gecko-dev/layout/style/nsAnimationManager.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

612 строки
23 KiB
C++
Исходник Обычный вид История

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
2012-05-21 15:12:37 +04:00
/* 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/. */
#include "nsAnimationManager.h"
#include "nsTransitionManager.h"
#include "mozilla/dom/CSSAnimationBinding.h"
#include "mozilla/AnimationEventDispatcher.h"
#include "mozilla/AnimationTarget.h"
#include "mozilla/EffectCompositor.h"
#include "mozilla/EffectSet.h"
#include "mozilla/MemoryReporting.h"
#include "mozilla/ServoStyleSet.h"
#include "mozilla/StyleAnimationValue.h"
#include "mozilla/dom/AnimationEffect.h"
#include "mozilla/dom/DocumentTimeline.h"
#include "mozilla/dom/KeyframeEffect.h"
#include "mozilla/dom/MutationObservers.h"
#include "nsPresContext.h"
#include "nsPresContextInlines.h"
#include "nsStyleChangeList.h"
#include "nsLayoutUtils.h"
#include "nsIFrame.h"
#include "mozilla/dom/Document.h"
#include "nsDOMMutationObserver.h"
#include "nsRFPService.h"
#include <algorithm> // std::stable_sort
#include <math.h>
using namespace mozilla;
using namespace mozilla::css;
using mozilla::dom::Animation;
using mozilla::dom::AnimationEffect;
using mozilla::dom::AnimationPlayState;
using mozilla::dom::CSSAnimation;
using mozilla::dom::KeyframeEffect;
typedef mozilla::ComputedTiming::AnimationPhase AnimationPhase;
////////////////////////// CSSAnimation ////////////////////////////
JSObject* CSSAnimation::WrapObject(JSContext* aCx,
JS::Handle<JSObject*> aGivenProto) {
return dom::CSSAnimation_Binding::Wrap(aCx, this, aGivenProto);
}
mozilla::dom::Promise* CSSAnimation::GetReady(ErrorResult& aRv) {
FlushUnanimatedStyle();
return Animation::GetReady(aRv);
}
void CSSAnimation::Play(ErrorResult& aRv, LimitBehavior aLimitBehavior) {
mPauseShouldStick = false;
Animation::Play(aRv, aLimitBehavior);
}
void CSSAnimation::Pause(ErrorResult& aRv) {
mPauseShouldStick = true;
Animation::Pause(aRv);
}
AnimationPlayState CSSAnimation::PlayStateFromJS() const {
// Flush style to ensure that any properties controlling animation state
// (e.g. animation-play-state) are fully updated.
FlushUnanimatedStyle();
return Animation::PlayStateFromJS();
}
bool CSSAnimation::PendingFromJS() const {
// Flush style since, for example, if the animation-play-state was just
// changed its possible we should now be pending.
FlushUnanimatedStyle();
return Animation::PendingFromJS();
}
void CSSAnimation::PlayFromJS(ErrorResult& aRv) {
// Note that flushing style below might trigger calls to
// PlayFromStyle()/PauseFromStyle() on this object.
FlushUnanimatedStyle();
Animation::PlayFromJS(aRv);
}
void CSSAnimation::PlayFromStyle() {
mIsStylePaused = false;
if (!mPauseShouldStick) {
ErrorResult rv;
Animation::Play(rv, Animation::LimitBehavior::Continue);
// play() should not throw when LimitBehavior is Continue
MOZ_ASSERT(!rv.Failed(), "Unexpected exception playing animation");
}
}
void CSSAnimation::PauseFromStyle() {
// Check if the pause state is being overridden
if (mIsStylePaused) {
return;
}
mIsStylePaused = true;
ErrorResult rv;
Animation::Pause(rv);
// pause() should only throw when *all* of the following conditions are true:
// - we are in the idle state, and
// - we have a negative playback rate, and
// - we have an infinitely repeating animation
// The first two conditions will never happen under regular style processing
// but could happen if an author made modifications to the Animation object
// and then updated animation-play-state. It's an unusual case and there's
// no obvious way to pass on the exception information so we just silently
// fail for now.
