Devtools Animations: Support multiple frames in the animation timeline

Devtools Animations: Support multiple frames in the animation timeline

Currently, the animation timeline only works for animations attached to
the main document frame. This change ensures the inspector agent is properly
attached to all frames within the document. All start times sent to the
front-end are normalised to the monotonic time. As the playback rate or
current time of an animation timeline are manipulated, these monotonic
time values are invalidated. Since these methods are controlled from the
front-end, the normalised start times are updated. This has no visible UI
change.

BUG=466827

[samli, 2015-03-31 06:56:07]

samli@chromium.org changed reviewers:
+ dgozman@chromium.org

[samli, 2015-03-31 06:56:08]

WIP.

Incomplete tasks:
- Check properly attached to all new incoming frames (and set playback rate on
them)
- on didCommitLoadForLocalFrame(), clear maps appropriately - this is causing an
issue at the moment
- Update start times when IAA::setCurrentTime() is called (just need to the
callback)

[dgozman, 2015-03-31 10:39:25]

dgozman@chromium.org changed reviewers:
+ pfeldman@chromium.org

[samli, 2015-03-31 11:02:23]

On 2015/03/31 at 06:56:08, samli wrote:
> WIP.
> 
> Incomplete tasks:
> - Check properly attached to all new incoming frames (and set playback rate on
them)
> - on didCommitLoadForLocalFrame(), clear maps appropriately - this is causing
an issue at the moment
> - Update start times when IAA::setCurrentTime() is called (just need to the
callback)

Complete now, PTAL. Lacking a test, haven't thought much about how to test this
since it deals with lots of cross-frame stuff and monotonic time. Haven't put
much thought into the testing yet.

[samli, 2015-04-01 01:14:47]

PTAL, has test now.

[pfeldman, 2015-04-01 13:06:45]

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
File Source/core/inspector/InspectorAnimationAgent.cpp (right):

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:177: return player.startTime()
/ player.timeline()->playbackRate() + player.timeline()->zeroTime() * 1000;
What time is this?

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:267: for (Frame* frame =
m_pageAgent->inspectedFrame(); frame; frame =
frame->tree().traverseNext(m_pageAgent->inspectedFrame())) {
It is strange that this synchronization takes place in the inspector code. It
seems unrelated to inspector, some timelines belonging to different documents
could just stay in sync on their own. You also need to sync the newly added
frames and I don't see this code here.

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/front_e...
File Source/devtools/front_end/elements/AnimationTimeline.js (right):

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/front_e...
Source/devtools/front_end/elements/AnimationTimeline.js:50:
target.animationAgent().setPlaybackRate(this._animationsPlaybackRate,
this._updateStartTimes.bind(this));
You should not be talking to the agents from outside the model. Model should
talk to agent and update its instances internally, you should not need to use
public methods to update model internals (updateStartTime).

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/protoco...
File Source/devtools/protocol.json (right):

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/protoco...
Source/devtools/protocol.json:4879: { "name": "id", "type": "string",
"description": "<code>AnimationPlayer</code>'s id." },
For the ids, we typically introduce a separate type "AnimationPlayerId" in this
case that inherits from string. Then we call parameters using full qualified
names such as animationPlayerId: AnimationPlayerId.

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/protoco...
Source/devtools/protocol.json:4880: { "name": "startTime", "type": "number",
"description": "Updated normalized <code>AnimationPlayer</code>'s start time." }
Is this monotonicallyIncreasingTime or wall?

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/protoco...
Source/devtools/protocol.json:4922: { "name": "normalisedCurrentTime", "type":
"number", "description": "Current time for the page animation timeline" }
normalizedCurrentTime

[samli, 2015-04-02 09:57:59]

samli@chromium.org changed reviewers:
- dgozman@chromium.org

[samli, 2015-04-02 09:58:00]

PTAL

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
File Source/core/inspector/InspectorAnimationAgent.cpp (right):

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:177: return player.startTime()
/ player.timeline()->playbackRate() + player.timeline()->zeroTime() * 1000;
On 2015/04/01 at 13:06:45, pfeldman wrote:
> What time is this?

We take relative timeline times and convert them to monotonic times.

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:267: for (Frame* frame =
m_pageAgent->inspectedFrame(); frame; frame =
frame->tree().traverseNext(m_pageAgent->inspectedFrame())) {
On 2015/04/01 at 13:06:45, pfeldman wrote:
> It is strange that this synchronization takes place in the inspector code. It
seems unrelated to inspector, some timelines belonging to different documents
could just stay in sync on their own. You also need to sync the newly added
frames and I don't see this code here.

This code isn't to sync timelines. This code is used to seek timelines. For
example, when 'replay' is hit or the scrubber is dragged, we want to seek back
in time. Instead of maintaining all timelines separately, the front-end only
understands monotonic time values. It then says: "take me back to time x", which
setCurrentTime() then seeks each animation timeline individually (converting
monotonic time back into timeline time).

