TRACE(
|
|
NS_FRAME_TRACE_CALLS,
|
|
("enter nsHTMLCanvasFrame::Reflow: availSize=%d,%d",
|
|
aReflowInput.AvailableWidth(), aReflowInput.AvailableHeight()));
|
|
|
|
MOZ_ASSERT(mState & NS_FRAME_IN_REFLOW, "frame is not in reflow");
|
|
|
|
WritingMode wm = aReflowInput.GetWritingMode();
|
|
LogicalSize finalSize = aReflowInput.ComputedSize();
|
|
|
|
// stash this away so we can compute our inner area later
|
|
mBorderPadding = aReflowInput.ComputedLogicalBorderPadding();
|
|
|
|
finalSize.ISize(wm) += mBorderPadding.IStartEnd(wm);
|
|
finalSize.BSize(wm) += mBorderPadding.BStartEnd(wm);
|
|
|
|
if (GetPrevInFlow()) {
|
|
nscoord y = GetContinuationOffset(&finalSize.ISize(wm));
|
|
finalSize.BSize(wm) -= y + mBorderPadding.BStart(wm);
|
|
finalSize.BSize(wm) = std::max(0, finalSize.BSize(wm));
|
|
}
|
|
|
|
aMetrics.SetSize(wm, finalSize);
|
|
aMetrics.SetOverflowAreasToDesiredBounds();
|
|
FinishAndStoreOverflow(&aMetrics);
|
|
|
|
// Reflow the single anon block child.
|
|
nsReflowStatus childStatus;
|
|
nsIFrame* childFrame = mFrames.FirstChild();
|
|
WritingMode childWM = childFrame->GetWritingMode();
|
|
LogicalSize availSize = aReflowInput.ComputedSize(childWM);
|
|
availSize.BSize(childWM) = NS_UNCONSTRAINEDSIZE;
|
|
NS_ASSERTION(!childFrame->GetNextSibling(), "HTML canvas should have 1 kid");
|
|
ReflowOutput childDesiredSize(aReflowInput.GetWritingMode());
|
|
ReflowInput childReflowInput(aPresContext, aReflowInput, childFrame,
|
|
availSize);
|
|
ReflowChild(childFrame, aPresContext, childDesiredSize, childReflowInput, 0,
|
|
0, ReflowChildFlags::Default, childStatus, nullptr);
|
|
FinishReflowChild(childFrame, aPresContext, childDesiredSize,
|
|
&childReflowInput, 0, 0, ReflowChildFlags::Default);
|
|
|
|
NS_FRAME_TRACE(NS_FRAME_TRACE_CALLS,
|
|
("exit nsHTMLCanvasFrame::Reflow: size=%d,%d",
|
|
aMetrics.ISize(wm), aMetrics.BSize(wm)));
|
|
NS_FRAME_SET_TRUNCATION(aStatus, aReflowInput, aMetrics);
|
|
}
|
|
|
|
// FIXME taken from nsImageFrame, but then had splittable frame stuff
|
|
// removed. That needs to be fixed.
|
|
// XXXdholbert As in nsImageFrame, this function's clients should probably
|
|
// just be calling GetContentRectRelativeToSelf().
|
|
nsRect nsHTMLCanvasFrame::GetInnerArea() const {
|
|
nsMargin bp = mBorderPadding.GetPhysicalMargin(GetWritingMode());
|
|
nsRect r;
|
|
r.x = bp.left;
|
|
r.y = bp.top;
|
|
r.width = mRect.width - bp.left - bp.right;
|
|
r.height = mRect.height - bp.top - bp.bottom;
|
|
return r;
|
|
}
|
|
|
|
already_AddRefed<Layer> nsHTMLCanvasFrame::BuildLayer(
|
|
nsDisplayListBuilder* aBuilder, LayerManager* aManager,
|
|
nsDisplayItem* aItem,
|
|
const ContainerLayerParameters& aContainerParameters) {
|
|
nsRect area = GetContentRectRelativeToSelf() + aItem->ToReferenceFrame();
|
|
HTMLCanvasElement* element = static_cast<HTMLCanvasElement*>(GetContent());
|
|
nsIntSize canvasSizeInPx = GetCanvasSize();
|
|
|
|
nsPresContext* presContext = PresContext();
|
|
element->HandlePrintCallback(presContext->Type());
|
|
|
|
if (canvasSizeInPx.width <= 0 || canvasSizeInPx.height <= 0 || area.IsEmpty())
|
|
return nullptr;
|
|
|
|
Layer* oldLayer =
|
|
aManager->GetLayerBuilder()->GetLeafLayerFor(aBuilder, aItem);
|
|
RefPtr<Layer> layer = element->GetCanvasLayer(aBuilder, oldLayer, aManager);
|
|
if (!layer) return nullptr;
|
|
|
|
IntrinsicSize intrinsicSize = IntrinsicSizeFromCanvasSize(canvasSizeInPx);
|
|
AspectRatio intrinsicRatio = IntrinsicRatioFromCanvasSize(canvasSizeInPx);
|
|