if (rv.Failed()) {
NS_WARNING("Unexpected exception pausing animation - silently failing");
}
}
void CSSAnimation::Tick() {
Animation::Tick();
QueueEvents();
}
bool CSSAnimation::HasLowerCompositeOrderThan(
const CSSAnimation& aOther) const {
MOZ_ASSERT(IsTiedToMarkup() && aOther.IsTiedToMarkup(),
"Should only be called for CSS animations that are sorted "
"as CSS animations (i.e. tied to CSS markup)");
// 0. Object-equality case
if (&aOther == this) {
return false;
}
// 1. Sort by document order
if (!mOwningElement.Equals(aOther.mOwningElement)) {
return mOwningElement.LessThan(
const_cast<CSSAnimation*>(this)->CachedChildIndexRef(),
aOther.mOwningElement,
const_cast<CSSAnimation*>(&aOther)->CachedChildIndexRef());
}
// 2. (Same element and pseudo): Sort by position in animation-name
return mAnimationIndex < aOther.mAnimationIndex;
}
void CSSAnimation::QueueEvents(const StickyTimeDuration& aActiveTime) {
// If the animation is pending, we ignore animation events until we finish
// pending.
if (mPendingState != PendingState::NotPending) {
return;
}
// CSS animations dispatch events at their owning element. This allows
// script to repurpose a CSS animation to target a different element,
// to use a group effect (which has no obvious "target element"), or
// to remove the animation effect altogether whilst still getting
// animation events.
//
// It does mean, however, that for a CSS animation that has no owning
// element (e.g. it was created using the CSSAnimation constructor or
// disassociated from CSS) no events are fired. If it becomes desirable
// for these animations to still fire events we should spec the concept
// of the "original owning element" or "event target" and allow script
// to set it when creating a CSSAnimation object.
if (!mOwningElement.IsSet()) {
return;
}
nsPresContext* presContext = mOwningElement.GetPresContext();
if (!presContext) {
return;
}
uint64_t currentIteration = 0;
ComputedTiming::AnimationPhase currentPhase;
StickyTimeDuration intervalStartTime;
StickyTimeDuration intervalEndTime;
StickyTimeDuration iterationStartTime;
if (!mEffect) {
currentPhase =
GetAnimationPhaseWithoutEffect<ComputedTiming::AnimationPhase>(*this);
if (currentPhase == mPreviousPhase) {
return;
}
} else {
ComputedTiming computedTiming = mEffect->GetComputedTiming();
currentPhase = computedTiming.mPhase;
currentIteration = computedTiming.mCurrentIteration;
if (currentPhase == mPreviousPhase &&
currentIteration == mPreviousIteration) {
return;
}
intervalStartTime = IntervalStartTime(computedTiming.mActiveDuration);
intervalEndTime = IntervalEndTime(computedTiming.mActiveDuration);
uint64_t iterationBoundary = mPreviousIteration > currentIteration
? currentIteration + 1
: currentIteration;
iterationStartTime = computedTiming.mDuration.MultDouble(
(iterationBoundary - computedTiming.mIterationStart));
}
TimeStamp startTimeStamp = ElapsedTimeToTimeStamp(intervalStartTime);
TimeStamp endTimeStamp = ElapsedTimeToTimeStamp(intervalEndTime);
TimeStamp iterationTimeStamp = ElapsedTimeToTimeStamp(iterationStartTime);
AutoTArray<AnimationEventInfo, 2> events;
auto appendAnimationEvent = [&](EventMessage aMessage,
const StickyTimeDuration& aElapsedTime,
const TimeStamp& aScheduledEventTimeStamp) {
double elapsedTime = aElapsedTime.ToSeconds();
if (aMessage == eAnimationCancel) {
// 0 is an inappropriate value for this callsite. What we need to do is
// use a single random value for all increasing times reportable.
// That is to say, whenever elapsedTime goes negative (because an
// animation restarts, something rewinds the animation, or otherwise)
// a new random value for the mix-in must be generated.