New frames are automatically captured in this. New frames have their timeline
playback rate set in the frontend.

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/front_e...
File Source/devtools/front_end/elements/AnimationTimeline.js (right):

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/front_e...
Source/devtools/front_end/elements/AnimationTimeline.js:50:
target.animationAgent().setPlaybackRate(this._animationsPlaybackRate,
this._updateStartTimes.bind(this));
On 2015/04/01 at 13:06:45, pfeldman wrote:
> You should not be talking to the agents from outside the model. Model should
talk to agent and update its instances internally, you should not need to use
public methods to update model internals (updateStartTime).

Done.

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/protoco...
File Source/devtools/protocol.json (right):

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/protoco...
Source/devtools/protocol.json:4879: { "name": "id", "type": "string",
"description": "<code>AnimationPlayer</code>'s id." },
On 2015/04/01 at 13:06:45, pfeldman wrote:
> For the ids, we typically introduce a separate type "AnimationPlayerId" in
this case that inherits from string. Then we call parameters using full
qualified names such as animationPlayerId: AnimationPlayerId.

Done.

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/protoco...
Source/devtools/protocol.json:4880: { "name": "startTime", "type": "number",
"description": "Updated normalized <code>AnimationPlayer</code>'s start time." }
On 2015/04/01 at 13:06:45, pfeldman wrote:
> Is this monotonicallyIncreasingTime or wall?

Monotonically increasing time.

https://codereview.chromium.org/1042143005/diff/40001/Source/devtools/protoco...
Source/devtools/protocol.json:4922: { "name": "normalisedCurrentTime", "type":
"number", "description": "Current time for the page animation timeline" }
On 2015/04/01 at 13:06:45, pfeldman wrote:
> normalizedCurrentTime

Done.

[pfeldman, 2015-04-02 14:42:39]

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
File Source/core/inspector/InspectorAnimationAgent.cpp (right):

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:177: return player.startTime()
/ player.timeline()->playbackRate() + player.timeline()->zeroTime() * 1000;
On 2015/04/02 09:58:00, samli wrote:
> On 2015/04/01 at 13:06:45, pfeldman wrote:
> > What time is this?
> 
> We take relative timeline times and convert them to monotonic times.

But zeroTime is already monotonicTime, why do multiply it by 1000?

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:267: for (Frame* frame =
m_pageAgent->inspectedFrame(); frame; frame =
frame->tree().traverseNext(m_pageAgent->inspectedFrame())) {
On 2015/04/02 09:58:00, samli wrote:
> On 2015/04/01 at 13:06:45, pfeldman wrote:
> > It is strange that this synchronization takes place in the inspector code.
It
> seems unrelated to inspector, some timelines belonging to different documents
> could just stay in sync on their own. You also need to sync the newly added
> frames and I don't see this code here.
> 
> This code isn't to sync timelines. This code is used to seek timelines. For
> example, when 'replay' is hit or the scrubber is dragged, we want to seek back
> in time. Instead of maintaining all timelines separately, the front-end only
> understands monotonic time values. It then says: "take me back to time x",
which
> setCurrentTime() then seeks each animation timeline individually (converting
> monotonic time back into timeline time).
> 
> New frames are automatically captured in this. New frames have their timeline
> playback rate set in the frontend.

Ok, I think I understand.

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:270:
timeline.setCurrentTime((normalisedCurrentTime - timeline.zeroTime() * 1000) *
timeline.playbackRate());
There is something wrong with the timeline units: you can't do
timeline.setCurrentTime(foo - timeline.zeroTime() * 1000). You should stick to
one scale within one timeline.

https://codereview.chromium.org/1042143005/diff/80001/Source/core/inspector/I...
File Source/core/inspector/InspectorAnimationAgent.cpp (right):

https://codereview.chromium.org/1042143005/diff/80001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:217:
.setStartTime(normalizedStartTime(player))
Different time scales is confusing, if you are using different scales,
properties should have corresponding suffixes.

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
File Source/devtools/front_end/elements/AnimationTimeline.js (right):

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
Source/devtools/front_end/elements/AnimationTimeline.js:92: if (!startTimes ||
!startTimes.length)
Why is it nullable?

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
Source/devtools/front_end/elements/AnimationTimeline.js:99: this._startTime =
player.startTime;
I don't understand how multiple players within multiple frames synchronize. Why
is this different from simply several players?

