Source code

Revision control

Copy as Markdown

Other Tools

/* -*- 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 */
// Main header first:
#include "SVGOuterSVGFrame.h"
// Keep others in (case-insensitive) order:
#include "gfxContext.h"
#include "nsDisplayList.h"
#include "nsIInterfaceRequestorUtils.h"
#include "nsLayoutUtils.h"
#include "nsObjectLoadingContent.h"
#include "nsSubDocumentFrame.h"
#include "mozilla/PresShell.h"
#include "mozilla/SVGUtils.h"
#include "mozilla/dom/BrowserChild.h"
#include "mozilla/dom/Document.h"
#include "mozilla/dom/Element.h"
#include "mozilla/dom/SVGSVGElement.h"
using namespace mozilla::dom;
using namespace mozilla::gfx;
using namespace mozilla::image;
// Implementation
nsContainerFrame* NS_NewSVGOuterSVGFrame(mozilla::PresShell* aPresShell,
mozilla::ComputedStyle* aStyle) {
return new (aPresShell)
mozilla::SVGOuterSVGFrame(aStyle, aPresShell->GetPresContext());
namespace mozilla {
SVGOuterSVGFrame::SVGOuterSVGFrame(ComputedStyle* aStyle,
nsPresContext* aPresContext)
: SVGDisplayContainerFrame(aStyle, aPresContext, kClassID) {
// Outer-<svg> has CSS layout, so remove this bit:
// The CSS Containment spec says that size-contained replaced elements must be
// treated as having an intrinsic width and height of 0. That's applicable to
// outer SVG frames, unless they're the outermost element (in which case
// they're not really "replaced", and there's no outer context to contain sizes
// from leaking into). Hence, we check for a parent element before we bother
// testing for 'contain:size'.
static inline ContainSizeAxes ContainSizeAxesIfApplicable(
const SVGOuterSVGFrame* aFrame) {
if (!aFrame->GetContent()->GetParent()) {
return ContainSizeAxes(false, false);
return aFrame->GetContainSizeAxes();
// This should match ImageDocument::GetZoomLevel.
float SVGOuterSVGFrame::ComputeFullZoom() const {
if (BrowsingContext* bc = PresContext()->Document()->GetBrowsingContext()) {
return bc->FullZoom();
return 1.0f;
void SVGOuterSVGFrame::Init(nsIContent* aContent, nsContainerFrame* aParent,
nsIFrame* aPrevInFlow) {
"Content is not an SVG 'svg' element!");
// Check for conditional processing attributes here rather than in
// nsCSSFrameConstructor::FindSVGData because we want to avoid
// simply giving failing outer <svg> elements an SVGContainerFrame.
// We don't create other SVG frames if PassesConditionalProcessingTests
// returns false, but since we do create SVGOuterSVGFrame frames we
// prevent them from painting by [ab]use NS_FRAME_IS_NONDISPLAY. The
// frame will be recreated via an nsChangeHint_ReconstructFrame restyle if
// the value returned by PassesConditionalProcessingTests changes.
auto* svg = static_cast<SVGSVGElement*>(aContent);
if (!svg->PassesConditionalProcessingTests()) {
SVGDisplayContainerFrame::Init(aContent, aParent, aPrevInFlow);
Document* doc = mContent->GetUncomposedDoc();
mIsRootContent = doc && doc->GetRootElement() == mContent;
if (mIsRootContent) {
if (nsCOMPtr<nsIDocShell> docShell = PresContext()->GetDocShell()) {
RefPtr<BrowsingContext> bc = docShell->GetBrowsingContext();
if (const Maybe<nsString>& type = bc->GetEmbedderElementType()) {
mIsInObjectOrEmbed =
nsGkAtoms::object->Equals(*type) || nsGkAtoms::embed->Equals(*type);
mIsInIframe = nsGkAtoms::iframe->Equals(*type);
if (!mIsInIframe) {
mFullZoom = ComputeFullZoom();
// nsQueryFrame methods
// nsIFrame methods
// reflowing
/* virtual */
nscoord SVGOuterSVGFrame::GetMinISize(gfxContext* aRenderingContext) {
return GetIntrinsicSize().ISize(GetWritingMode()).valueOr(0);
/* virtual */
nscoord SVGOuterSVGFrame::GetPrefISize(gfxContext* aRenderingContext) {
nscoord result;
SVGSVGElement* svg = static_cast<SVGSVGElement*>(GetContent());
WritingMode wm = GetWritingMode();
const SVGAnimatedLength& isize =
wm.IsVertical() ? svg->mLengthAttributes[SVGSVGElement::ATTR_HEIGHT]
: svg->mLengthAttributes[SVGSVGElement::ATTR_WIDTH];
if (Maybe<nscoord> containISize =
ContainSizeAxesIfApplicable(this).ContainIntrinsicISize(*this)) {
result = *containISize;
} else if (isize.IsPercentage()) {
// If we are here, our inline size attribute is a percentage either
// explicitly (via an attribute value) or implicitly (by being unset, which
// is treated as 100%). The following if-condition, deciding to return
// either the fallback intrinsic size or zero, is made to match blink and
// webkit's behavior for webcompat.