elapsedTime = nsRFPService::ReduceTimePrecisionAsSecs(
elapsedTime, 0, TimerPrecisionType::RFPOnly);
}
events.AppendElement(
AnimationEventInfo(mAnimationName, mOwningElement.Target(), aMessage,
elapsedTime, aScheduledEventTimeStamp, this));
};
// Handle cancel event first
if ((mPreviousPhase != AnimationPhase::Idle &&
mPreviousPhase != AnimationPhase::After) &&
currentPhase == AnimationPhase::Idle) {
appendAnimationEvent(eAnimationCancel, aActiveTime,
GetTimelineCurrentTimeAsTimeStamp());
}
switch (mPreviousPhase) {
case AnimationPhase::Idle:
case AnimationPhase::Before:
if (currentPhase == AnimationPhase::Active) {
appendAnimationEvent(eAnimationStart, intervalStartTime,
startTimeStamp);
} else if (currentPhase == AnimationPhase::After) {
appendAnimationEvent(eAnimationStart, intervalStartTime,
startTimeStamp);
appendAnimationEvent(eAnimationEnd, intervalEndTime, endTimeStamp);
}
break;
case AnimationPhase::Active:
if (currentPhase == AnimationPhase::Before) {
appendAnimationEvent(eAnimationEnd, intervalStartTime, startTimeStamp);
} else if (currentPhase == AnimationPhase::Active) {
// The currentIteration must have changed or element we would have
// returned early above.
MOZ_ASSERT(currentIteration != mPreviousIteration);
appendAnimationEvent(eAnimationIteration, iterationStartTime,
iterationTimeStamp);
} else if (currentPhase == AnimationPhase::After) {
appendAnimationEvent(eAnimationEnd, intervalEndTime, endTimeStamp);
}
break;
case AnimationPhase::After:
if (currentPhase == AnimationPhase::Before) {
appendAnimationEvent(eAnimationStart, intervalEndTime, startTimeStamp);
appendAnimationEvent(eAnimationEnd, intervalStartTime, endTimeStamp);
} else if (currentPhase == AnimationPhase::Active) {
appendAnimationEvent(eAnimationStart, intervalEndTime, endTimeStamp);
}
break;
}
mPreviousPhase = currentPhase;
mPreviousIteration = currentIteration;
if (!events.IsEmpty()) {
presContext->AnimationEventDispatcher()->QueueEvents(std::move(events));
}
}
void CSSAnimation::UpdateTiming(SeekFlag aSeekFlag,
SyncNotifyFlag aSyncNotifyFlag) {
if (mNeedsNewAnimationIndexWhenRun &&
PlayState() != AnimationPlayState::Idle) {
mAnimationIndex = sNextAnimationIndex++;
mNeedsNewAnimationIndexWhenRun = false;
}
Animation::UpdateTiming(aSeekFlag, aSyncNotifyFlag);
}
////////////////////////// nsAnimationManager ////////////////////////////
// Find the matching animation by |aName| in the old list
// of animations and remove the matched animation from the list.
static already_AddRefed<CSSAnimation> PopExistingAnimation(
const nsAtom* aName,
nsAnimationManager::CSSAnimationCollection* aCollection) {
if (!aCollection) {
return nullptr;
}
// Animations are stored in reverse order to how they appear in the
// animation-name property. However, we want to match animations beginning
// from the end of the animation-name list, so we iterate *forwards*
// through the collection.
for (size_t idx = 0, length = aCollection->mAnimations.Length();
idx != length; ++idx) {
CSSAnimation* cssAnim = aCollection->mAnimations[idx];
if (cssAnim->AnimationName() == aName) {
RefPtr<CSSAnimation> match = cssAnim;
aCollection->mAnimations.RemoveElementAt(idx);
return match.forget();
}
}
return nullptr;
}
class MOZ_STACK_CLASS ServoCSSAnimationBuilder final {
public:
explicit ServoCSSAnimationBuilder(const ComputedStyle* aComputedStyle)
: mComputedStyle(aComputedStyle) {
MOZ_ASSERT(aComputedStyle);
}
bool BuildKeyframes(const Element& aElement, nsPresContext* aPresContext,
nsAtom* aName, const nsTimingFunction& aTimingFunction,
nsTArray<Keyframe>& aKeyframes) {
return aPresContext->StyleSet()->GetKeyframesForName(
aElement, *mComputedStyle, aName, aTimingFunction, aKeyframes);
}
void SetKeyframes(KeyframeEffect& aEffect, nsTArray<Keyframe>&& aKeyframes) {
aEffect.SetKeyframes(std::move(aKeyframes), mComputedStyle);
}
Bug 1405548 - Post restyles when creating or removing new CSS animations when using the Servo backend; r=hiro As explained in the extended comment in this patch, for Servo we want to post restyles when creating new animations so that we run a second animation restyle and incorporate the result of new animations into style immediately. (Gecko does everything in the one restyle, and although this causes other bugs related to triggering transitions, at least it means it does not require restyles to be posted here). It turns out that we normally end up posting a restyle anyway in CSSAnimation::SetAnimationIndex. Bug 1332958 was supposed to drop that but it never landed. However, CSSAnimation::SetAnimationIndex only posts a restyle when there is a change to the animation index. It turns out that, by chance, there normally *is* a change to a CSSAnimation's animation index when it is created. Initially it takes its animation index from