[samli, 2015-04-02 20:13:18]

samli@chromium.org changed reviewers:
+ dstockwell@chromium.org

[samli, 2015-04-02 20:13:19]

+dstockwell for animation related changes

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
File Source/core/inspector/InspectorAnimationAgent.cpp (right):

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:177: return player.startTime()
/ player.timeline()->playbackRate() + player.timeline()->zeroTime() * 1000;
On 2015/04/02 14:42:38, pfeldman wrote:
> On 2015/04/02 09:58:00, samli wrote:
> > On 2015/04/01 at 13:06:45, pfeldman wrote:
> > > What time is this?
> > 
> > We take relative timeline times and convert them to monotonic times.
> 
> But zeroTime is already monotonicTime, why do multiply it by 1000?

Chrome's internal representations are in seconds. Public Web Animations API uses
milliseconds. As such, front end only understands milliseconds.

To clarify: zeroTime() is in seconds, startTime() is milliseconds since it's a
public API.

https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:270:
timeline.setCurrentTime((normalisedCurrentTime - timeline.zeroTime() * 1000) *
timeline.playbackRate());
On 2015/04/02 14:42:38, pfeldman wrote:
> There is something wrong with the timeline units: you can't do
> timeline.setCurrentTime(foo - timeline.zeroTime() * 1000). You should stick to
> one scale within one timeline.

I think this is the milliseconds confusion?

https://codereview.chromium.org/1042143005/diff/80001/Source/core/inspector/I...
File Source/core/inspector/InspectorAnimationAgent.cpp (right):

https://codereview.chromium.org/1042143005/diff/80001/Source/core/inspector/I...
Source/core/inspector/InspectorAnimationAgent.cpp:217:
.setStartTime(normalizedStartTime(player))
On 2015/04/02 14:42:38, pfeldman wrote:
> Different time scales is confusing, if you are using different scales,
> properties should have corresponding suffixes.

Will fix

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
File Source/devtools/front_end/elements/AnimationTimeline.js (right):

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
Source/devtools/front_end/elements/AnimationTimeline.js:92: if (!startTimes ||
!startTimes.length)
On 2015/04/02 14:42:38, pfeldman wrote:
> Why is it nullable?

In the case of protocol error it's called with null

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
Source/devtools/front_end/elements/AnimationTimeline.js:99: this._startTime =
player.startTime;
On 2015/04/02 14:42:38, pfeldman wrote:
> I don't understand how multiple players within multiple frames synchronize.
Why
> is this different from simply several players?

All animation players have times calculated based on their attached animation
timeline which in turn are based on monotonic time. Ie: a player can return a
start time of "2000" which means it started 2000ms after the timeline started.
We can compare players in one Crame and have been doing so for a while. However,
we can't compare players in separate frames (timelines) without factoring in the
timeline start times, which is what we're doing in this patch.

[pfeldman, 2015-04-02 20:25:19]

On 2015/04/02 20:13:19, samli wrote:
> +dstockwell for animation related changes
> 
>
https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
> File Source/core/inspector/InspectorAnimationAgent.cpp (right):
> 
>
https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
> Source/core/inspector/InspectorAnimationAgent.cpp:177: return
player.startTime()
> / player.timeline()->playbackRate() + player.timeline()->zeroTime() * 1000;
> On 2015/04/02 14:42:38, pfeldman wrote:
> > On 2015/04/02 09:58:00, samli wrote:
> > > On 2015/04/01 at 13:06:45, pfeldman wrote:
> > > > What time is this?
> > > 
> > > We take relative timeline times and convert them to monotonic times.
> > 
> > But zeroTime is already monotonicTime, why do multiply it by 1000?
> 
> Chrome's internal representations are in seconds. Public Web Animations API
uses
> milliseconds. As such, front end only understands milliseconds.
> 
> To clarify: zeroTime() is in seconds, startTime() is milliseconds since it's a
> public API.
> 
>
https://codereview.chromium.org/1042143005/diff/40001/Source/core/inspector/I...
> Source/core/inspector/InspectorAnimationAgent.cpp:270:
> timeline.setCurrentTime((normalisedCurrentTime - timeline.zeroTime() * 1000) *
> timeline.playbackRate());
> On 2015/04/02 14:42:38, pfeldman wrote:
> > There is something wrong with the timeline units: you can't do
> > timeline.setCurrentTime(foo - timeline.zeroTime() * 1000). You should stick
to
> > one scale within one timeline.
> 
> I think this is the milliseconds confusion?
> 
>
https://codereview.chromium.org/1042143005/diff/80001/Source/core/inspector/I...
> File Source/core/inspector/InspectorAnimationAgent.cpp (right):
> 
>
https://codereview.chromium.org/1042143005/diff/80001/Source/core/inspector/I...
> Source/core/inspector/InspectorAnimationAgent.cpp:217:
> .setStartTime(normalizedStartTime(player))
> On 2015/04/02 14:42:38, pfeldman wrote:
> > Different time scales is confusing, if you are using different scales,
> > properties should have corresponding suffixes.
> 
> Will fix
> 
>
https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
> File Source/devtools/front_end/elements/AnimationTimeline.js (right):
> 
>
https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
> Source/devtools/front_end/elements/AnimationTimeline.js:92: if (!startTimes ||
> !startTimes.length)
> On 2015/04/02 14:42:38, pfeldman wrote:
> > Why is it nullable?
> 
> In the case of protocol error it's called with null
> 
>
https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/front_e...
> Source/devtools/front_end/elements/AnimationTimeline.js:99: this._startTime =
> player.startTime;
> On 2015/04/02 14:42:38, pfeldman wrote:
> > I don't understand how multiple players within multiple frames synchronize.
> Why
> > is this different from simply several players?
> 
> All animation players have times calculated based on their attached animation
> timeline which in turn are based on monotonic time. Ie: a player can return a
> start time of "2000" which means it started 2000ms after the timeline started.
> We can compare players in one Crame and have been doing so for a while.
However,
> we can't compare players in separate frames (timelines) without factoring in
the
> timeline start times, which is what we're doing in this patch.