if (isize.IsExplicitlySet() || StylePosition()->ISize(wm).HasPercent() ||
!GetAspectRatio()) {
result = wm.IsVertical() ? kFallbackIntrinsicSize.height
: kFallbackIntrinsicSize.width;
} else {
result = nscoord(0);
} else {
result =
if (result < 0) {
result = nscoord(0);
return result;
/* virtual */
IntrinsicSize SVGOuterSVGFrame::GetIntrinsicSize() {
// XXXjwatt Note that here we want to return the CSS width/height if they're
// specified and we're embedded inside an nsIObjectLoadingContent.
const auto containAxes = ContainSizeAxesIfApplicable(this);
if (containAxes.IsBoth()) {
// Intrinsic size of 'contain:size' replaced elements is determined by
// contain-intrinsic-size, defaulting to 0x0.
return FinishIntrinsicSize(containAxes, IntrinsicSize(0, 0));
SVGSVGElement* content = static_cast<SVGSVGElement*>(GetContent());
const SVGAnimatedLength& width =
const SVGAnimatedLength& height =
IntrinsicSize intrinsicSize;
if (!width.IsPercentage()) {
nscoord val =
intrinsicSize.width.emplace(std::max(val, 0));
if (!height.IsPercentage()) {
nscoord val = nsPresContext::CSSPixelsToAppUnits(
intrinsicSize.height.emplace(std::max(val, 0));
return FinishIntrinsicSize(containAxes, intrinsicSize);
/* virtual */
AspectRatio SVGOuterSVGFrame::GetIntrinsicRatio() const {
if (ContainSizeAxesIfApplicable(this).IsAny()) {
return AspectRatio();
// We only have an intrinsic size/ratio if our width and height attributes
// are both specified and set to non-percentage values, or we have a viewBox
auto* content = static_cast<SVGSVGElement*>(GetContent());
const SVGAnimatedLength& width =
const SVGAnimatedLength& height =
if (!width.IsPercentage() && !height.IsPercentage()) {
// Use width/height ratio only if
// 1. it's not a degenerate ratio, and
// 2. width and height are non-negative numbers.
// Otherwise, we use the viewbox rect.
// Note width/height may have different units and therefore be
// affected by zoom in different ways.
const float w = width.GetAnimValueWithZoom(content);
const float h = height.GetAnimValueWithZoom(content);
if (w > 0.0f && h > 0.0f) {
return AspectRatio::FromSize(w, h);
const auto& viewBox = content->GetViewBoxInternal();
if (viewBox.HasRect()) {
float zoom = Style()->EffectiveZoom().ToFloat();
const auto& anim = viewBox.GetAnimValue() * zoom;
return AspectRatio::FromSize(anim.width, anim.height);
return SVGDisplayContainerFrame::GetIntrinsicRatio();
/* virtual */
nsIFrame::SizeComputationResult SVGOuterSVGFrame::ComputeSize(
gfxContext* aRenderingContext, WritingMode aWritingMode,
const LogicalSize& aCBSize, nscoord aAvailableISize,
const LogicalSize& aMargin, const LogicalSize& aBorderPadding,
const StyleSizeOverrides& aSizeOverrides, ComputeSizeFlags aFlags) {
if (IsRootOfImage() || mIsInObjectOrEmbed) {
// The embedding element has sized itself using the CSS replaced element
// sizing rules, using our intrinsic dimensions as necessary. The SVG spec
// says that the width and height of embedded SVG is overridden by the
// width and height of the embedding element, so we just need to size to
// the viewport that the embedding element has established for us.
