In the next patch in this series we would like to use this functionality in
nsSMILController as well so this patch moves it to somewhere we can share it.
MozReview-Commit-ID: 1IzWoCCw4aD
It's not great that we traverse up the tree looking for the subtree root for
each animating element that is in mElementsToRestyle. An alternative could
be to traverse the entire subtree and build a hash set of the elements
within it, so that each test in the mElementsToRestyle loop can be done
quickly. I suspect that most of the time looking up the tree for
each animation will be quicker, but it does have worse worst case behavior.
MozReview-Commit-ID: FWoQ7fD9YZC
Before this patch, we store each computed values in a hashtable,
nsRefPtrHashtable<nsUint32HashKey, RawServoAnimationValue>, for all
KeyframeEffectReadOnly on an element, and convert the ServoAnimationValues of
the hashtable into an nsTArray<ServoAnimationValue*> and then convert
the ServoAnimationValues of the nsTArray into PropertyDeclarationBlock
in rust. This way was really inefficient.
In this patch, we store the computed values into AnimationValueMap and
convert all AnimationValue in the map into PropertyDeclarationBlock
after EffectCompositor::GetAnimationRule.
MozReview-Commit-ID: EJ2Kl65fVeF
Later in this patch series we will replace ServoAnimationRule with a hashmap.
At that point, we would like to pass the hashmap to ComposeStyle. In order
to achieve that, this patch templatizes the 'animation rule' parameter of
ComposeStyle in both Animation and KeyframeEffectReadOnly so that it can
represent a hashmap instead.
MozReview-Commit-ID: H5bMxjN3W8n
The only one caller of GetServoAnimationRule is Gecko_GetAnimationRule.
There is no risk that pres context is destroyed in Gecko_GetAnimationRule.
MozReview-Commit-ID: 52pxbCiCNoJ
We always call TElement::get_animation_rules() for both cascade levels while
there are animations or transitions, so we want to handle the following cases:
1. Have both CascadeLevel::Animations and CascadeLevel::Transitions:
* We use EffectSet::mPropertiesForAnimationsLevel to filter out
CascadeLevel::Transitions because transitions are suppressed when both
are presented.
2. Have CascadeLevel::Animations, but don't have CascadeLevel::Transitions:
* We also use EffectSet::mPropertiesForAnimationsLevel to filter out
the unwanted CascadeLevel::Transitions.
3. Don't Have CascadeLevel::Animations, but have CascadeLevel::Transitions:
* EffectSet::mPropertiesForAnimationsLevel doesn't work for this case.
In Gecko, mElementsToRestyle can help us to filter out the unwanted
CascadeLevel::Animations, However, mElementsToRestyle is cleared in
Pre-Traversal, so now we rely on the cascade ordering of transitions to
override animations. I think we still need to optimize this eventually.
4. No animations:
* EffectSet helps us to check if there is any animations/transitions.
Therefore, we need to call MaybeUpdateCascadeResults(), which updates
mPropertiesForAnimationsLevel, in Pre-Traversal.
MozReview-Commit-ID: IHYw56EX7Ta
The restyle request during restyling is a result of creating/updating/removing
CSS animations that will come from a SequentialTask which will be implemented
in a subsequent patch.
MozReview-Commit-ID: JoAqvcN3y51
We don't need to update the time during composing style, we just need to
update when transform animations are sent to the compositor. The most
recent refresh time of nsRefreshDriver should be the same in either side.
If the time is skewed, that means we skip composing style, if it happened
that's another bug.
MozReview-Commit-ID: Dxtuocf1ml1
We no longer need separate time for cascade levels respectively since we do
send all type of animations on the same element to the compositor regardless
of cascade level.
MozReview-Commit-ID: 7cGQGE3yHvm
Make FindAnimationsForCompositor return false because we don't support
OMTA, and use warnings, instead of assertions, because animations sill
work.
We call RequestRestyle(Layer) not only for OMTA but also for
non-compositor-driven animations, so just skip the animation generation part,
and use warnings, instead of assertions.
MozReview-Commit-ID: LNw5aC8jOvh
1. Add one new FFI, Gecko_GetAnimationRule, which will try
to update the animation rule and retrieve a ServoAnimationRule
from EffectSet.
2. Add GetServoAnimationRule, which calls Animation::ComposeStyle and updates
mElementsToRestyle.
3. There is no eRestyle_{CSSAnimations|CSSTransitions}, so we use
eRestyle_Self and eRestyle_Subtree for stylo.