Animation::sNextAnimationIndex which is incremented each time it is assigned to an animation. If the first Animation we create for a given content process is a CSSAnimation then sNextAnimationIndex will be zero and so we will initially assign an animation index of zero. If that CSS animation is also the first in the list of animations in animation-name, when we call SetAnimationIndex we will pass zero as the index to use, and when we go to update the animation index we will detect that there is no change, and will NOT post an animation restyle. As a result the target element's style will NOT reflect the animated style. To fix this we need to ensure that *new* CSS animations trigger a restyle. For *changes* to animations, the corresponding calls to SetKeyframes and SetSpecifiedTiming post restyles so the behavior should be correct in those cases. For *removed* animations I observed that in at least some cases we successfully post a restyle. However, this appeared to be as much by chance as anything so this patch also posts a restyle for removed animations. (Note that the EffectCompositor will ignore redundant restyle requests so this is ok.) This patch deliberately does not expose Animation::PostUpdate and call that because the code introduced here is intended to be temporary. Long-term we should remove the Gecko style backend and allow the calls to PlayFromStyle, PauseFromStyle, CancelFromStyle etc. to post restyles just like calls to Play, Pause, and Cancel do. At that point this code can also be removed. MozReview-Commit-ID: 4c3vJdLBqeY --HG-- extra : rebase_source : 684cb483562709161b2d5635e173e55319509a70
2017-10-11 10:36:36 +03:00
// Currently all the animation building code in this file is based on
// assumption that creating and removing animations should *not* trigger
// additional restyles since those changes will be handled within the same
// restyle.
//
// While that is true for the Gecko style backend, it is not true for the
// Servo style backend where we want restyles to be triggered so that we
// perform a second animation restyle where we will incorporate the changes
// arising from creating and removing animations.
//
// Fortunately, our attempts to avoid posting extra restyles as part of the
// processing here are imperfect and most of the time we happen to post
// them anyway. Occasionally, however, we don't. For example, we don't post
// a restyle when we create a new animation whose an animation index matches
// the default value it was given already (which is typically only true when
// the CSSAnimation we create is the first Animation created in a particular
// content process).
//
// As a result, when we are using the Servo backend, whenever we have an added
// or removed animation we need to explicitly trigger a restyle.
//
// This code should eventually disappear along with the Gecko style backend
// and we should simply call Play() / Pause() / Cancel() etc. which will
// post the required restyles.
void NotifyNewOrRemovedAnimation(const Animation& aAnimation) {
dom::AnimationEffect* effect = aAnimation.GetEffect();
Bug 1405548 - Post restyles when creating or removing new CSS animations when using the Servo backend; r=hiro As explained in the extended comment in this patch, for Servo we want to post restyles when creating new animations so that we run a second animation restyle and incorporate the result of new animations into style immediately. (Gecko does everything in the one restyle, and although this causes other bugs related to triggering transitions, at least it means it does not require restyles to be posted here). It turns out that we normally end up posting a restyle anyway in CSSAnimation::SetAnimationIndex. Bug 1332958 was supposed to drop that but it never landed. However, CSSAnimation::SetAnimationIndex only posts a restyle when there is a change to the animation index. It turns out that, by chance, there normally *is* a change to a CSSAnimation's animation index when it is created. Initially it takes its animation index from Animation::sNextAnimationIndex which is incremented each time it is assigned to an animation. If the first Animation we create for a given content process is a CSSAnimation then sNextAnimationIndex will be zero and so we will initially assign an animation index of zero. If that CSS animation is also the first in the list of animations in animation-name, when we call SetAnimationIndex we will pass zero as the index to use, and when we go to update the animation index we will detect that there is no change, and will NOT post an animation restyle. As a result the target element's style will NOT reflect the animated style. To fix this we need to ensure that *new* CSS animations trigger a restyle. For *changes* to animations, the corresponding calls to