return {aCBSize, AspectRatioUsage::None};
LogicalSize cbSize = aCBSize;
IntrinsicSize intrinsicSize = GetIntrinsicSize();
if (mIsRootContent) {
// We're the root of the outermost browsing context, so we need to scale
// cbSize by the full-zoom so that SVGs with percentage width/height zoom:
aCBSize.BSize(aWritingMode) != NS_UNCONSTRAINEDSIZE,
"root should not have auto-width/height containing block");
if (!mIsInIframe) {
// NOTE: We can't just use mFullZoom because this can run before Reflow()
// updates it.
const float zoom = ComputeFullZoom();
cbSize.ISize(aWritingMode) *= zoom;
cbSize.BSize(aWritingMode) *= zoom;
// We also need to honour the width and height attributes' default values
// of 100% when we're the root of a browsing context. (GetIntrinsicSize()
// doesn't report these since there's no such thing as a percentage
// intrinsic size. Also note that explicit percentage values are mapped
// into style, so the following isn't for them.)
auto* content = static_cast<SVGSVGElement*>(GetContent());
const SVGAnimatedLength& width =
if (width.IsPercentage()) {
"GetIntrinsicSize should have reported no intrinsic width");
float val = width.GetAnimValInSpecifiedUnits() / 100.0f;
intrinsicSize.width.emplace(std::max(val, 0.0f) *
const SVGAnimatedLength& height =
"root should not have auto-height containing block");
if (height.IsPercentage()) {
"GetIntrinsicSize should have reported no intrinsic height");
float val = height.GetAnimValInSpecifiedUnits() / 100.0f;
intrinsicSize.height.emplace(std::max(val, 0.0f) *
MOZ_ASSERT(intrinsicSize.height && intrinsicSize.width,
"We should have just handled the only situation where"
"we lack an intrinsic height or width.");
return {ComputeSizeWithIntrinsicDimensions(
aRenderingContext, aWritingMode, intrinsicSize, GetAspectRatio(),
cbSize, aMargin, aBorderPadding, aSizeOverrides, aFlags),
void SVGOuterSVGFrame::Reflow(nsPresContext* aPresContext,
ReflowOutput& aDesiredSize,
const ReflowInput& aReflowInput,
nsReflowStatus& aStatus) {
MOZ_ASSERT(aStatus.IsEmpty(), "Caller should pass a fresh reflow status!");
("enter SVGOuterSVGFrame::Reflow: availSize=%d,%d",
aReflowInput.AvailableWidth(), aReflowInput.AvailableHeight()));
MOZ_ASSERT(HasAnyStateBits(NS_FRAME_IN_REFLOW), "frame is not in reflow");
const auto wm = GetWritingMode();
aDesiredSize.SetSize(wm, aReflowInput.ComputedSizeWithBorderPadding(wm));
NS_ASSERTION(!GetPrevInFlow(), "SVG can't currently be broken across pages.");
SVGSVGElement* svgElem = static_cast<SVGSVGElement*>(GetContent());
auto* anonKid = static_cast<SVGOuterSVGAnonChildFrame*>(
if (HasAnyStateBits(NS_FRAME_FIRST_REFLOW)) {
// Initialize
// If our SVG viewport has changed, update our content and notify.
gfx::Size newViewportSize(
uint32_t changeBits = 0;
if (newViewportSize != svgElem->GetViewportSize()) {
// When our viewport size changes, we may need to update the overflow rects
// of our child frames. This is the case if:
// * We have a real/synthetic viewBox (a children-only transform), since
// the viewBox transform will change as the viewport dimensions change.
// * We do not have a real/synthetic viewBox, but the last time we
// reflowed (or the last time UpdateOverflow() was called) we did.
// We only handle the former case here, in which case we mark all our child
// frames as dirty so that we reflow them below and update their overflow
// rects.