|
|
nsRect dest = nsLayoutUtils::ComputeObjectDestRect(
|
|
area, intrinsicSize, intrinsicRatio, StylePosition());
|
|
|
|
gfxRect destGFXRect = presContext->AppUnitsToGfxUnits(dest);
|
|
|
|
// Transform the canvas into the right place
|
|
gfxPoint p = destGFXRect.TopLeft() + aContainerParameters.mOffset;
|
|
Matrix transform = Matrix::Translation(p.x, p.y);
|
|
transform.PreScale(destGFXRect.Width() / canvasSizeInPx.width,
|
|
destGFXRect.Height() / canvasSizeInPx.height);
|
|
layer->SetBaseTransform(gfx::Matrix4x4::From2D(transform));
|
|
if (layer->GetType() == layers::Layer::TYPE_CANVAS) {
|
|
RefPtr<CanvasLayer> canvasLayer = static_cast<CanvasLayer*>(layer.get());
|
|
canvasLayer->SetSamplingFilter(
|
|
nsLayoutUtils::GetSamplingFilterForFrame(this));
|
|
nsIntRect bounds;
|
|
bounds.SetRect(0, 0, canvasSizeInPx.width, canvasSizeInPx.height);
|
|
canvasLayer->SetBounds(bounds);
|
|
} else if (layer->GetType() == layers::Layer::TYPE_IMAGE) {
|
|
RefPtr<ImageLayer> imageLayer = static_cast<ImageLayer*>(layer.get());
|
|
imageLayer->SetSamplingFilter(
|
|
nsLayoutUtils::GetSamplingFilterForFrame(this));
|
|
}
|
|
|
|
return layer.forget();
|
|
}
|
|
|
|
bool nsHTMLCanvasFrame::UpdateWebRenderCanvasData(
|
|
nsDisplayListBuilder* aBuilder, WebRenderCanvasData* aCanvasData) {
|
|
HTMLCanvasElement* element = static_cast<HTMLCanvasElement*>(GetContent());
|
|
return element->UpdateWebRenderCanvasData(aBuilder, aCanvasData);
|
|
}
|
|
|
|
void nsHTMLCanvasFrame::BuildDisplayList(nsDisplayListBuilder* aBuilder,
|
|
const nsDisplayListSet& aLists) {
|
|
if (!IsVisibleForPainting()) return;
|
|
|
|
DisplayBorderBackgroundOutline(aBuilder, aLists);
|
|
|
|
uint32_t clipFlags =
|
|
nsStyleUtil::ObjectPropsMightCauseOverflow(StylePosition())
|
|
? 0
|
|
: DisplayListClipState::ASSUME_DRAWING_RESTRICTED_TO_CONTENT_RECT;
|
|
|
|
DisplayListClipState::AutoClipContainingBlockDescendantsToContentBox clip(
|
|
aBuilder, this, clipFlags);
|
|
|
|
aLists.Content()->AppendNewToTop<nsDisplayCanvas>(aBuilder, this);
|
|
|
|
DisplaySelectionOverlay(aBuilder, aLists.Content(),
|
|
nsISelectionDisplay::DISPLAY_IMAGES);
|
|
}
|
|
|
|
// get the offset into the content area of the image where aImg starts if it is
|
|
// a continuation. from nsImageFrame
|
|
nscoord nsHTMLCanvasFrame::GetContinuationOffset(nscoord* aWidth) const {
|
|
nscoord offset = 0;
|
|
if (aWidth) {
|
|
*aWidth = 0;
|
|
}
|
|
|
|
if (GetPrevInFlow()) {
|
|
for (nsIFrame* prevInFlow = GetPrevInFlow(); prevInFlow;
|
|
prevInFlow = prevInFlow->GetPrevInFlow()) {
|
|
nsRect rect = prevInFlow->GetRect();
|
|
if (aWidth) {
|
|
*aWidth = rect.width;
|
|
}
|
|
offset += rect.height;
|
|
}
|
|
offset -= mBorderPadding.GetPhysicalMargin(GetWritingMode()).top;
|
|
offset = std::max(0, offset);
|
|
}
|
|
return offset;
|
|
}
|
|
|
|
void nsHTMLCanvasFrame::AppendDirectlyOwnedAnonBoxes(
|
|
nsTArray<OwnedAnonBox>& aResult) {
|
|
MOZ_ASSERT(mFrames.FirstChild(), "Must have our canvas content anon box");
|
|
MOZ_ASSERT(!mFrames.FirstChild()->GetNextSibling(),
|
|
"Must only have our canvas content anon box");
|
|
aResult.AppendElement(OwnedAnonBox(mFrames.FirstChild()));
|
|
}
|
|
|
|
#ifdef ACCESSIBILITY
|
|
a11y::AccType nsHTMLCanvasFrame::AccessibleType() {
|
|
return a11y::eHTMLCanvasType;
|
|
}
|
|
#endif
|
|
|
|
#ifdef DEBUG_FRAME_DUMP
|
|
nsresult nsHTMLCanvasFrame::GetFrameName(nsAString& aResult) const {
|
|
return MakeFrameName(NS_LITERAL_STRING("HTMLCanvas"), aResult);
|
|
}
|
|
#endif
|