SetKeyframes and SetSpecifiedTiming post restyles so the behavior should be correct in those cases. For *removed* animations I observed that in at least some cases we successfully post a restyle. However, this appeared to be as much by chance as anything so this patch also posts a restyle for removed animations. (Note that the EffectCompositor will ignore redundant restyle requests so this is ok.) This patch deliberately does not expose Animation::PostUpdate and call that because the code introduced here is intended to be temporary. Long-term we should remove the Gecko style backend and allow the calls to PlayFromStyle, PauseFromStyle, CancelFromStyle etc. to post restyles just like calls to Play, Pause, and Cancel do. At that point this code can also be removed. MozReview-Commit-ID: 4c3vJdLBqeY --HG-- extra : rebase_source : 684cb483562709161b2d5635e173e55319509a70
2017-10-11 10:36:36 +03:00
if (!effect) {
return;
}
KeyframeEffect* keyframeEffect = effect->AsKeyframeEffect();
Bug 1405548 - Post restyles when creating or removing new CSS animations when using the Servo backend; r=hiro As explained in the extended comment in this patch, for Servo we want to post restyles when creating new animations so that we run a second animation restyle and incorporate the result of new animations into style immediately. (Gecko does everything in the one restyle, and although this causes other bugs related to triggering transitions, at least it means it does not require restyles to be posted here). It turns out that we normally end up posting a restyle anyway in CSSAnimation::SetAnimationIndex. Bug 1332958 was supposed to drop that but it never landed. However, CSSAnimation::SetAnimationIndex only posts a restyle when there is a change to the animation index. It turns out that, by chance, there normally *is* a change to a CSSAnimation's animation index when it is created. Initially it takes its animation index from Animation::sNextAnimationIndex which is incremented each time it is assigned to an animation. If the first Animation we create for a given content process is a CSSAnimation then sNextAnimationIndex will be zero and so we will initially assign an animation index of zero. If that CSS animation is also the first in the list of animations in animation-name, when we call SetAnimationIndex we will pass zero as the index to use, and when we go to update the animation index we will detect that there is no change, and will NOT post an animation restyle. As a result the target element's style will NOT reflect the animated style. To fix this we need to ensure that *new* CSS animations trigger a restyle. For *changes* to animations, the corresponding calls to SetKeyframes and SetSpecifiedTiming post restyles so the behavior should be correct in those cases. For *removed* animations I observed that in at least some cases we successfully post a restyle. However, this appeared to be as much by chance as anything so this patch also posts a restyle for removed animations. (Note that the EffectCompositor will ignore redundant restyle requests so this is ok.) This patch deliberately does not expose Animation::PostUpdate and call that because the code introduced here is intended to be temporary. Long-term we should remove the Gecko style backend and allow the calls to PlayFromStyle, PauseFromStyle, CancelFromStyle etc. to post restyles just like calls to Play, Pause, and Cancel do. At that point this code can also be removed. MozReview-Commit-ID: 4c3vJdLBqeY --HG-- extra : rebase_source : 684cb483562709161b2d5635e173e55319509a70
2017-10-11 10:36:36 +03:00
if (!keyframeEffect) {
return;
}
keyframeEffect->RequestRestyle(EffectCompositor::RestyleType::Standard);
}
private:
const ComputedStyle* mComputedStyle;
};
static void UpdateOldAnimationPropertiesWithNew(
CSSAnimation& aOld, TimingParams&& aNewTiming,
nsTArray<Keyframe>&& aNewKeyframes, bool aNewIsStylePaused,
ServoCSSAnimationBuilder& aBuilder) {
bool animationChanged = false;
// Update the old from the new so we can keep the original object
// identity (and any expando properties attached to it).
if (aOld.GetEffect()) {
dom::AnimationEffect* oldEffect = aOld.GetEffect();
animationChanged = oldEffect->SpecifiedTiming() != aNewTiming;
oldEffect->SetSpecifiedTiming(std::move(aNewTiming));
KeyframeEffect* oldKeyframeEffect = oldEffect->AsKeyframeEffect();
if (oldKeyframeEffect) {
aBuilder.SetKeyframes(*oldKeyframeEffect, std::move(aNewKeyframes));
}
}
// Handle changes in play state. If the animation is idle, however,
// changes to animation-play-state should *not* restart it.
if (aOld.PlayState() != AnimationPlayState::Idle) {
// CSSAnimation takes care of override behavior so that,
// for example, if the author has called pause(), that will
// override the animation-play-state.