// In the latter case, updating of overflow rects is handled for removal of
// real viewBox (the viewBox attribute) in AttributeChanged. Synthetic
// viewBox "removal" (e.g. a document references the same SVG via both an
// <svg:image> and then as a CSS background image (a synthetic viewBox is
// used when painting the former, but not when painting the latter)) is
// handled in SVGSVGElement::FlushImageTransformInvalidation.
if (svgElem->HasViewBoxOrSyntheticViewBox()) {
nsIFrame* anonChild = PrincipalChildList().FirstChild();
for (nsIFrame* child : anonChild->PrincipalChildList()) {
if (mIsRootContent && !mIsInIframe) {
const auto oldZoom = mFullZoom;
mFullZoom = ComputeFullZoom();
if (oldZoom != mFullZoom) {
changeBits |= FULL_ZOOM_CHANGED;
if (changeBits && !HasAnyStateBits(NS_FRAME_FIRST_REFLOW)) {
// Now that we've marked the necessary children as dirty, call
// ReflowSVG() or ReflowSVGNonDisplayText() on them, depending
// on whether we are non-display.
mCallingReflowSVG = true;
if (HasAnyStateBits(NS_FRAME_IS_NONDISPLAY)) {
} else {
// Update the mRects and ink overflow rects of all our descendants,
// including our anonymous wrapper kid:
"We should have one anonymous child frame wrapping our real "
mCallingReflowSVG = false;
// Set our anonymous kid's offset from our border box:
// Including our size in our overflow rects regardless of the value of
// 'background', 'border', etc. makes sure that we usually (when we clip to
// our content area) don't have to keep changing our overflow rects as our
// descendants move about (see perf comment below). Including our size in our
// scrollable overflow rect also makes sure that we scroll if we're too big
// for our viewport.
// <svg> never allows scrolling to anything outside its mRect (only panning),
// so we must always keep our scrollable overflow set to our size.
// With regards to ink overflow, we always clip root-<svg> (see our
// BuildDisplayList method) regardless of the value of the 'overflow'
// property since that is per-spec, even for the initial 'visible' value. For
// that reason there's no point in adding descendant ink overflow to our
// own when this frame is for a root-<svg>. That said, there's also a very
// good performance reason for us wanting to avoid doing so. If we did, then
// the frame's overflow would often change as descendants that are partially
// or fully outside its rect moved (think animation on/off screen), and that
// would cause us to do a full NS_FRAME_IS_DIRTY reflow and repaint of the
// entire document tree each such move (see bug 875175).
// So it's only non-root outer-<svg> that has the ink overflow of its
// descendants added to its own. (Note that the default user-agent style
// sheet makes 'hidden' the default value for :not(root(svg)), so usually
// FinishAndStoreOverflow will still clip this back to the frame's rect.)
// WARNING!! Keep UpdateBounds below in sync with whatever we do for our
// overflow rects here! (Again, see bug 875175.)
// An outer SVG will be here as a nondisplay if it fails the conditional
// processing test. In that case, we don't maintain its overflow.
if (!HasAnyStateBits(NS_FRAME_IS_NONDISPLAY)) {
if (!mIsRootContent) {
anonKid->InkOverflowRect() + anonKid->GetPosition());
("exit SVGOuterSVGFrame::Reflow: size=%d,%d",
aDesiredSize.Width(), aDesiredSize.Height()));
void SVGOuterSVGFrame::DidReflow(nsPresContext* aPresContext,
const ReflowInput* aReflowInput) {
SVGDisplayContainerFrame::DidReflow(aPresContext, aReflowInput);
// Make sure elements styled by :hover get updated if script/animation moves
// them under or out from under the pointer:
/* virtual */
void SVGOuterSVGFrame::UnionChildOverflow(OverflowAreas& aOverflowAreas) {
// See the comments in Reflow above.