MozReview-Commit-ID: 9RpJurPSFMk
1. Introduce ServoAnimationRule, which is an equivalent of AnimValuesStyleRule.
2. Wrap ServoAnimationRule and AnimValuesStyleRule into a struct, and
use it as the parameter of Animation::ComposeStyle and
KeyframeEffectReadOnly::ComposeStyle.
3. Uncompute the RawServoAnimationValues in ServoAnimationRule when we
need it.
MozReview-Commit-ID: HahXDYBCAhH
Unlike the other one this function does not need an old nsStyleContext.
This function just resolves style without all animation data in the first place.
The existing ResolveStyleWithoutAnimation is renamed to
ResolveStyleByRemovingAnimation.
MozReview-Commit-ID: 1RmCQNw0MmW
Currently this function is used when an nsIFrame is generated by reframing. In
reframing, we call EffectCompositor::UpdateEffectProperties, as a result, we
have to clear the base style that we resolved once. So, for such cases, we need
to re-resolve the base style with the style context that is associated with the
nsIFrame.
MozReview-Commit-ID: F90OuF44SPI
This should be easier to read and provide us a convenient place to check for
other cases where we need to synchronize with the main thread (such as the
change introduced in this bug where we synchronize with other animations
started at the same time).
MozReview-Commit-ID: 8iuA7P4ycwM
This patch adds a hashtable to store the non-animated base style of each
property in EffectSet class if the following conditions are met.
1) the effect is the lowest priority of the effect
2) the effect is additive or accumulative
The styles are stored as StyleAnimationValue objects since it's handy class to
store a CSS value for different properties.
MozReview-Commit-ID: 1MZV7MnqzfI
If all of animations on an element are paused, finished or zero playback rate,
we don't send those animations to the compositor.
Also in this change, we send zero active duration animations to the compositor
in the same way as normail animations.
MozReview-Commit-ID: CHjv6Buy5fa
When we are composing style for the Animations level of the cascade,
if we have transitions-level animations, we *also* need to compose them if we
have animations-level animations that build on top of them using additive or
accumulative composite modes.
However, we should not build those transitions-level animations unless they will
be built on or overridden by a regular animations-level animation. Otherwise
we will end up with transitions-level animations in the animations-level and
while transitions-level will override the animations-level in the cascade,
once the transition finishes there will be nothing to remove the cached
animations-level animation rule.
MozReview-Commit-ID: LaRabzDSsO5
We needed this polyfill for <initializer_list> when some of our C++
standard libraries did not support said header. They all do now, so the
polyfill is redundant.
To send animations to compositor in the delay phase we need to
modify Animation::IsPlaying returning true in the delay phase.
Note about background-position-in-delay.html:
After this patch, background-position animation also creates an active layer
from its delay phase.
Also note about test cases in test_animations_omta.html:
After landing bug 1279071, getOMTAStyle() returns the style value only
specified by animations, also in this patch we don't apply any opacity or
transform values in the delay phase, as a result we can't tell animating
value during delay phase on the compositor.
MozReview-Commit-ID: ILYKig3c08d
This patch introduces a new functions named HasEffectiveAnimationOfProperty.
This function checks that a given CSS property is overridden by !important
rules.
On the other hand, now KeyframeEffetReadOnly::HasAnimationOfProperty() does
just check that the effect has a given CSS property. This is used to create
a stacking context because we should create a stacking context for opacity or
transform animations even if the property is overridden by !important rules.
Note about no-stacking-context-(opacity|transform)-removing-animation-in-delay.html
Before this patch we don't create any stacking context for animations overridden
by !important rules, but after this patch we do create a stacking context for
such animations. As a result, in the test case we did paint a stacking context
in the first rAF callback and then in the second rAF callback we did clear the
painted stacking context. Unfortunately sometimes the second rAF callback was
called prior to clear the stacking context on the compositor because of
compositor delay. To avoid this situation, we have to wait for MozAfterPaint
instead of rAF callback.
MozReview-Commit-ID: AG1Y0IgoB3U
Also we skip to update the cascade and compose style for orphaned elements if
the animation reqeusted restyle but the target element has been removed from
document before a subsequent tick.
MozReview-Commit-ID: 3I6uyvuSPO5
This patch also makes composite order lowest to highest, as a result we also
need to replace mWinsInCascade checks with the the properties.
The mWinsInCascade membed itself will be removed in a subsequent patch.
Now we call RequestRestyle(Layer) respectively for transition and animation,
so a test case in test_restyles.html works as expected.
And now lower-priority animations are also sent to the compositor so this patch
fixed some tests in test_running_on_compositor.html and
test_animation_performance_warning.html
MozReview-Commit-ID: BchUsJbmatg