// (We should check aNew->IsStylePaused() but that requires
// downcasting to CSSAnimation and we happen to know that
// aNew will only ever be paused by calling PauseFromStyle
// making IsPausedOrPausing synonymous in this case.)
if (!aOld.IsStylePaused() && aNewIsStylePaused) {
aOld.PauseFromStyle();
animationChanged = true;
} else if (aOld.IsStylePaused() && !aNewIsStylePaused) {
aOld.PlayFromStyle();
animationChanged = true;
}
}
// Updating the effect timing above might already have caused the
// animation to become irrelevant so only add a changed record if
// the animation is still relevant.
if (animationChanged && aOld.IsRelevant()) {
MutationObservers::NotifyAnimationChanged(&aOld);
}
}
// Returns a new animation set up with given StyleAnimation.
// Or returns an existing animation matching StyleAnimation's name updated
// with the new StyleAnimation.
static already_AddRefed<CSSAnimation> BuildAnimation(
nsPresContext* aPresContext, const NonOwningAnimationTarget& aTarget,
const nsStyleDisplay& aStyleDisplay, uint32_t animIdx,
ServoCSSAnimationBuilder& aBuilder,
nsAnimationManager::CSSAnimationCollection* aCollection) {
MOZ_ASSERT(aPresContext);
nsAtom* animationName = aStyleDisplay.GetAnimationName(animIdx);
nsTArray<Keyframe> keyframes;
if (!aBuilder.BuildKeyframes(
*aTarget.mElement, aPresContext, animationName,
aStyleDisplay.GetAnimationTimingFunction(animIdx), keyframes)) {
return nullptr;
}
TimingParams timing = TimingParamsFromCSSParams(
aStyleDisplay.GetAnimationDuration(animIdx),
aStyleDisplay.GetAnimationDelay(animIdx),
aStyleDisplay.GetAnimationIterationCount(animIdx),
aStyleDisplay.GetAnimationDirection(animIdx),
aStyleDisplay.GetAnimationFillMode(animIdx));
bool isStylePaused = aStyleDisplay.GetAnimationPlayState(animIdx) ==
StyleAnimationPlayState::Paused;
// Find the matching animation with animation name in the old list
// of animations and remove the matched animation from the list.
RefPtr<CSSAnimation> oldAnim =
PopExistingAnimation(animationName, aCollection);
if (oldAnim) {
// Copy over the start times and (if still paused) pause starts
// for each animation (matching on name only) that was also in the
// old list of animations.
// This means that we honor dynamic changes, which isn't what the
// spec says to do, but WebKit seems to honor at least some of
// them. See
// http://lists.w3.org/Archives/Public/www-style/2011Apr/0079.html
// In order to honor what the spec said, we'd copy more data over.
UpdateOldAnimationPropertiesWithNew(*oldAnim, std::move(timing),
std::move(keyframes), isStylePaused,
aBuilder);
return oldAnim.forget();
}
KeyframeEffectParams effectOptions;
RefPtr<KeyframeEffect> effect = new KeyframeEffect(
aPresContext->Document(),
OwningAnimationTarget(aTarget.mElement, aTarget.mPseudoType),
std::move(timing), effectOptions);
aBuilder.SetKeyframes(*effect, std::move(keyframes));
RefPtr<CSSAnimation> animation = new CSSAnimation(
aPresContext->Document()->GetScopeObject(), animationName);
animation->SetOwningElement(
OwningElementRef(*aTarget.mElement, aTarget.mPseudoType));
animation->SetTimelineNoUpdate(aTarget.mElement->OwnerDoc()->Timeline());
animation->SetEffectNoUpdate(effect);
if (isStylePaused) {
animation->PauseFromStyle();
} else {
animation->PlayFromStyle();
}
Bug 1405548 - Post restyles when creating or removing new CSS animations when using the Servo backend; r=hiro As explained in the extended comment in this patch, for Servo we want to post restyles when creating new animations so that we run a second animation restyle and incorporate the result of new animations into style immediately. (Gecko does everything in the one restyle, and although this causes other bugs related to triggering transitions, at least it means it does not require restyles to be posted here). It turns out that we normally end up posting a restyle anyway in CSSAnimation::SetAnimationIndex. Bug 1332958 was supposed to drop that but it never landed. However, CSSAnimation::SetAnimationIndex only