// WARNING!! Keep this in sync with Reflow above!
if (!mIsRootContent) {
nsIFrame* anonKid = PrincipalChildList().FirstChild();
anonKid->InkOverflowRect() + anonKid->GetPosition());
// container methods
nsresult SVGOuterSVGFrame::AttributeChanged(int32_t aNameSpaceID,
nsAtom* aAttribute,
int32_t aModType) {
if (aNameSpaceID == kNameSpaceID_None &&
if (aAttribute == nsGkAtoms::viewBox ||
aAttribute == nsGkAtoms::preserveAspectRatio ||
aAttribute == nsGkAtoms::transform) {
// make sure our cached transform matrix gets (lazily) updated
mCanvasTM = nullptr;
aAttribute == nsGkAtoms::viewBox
if (aAttribute != nsGkAtoms::transform) {
if (aAttribute == nsGkAtoms::width || aAttribute == nsGkAtoms::height ||
aAttribute == nsGkAtoms::viewBox) {
// Don't call ChildrenOnlyTransformChanged() here, since we call it
// under Reflow if the width/height/viewBox actually changed.
if (!mIsInObjectOrEmbed) {
// We are not embedded by reference, so our 'width' and 'height'
// attributes are not overridden (and viewBox may influence our
// intrinsic aspect ratio). We need to reflow.
this, IntrinsicDirty::FrameAncestorsAndDescendants,
return NS_OK;
bool SVGOuterSVGFrame::IsSVGTransformed(Matrix* aOwnTransform,
Matrix* aFromParentTransform) const {
// Our anonymous child's HasChildrenOnlyTransform() implementation makes sure
// our children-only transforms are applied to our children. We only care
// about transforms that transform our own frame here.
bool foundTransform = false;
SVGSVGElement* content = static_cast<SVGSVGElement*>(GetContent());
SVGAnimatedTransformList* transformList = content->GetAnimatedTransformList();
if ((transformList && transformList->HasTransform()) ||
content->GetAnimateMotionTransform()) {
if (aOwnTransform) {
*aOwnTransform = gfx::ToMatrix(
content->PrependLocalTransformsTo(gfxMatrix(), eUserSpaceToParent));
foundTransform = true;
return foundTransform;
// painting
void SVGOuterSVGFrame::BuildDisplayList(nsDisplayListBuilder* aBuilder,
const nsDisplayListSet& aLists) {
if (HasAnyStateBits(NS_FRAME_IS_NONDISPLAY)) {
DisplayBorderBackgroundOutline(aBuilder, aLists);
nsRect visibleRect = aBuilder->GetVisibleRect();
nsRect dirtyRect = aBuilder->GetDirtyRect();
// Per-spec, we always clip root-<svg> even when 'overflow' has its initial
// value of 'visible'. See also the "ink overflow" comments in Reflow.
DisplayListClipState::AutoSaveRestore autoSR(aBuilder);
if (mIsRootContent || StyleDisplay()->IsScrollableOverflow()) {
autoSR.ClipContainingBlockDescendantsToContentBox(aBuilder, this);
visibleRect = visibleRect.Intersect(GetContentRectRelativeToSelf());
dirtyRect = dirtyRect.Intersect(GetContentRectRelativeToSelf());
nsDisplayListBuilder::AutoBuildingDisplayList building(
aBuilder, this, visibleRect, dirtyRect);
nsDisplayList* contentList = aLists.Content();
nsDisplayListSet set(contentList, contentList, contentList, contentList,
contentList, contentList);
BuildDisplayListForNonBlockChildren(aBuilder, set);
// ISVGSVGFrame methods:
void SVGOuterSVGFrame::NotifyViewportOrTransformChanged(uint32_t aFlags) {
"Unexpected aFlags value");
SVGSVGElement* content = static_cast<SVGSVGElement*>(GetContent());
if (content->HasViewBox()) {
// Percentage lengths on children resolve against the viewBox rect so we
// don't need to notify them of the viewport change, but the viewBox
// transform will have changed, so we need to notify them of that instead.
} else if (content->ShouldSynthesizeViewBox()) {
// In the case of a synthesized viewBox, the synthetic viewBox's rect
// changes as the viewport changes. As a result we need to maintain the
} else if (mCanvasTM && mCanvasTM->IsSingular()) {
// A width/height of zero will result in us having a singular mCanvasTM
// even when we don't have a viewBox. So we also want to recompute our
// mCanvasTM for this width/height change even though we don't have a
// viewBox.