posts a restyle when there is a change to the animation index. It turns out that, by chance, there normally *is* a change to a CSSAnimation's animation index when it is created. Initially it takes its animation index from Animation::sNextAnimationIndex which is incremented each time it is assigned to an animation. If the first Animation we create for a given content process is a CSSAnimation then sNextAnimationIndex will be zero and so we will initially assign an animation index of zero. If that CSS animation is also the first in the list of animations in animation-name, when we call SetAnimationIndex we will pass zero as the index to use, and when we go to update the animation index we will detect that there is no change, and will NOT post an animation restyle. As a result the target element's style will NOT reflect the animated style. To fix this we need to ensure that *new* CSS animations trigger a restyle. For *changes* to animations, the corresponding calls to SetKeyframes and SetSpecifiedTiming post restyles so the behavior should be correct in those cases. For *removed* animations I observed that in at least some cases we successfully post a restyle. However, this appeared to be as much by chance as anything so this patch also posts a restyle for removed animations. (Note that the EffectCompositor will ignore redundant restyle requests so this is ok.) This patch deliberately does not expose Animation::PostUpdate and call that because the code introduced here is intended to be temporary. Long-term we should remove the Gecko style backend and allow the calls to PlayFromStyle, PauseFromStyle, CancelFromStyle etc. to post restyles just like calls to Play, Pause, and Cancel do. At that point this code can also be removed. MozReview-Commit-ID: 4c3vJdLBqeY --HG-- extra : rebase_source : 684cb483562709161b2d5635e173e55319509a70
2017-10-11 10:36:36 +03:00
aBuilder.NotifyNewOrRemovedAnimation(*animation);
return animation.forget();
}
static nsAnimationManager::OwningCSSAnimationPtrArray BuildAnimations(
nsPresContext* aPresContext, const NonOwningAnimationTarget& aTarget,
const nsStyleDisplay& aStyleDisplay, ServoCSSAnimationBuilder& aBuilder,
nsAnimationManager::CSSAnimationCollection* aCollection,
nsTHashtable<nsRefPtrHashKey<nsAtom>>& aReferencedAnimations) {
nsAnimationManager::OwningCSSAnimationPtrArray result;
for (size_t animIdx = aStyleDisplay.mAnimationNameCount; animIdx-- != 0;) {
nsAtom* name = aStyleDisplay.GetAnimationName(animIdx);
// CSS Animations whose animation-name does not match a @keyframes rule do
// not generate animation events. This includes when the animation-name is
// "none" which is represented by an empty name in the StyleAnimation.
// Since such animations neither affect style nor dispatch events, we do
// not generate a corresponding CSSAnimation for them.
if (name == nsGkAtoms::_empty) {
continue;
}
aReferencedAnimations.PutEntry(name);
RefPtr<CSSAnimation> dest = BuildAnimation(
aPresContext, aTarget, aStyleDisplay, animIdx, aBuilder, aCollection);
if (!dest) {
continue;
}
dest->SetAnimationIndex(static_cast<uint64_t>(animIdx));
result.AppendElement(dest);
}
return result;
}
void nsAnimationManager::UpdateAnimations(dom::Element* aElement,
PseudoStyleType aPseudoType,
const ComputedStyle* aComputedStyle) {
MOZ_ASSERT(mPresContext->IsDynamic(),
"Should not update animations for print or print preview");
MOZ_ASSERT(aElement->IsInComposedDoc(),
"Should not update animations that are not attached to the "
"document tree");
const nsStyleDisplay* disp =
aComputedStyle ? aComputedStyle->StyleDisplay() : nullptr;
if (!disp || disp->mDisplay == StyleDisplay::None) {
// If we are in a display:none subtree we will have no computed values.
// However, if we are on the root of display:none subtree, the computed
// values might not have been cleared yet.
// In either case, since CSS animations should not run in display:none
// subtrees we should stop (actually, destroy) any animations on this
// element here.
StopAnimationsForElement(aElement, aPseudoType);
return;
}
NonOwningAnimationTarget target(aElement, aPseudoType);
ServoCSSAnimationBuilder builder(aComputedStyle);
DoUpdateAnimations(target, *disp, builder);
}
void nsAnimationManager::DoUpdateAnimations(
const NonOwningAnimationTarget& aTarget,
const nsStyleDisplay& aStyleDisplay, ServoCSSAnimationBuilder& aBuilder) {
// Everything that causes our animation data to change triggers a
// style change, which in turn triggers a non-animation restyle.
// Likewise, when we initially construct frames, we're not in a
// style change, but also not in an animation restyle.
CSSAnimationCollection* collection =
CSSAnimationCollection::GetAnimationCollection(aTarget.mElement,
aTarget.mPseudoType);
if (!collection && aStyleDisplay.mAnimationNameCount == 1 &&
aStyleDisplay.mAnimations[0].GetName() == nsGkAtoms::_empty) {
return;
}
nsAutoAnimationMutationBatch mb(aTarget.mElement->OwnerDoc());
// Build the updated animations list, extracting matching animations from
// the existing collection as we go.
OwningCSSAnimationPtrArray newAnimations =
BuildAnimations(mPresContext, aTarget, aStyleDisplay, aBuilder,
collection, mMaybeReferencedAnimations);
if (newAnimations.IsEmpty()) {
if (collection) {
collection->Destroy();
}
return;
}
if (!collection) {
bool createdCollection = false;
collection = CSSAnimationCollection::GetOrCreateAnimationCollection(
aTarget.mElement, aTarget.mPseudoType, &createdCollection);
if (!collection) {
MOZ_ASSERT(!createdCollection, "outparam should agree with return value");
NS_WARNING("allocating collection failed");
return;
}
if (createdCollection) {
AddElementCollection(collection);
}
}
collection->mAnimations.SwapElements(newAnimations);
// Cancel removed animations
for (size_t newAnimIdx = newAnimations.Length(); newAnimIdx-- != 0;) {
Bug 1405548 - Post restyles when creating or removing new CSS animations when using the Servo backend; r=hiro As explained in the extended comment in this patch, for Servo we want to post restyles when creating new animations so that we run a second animation restyle and incorporate the result of new animations into style immediately. (Gecko does everything in the one restyle, and although this causes other bugs related to triggering transitions, at least it means it does not require restyles to be posted here). It turns out that we normally end up posting a restyle anyway in CSSAnimation::SetAnimationIndex. Bug 1332958 was supposed to drop that but it never landed. However, CSSAnimation::SetAnimationIndex only posts a restyle when there is a change to the animation index. It turns out that, by chance, there normally *is* a change to a CSSAnimation's animation index when it is created. Initially it takes its animation index from Animation::sNextAnimationIndex which is incremented each time it is assigned to an animation. If the first Animation we create for a given content process is a CSSAnimation then sNextAnimationIndex will be zero and so we will initially assign an animation index of zero. If that CSS animation is also the first in the list of animations in animation-name, when we call SetAnimationIndex we will pass zero as the index to use, and when we go to update the animation index we will detect that there is no change, and will NOT post an animation restyle. As a result the target element's style will NOT reflect the animated style. To fix this we need to ensure that *new* CSS animations trigger a restyle. For *changes* to animations, the corresponding calls to SetKeyframes and SetSpecifiedTiming post restyles so the behavior should be correct in those cases. For *removed* animations I observed that in at least some cases we successfully post a restyle. However, this appeared to be as much by chance as anything so this patch also posts a restyle for removed animations. (Note that the EffectCompositor will ignore redundant restyle requests so this is ok.) This patch deliberately does not expose Animation::PostUpdate and call that because the code introduced here is intended to be temporary. Long-term we should remove the Gecko style backend and allow the calls to PlayFromStyle, PauseFromStyle, CancelFromStyle etc. to post restyles just like calls to Play, Pause, and Cancel do. At that point this code can also be removed. MozReview-Commit-ID: 4c3vJdLBqeY --HG-- extra : rebase_source : 684cb483562709161b2d5635e173e55319509a70
2017-10-11 10:36:36 +03:00
aBuilder.NotifyNewOrRemovedAnimation(*newAnimations[newAnimIdx]);
newAnimations[newAnimIdx]->CancelFromStyle(PostRestyleMode::IfNeeded);
}
}