bool haveNonFulLZoomTransformChange = (aFlags & TRANSFORM_CHANGED);
if (aFlags & FULL_ZOOM_CHANGED) {
// Make sure our canvas transform matrix gets (lazily) recalculated:
mCanvasTM = nullptr;
if (haveNonFulLZoomTransformChange &&
uint32_t flags = HasAnyStateBits(NS_FRAME_IN_REFLOW)
? SVGSVGElement::eDuringReflow
: 0;
// ISVGDisplayableFrame methods:
void SVGOuterSVGFrame::PaintSVG(gfxContext& aContext,
const gfxMatrix& aTransform,
imgDrawingParams& aImgParams) {
PrincipalChildList().FirstChild()->IsSVGOuterSVGAnonChildFrame() &&
"We should have a single, anonymous, child");
auto* anonKid = static_cast<SVGOuterSVGAnonChildFrame*>(
anonKid->PaintSVG(aContext, aTransform, aImgParams);
SVGBBox SVGOuterSVGFrame::GetBBoxContribution(
const gfx::Matrix& aToBBoxUserspace, uint32_t aFlags) {
PrincipalChildList().FirstChild()->IsSVGOuterSVGAnonChildFrame() &&
"We should have a single, anonymous, child");
// We must defer to our child so that we don't include our
// content->PrependLocalTransformsTo() transforms.
auto* anonKid = static_cast<SVGOuterSVGAnonChildFrame*>(
return anonKid->GetBBoxContribution(aToBBoxUserspace, aFlags);
// SVGContainerFrame methods:
gfxMatrix SVGOuterSVGFrame::GetCanvasTM() {
if (!mCanvasTM) {
SVGSVGElement* content = static_cast<SVGSVGElement*>(GetContent());
float devPxPerCSSPx = 1.0f / nsPresContext::AppUnitsToFloatCSSPixels(
gfxMatrix tm = content->PrependLocalTransformsTo(
gfxMatrix::Scaling(devPxPerCSSPx, devPxPerCSSPx));
mCanvasTM = MakeUnique<gfxMatrix>(tm);
return *mCanvasTM;
// Implementation helpers
bool SVGOuterSVGFrame::IsRootOfImage() {
if (!mContent->GetParent()) {
// Our content is the document element
Document* doc = mContent->GetUncomposedDoc();
if (doc && doc->IsBeingUsedAsImage()) {
// Our document is being used as an image
return true;
return false;
bool SVGOuterSVGFrame::VerticalScrollbarNotNeeded() const {
const SVGAnimatedLength& height =
return height.IsPercentage() && height.GetBaseValInSpecifiedUnits() <= 100;
void SVGOuterSVGFrame::AppendDirectlyOwnedAnonBoxes(
nsTArray<OwnedAnonBox>& aResult) {
nsIFrame* anonKid = PrincipalChildList().FirstChild();
void SVGOuterSVGFrame::MaybeSendIntrinsicSizeAndRatioToEmbedder() {
void SVGOuterSVGFrame::MaybeSendIntrinsicSizeAndRatioToEmbedder(
Maybe<IntrinsicSize> aIntrinsicSize, Maybe<AspectRatio> aIntrinsicRatio) {
if (!mIsInObjectOrEmbed) {
nsCOMPtr<nsIDocShell> docShell = PresContext()->GetDocShell();
if (!docShell) {
BrowsingContext* bc = docShell->GetBrowsingContext();
if (bc->GetParent()->IsInProcess()) {
if (Element* embedder = bc->GetEmbedderElement()) {
if (nsCOMPtr<nsIObjectLoadingContent> olc = do_QueryInterface(embedder)) {
if (BrowserChild* browserChild = BrowserChild::GetFrom(docShell)) {
Unused << browserChild->SendIntrinsicSizeOrRatioChanged(aIntrinsicSize,
void SVGOuterSVGFrame::DidSetComputedStyle(ComputedStyle* aOldComputedStyle) {
if (!aOldComputedStyle) {
if (aOldComputedStyle->StylePosition()->mAspectRatio !=
StylePosition()->mAspectRatio) {
// Our aspect-ratio property value changed, and an embedding <object> or
// <embed> might care about that.
void SVGOuterSVGFrame::Destroy(DestroyContext& aContext) {
// This handles both the case when the root <svg> element is made display:none
// (and thus loses its intrinsic size and aspect ratio), and when the frame
// is navigated elsewhere & we need to reset parent <object>/<embed>'s
// recorded intrinsic size/ratio values.
MaybeSendIntrinsicSizeAndRatioToEmbedder(Nothing(), Nothing());
} // namespace mozilla
// Implementation of SVGOuterSVGAnonChildFrame
nsContainerFrame* NS_NewSVGOuterSVGAnonChildFrame(
mozilla::PresShell* aPresShell, mozilla::ComputedStyle* aStyle) {
return new (aPresShell)
mozilla::SVGOuterSVGAnonChildFrame(aStyle, aPresShell->GetPresContext());
namespace mozilla {
#ifdef DEBUG
void SVGOuterSVGAnonChildFrame::Init(nsIContent* aContent,
nsContainerFrame* aParent,
nsIFrame* aPrevInFlow) {
MOZ_ASSERT(aParent->IsSVGOuterSVGFrame(), "Unexpected parent");
SVGDisplayContainerFrame::Init(aContent, aParent, aPrevInFlow);
void SVGOuterSVGAnonChildFrame::BuildDisplayList(
nsDisplayListBuilder* aBuilder, const nsDisplayListSet& aLists) {
// Wrap our contents into an nsDisplaySVGWrapper.
// We wrap this frame instead of the SVGOuterSVGFrame so that the wrapper
// doesn't contain the <svg> element's CSS styles, like backgrounds or
// borders. Creating the nsDisplaySVGWrapper here also means that it'll be
// inside the nsDisplayTransform for our viewbox transform. The
// nsDisplaySVGWrapper's reference frame is this frame, because this frame
// always returns true from IsSVGTransformed.
nsDisplayList newList(aBuilder);
nsDisplayListSet set(&newList, &newList, &newList, &newList, &newList,
BuildDisplayListForNonBlockChildren(aBuilder, set);
aLists.Content()->AppendNewToTop<nsDisplaySVGWrapper>(aBuilder, this,
static Matrix ComputeOuterSVGAnonChildFrameTransform(
const SVGOuterSVGAnonChildFrame* aFrame) {
// Our elements 'transform' attribute is applied to our SVGOuterSVGFrame
// parent, and the element's children-only transforms are applied to us, the
// anonymous child frame. Since we are the child frame, we apply the
// children-only transforms as if they are our own transform.
SVGSVGElement* content = static_cast<SVGSVGElement*>(aFrame->GetContent());
if (!content->HasChildrenOnlyTransform()) {
return Matrix();
// Outer-<svg> doesn't use x/y, so we can pass eChildToUserSpace here.
gfxMatrix ownMatrix =
content->PrependLocalTransformsTo(gfxMatrix(), eChildToUserSpace);
if (ownMatrix.HasNonTranslation()) {
// viewBox, currentScale and currentTranslate should only produce a
// rectilinear transform.
"Non-rectilinear transform will break the following logic");
// The nsDisplayTransform code will apply this transform to our frame,
// including to our frame position. We don't want our frame position to
// be scaled though, so we need to correct for that in the transform.
// XXX Yeah, this is a bit hacky.
CSSPoint pos = CSSPixel::FromAppUnits(aFrame->GetPosition());
CSSPoint scaledPos = CSSPoint(ownMatrix._11 * pos.x, ownMatrix._22 * pos.y);
CSSPoint deltaPos = scaledPos - pos;
ownMatrix *= gfxMatrix::Translation(-deltaPos.x, -deltaPos.y);
return gfx::ToMatrix(ownMatrix);
// We want this frame to be a reference frame. An easy way to achieve that is
// to always return true from this method, even for identity transforms.
// This frame being a reference frame ensures that the offset between this
// <svg> element and the parent reference frame is completely absorbed by the
// nsDisplayTransform that's created for this frame, and that this offset does
// not affect our descendants' transforms. Consequently, if the <svg> element
// moves, e.g. during scrolling, the transform matrices of our contents are
// unaffected. This simplifies invalidation.
bool SVGOuterSVGAnonChildFrame::IsSVGTransformed(
Matrix* aOwnTransform, Matrix* aFromParentTransform) const {
if (aOwnTransform) {
*aOwnTransform = ComputeOuterSVGAnonChildFrameTransform(this);
return true;
} // namespace mozilla