Can we make all the time that goes through the protocol monotonic?

[samli, 2015-04-02 21:50:37]

That's what this patch is :)

[pfeldman, 2015-04-03 09:19:41]

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
File Source/devtools/protocol.json (right):

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4834: { "name": "startTime", "type": "number",
"description": "<code>AnimationPlayer</code>'s start time." },
Is this monotonicallyIncreasing time?

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4835: { "name": "currentTime", "type": "number",
"description": "<code>AnimationPlayer</code>'s current time." },
Is this monotonicallyIncreasing time?

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4885: { "name": "startTime", "type": "number",
"description": "Updated normalized <code>AnimationPlayer</code>'s start time." }
Is this monotonicallyIncreasing time?

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4927: { "name": "normalizedCurrentTime", "type":
"number", "description": "Current time for the page animation timeline" }
Is this monotonicallyIncreasing time?

[samli, 2015-04-03 09:56:01]

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
File Source/devtools/protocol.json (right):

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4834: { "name": "startTime", "type": "number",
"description": "<code>AnimationPlayer</code>'s start time." },
On 2015/04/03 09:19:41, pfeldman wrote:
> Is this monotonicallyIncreasing time?

Yes.

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4835: { "name": "currentTime", "type": "number",
"description": "<code>AnimationPlayer</code>'s current time." },
On 2015/04/03 09:19:41, pfeldman wrote:
> Is this monotonicallyIncreasing time?

This is always relative to the player. It's start time + current time

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4885: { "name": "startTime", "type": "number",
"description": "Updated normalized <code>AnimationPlayer</code>'s start time." }
On 2015/04/03 09:19:41, pfeldman wrote:
> Is this monotonicallyIncreasing time?

Yes.

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4927: { "name": "normalizedCurrentTime", "type":
"number", "description": "Current time for the page animation timeline" }
On 2015/04/03 09:19:41, pfeldman wrote:
> Is this monotonicallyIncreasing time?

Yes

[pfeldman, 2015-04-03 10:36:49]

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
File Source/devtools/protocol.json (right):

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4835: { "name": "currentTime", "type": "number",
"description": "<code>AnimationPlayer</code>'s current time." },
Lets rename it to currentOffset?

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4885: { "name": "startTime", "type": "number",
"description": "Updated normalized <code>AnimationPlayer</code>'s start time." }
On 2015/04/03 09:56:00, samli wrote:
> On 2015/04/03 09:19:41, pfeldman wrote:
> > Is this monotonicallyIncreasing time?
> 
> Yes.

Why is it called normalized then?

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4920: { "name": "animationStartTimes", "type":
"array", "items": { "$ref": "AnimationStartTime" }, "description": "List of
updated animation start times." }
So why do you want to collect all start times upon setPlaybackRate and
setCurrentTimes? When does the start time of the animation player change?

[samli, 2015-04-03 11:31:32]

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
File Source/devtools/protocol.json (right):

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4835: { "name": "currentTime", "type": "number",
"description": "<code>AnimationPlayer</code>'s current time." },
On 2015/04/03 10:36:49, pfeldman wrote:
> Lets rename it to currentOffset?

No, this matches the Web Animations spec.
https://w3c.github.io/web-animations/

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4885: { "name": "startTime", "type": "number",
"description": "Updated normalized <code>AnimationPlayer</code>'s start time." }
On 2015/04/03 10:36:49, pfeldman wrote:
> On 2015/04/03 09:56:00, samli wrote:
> > On 2015/04/03 09:19:41, pfeldman wrote:
> > > Is this monotonicallyIncreasing time?
> > 
> > Yes.
> 
> Why is it called normalized then?

Normalise means to "bring to a standard". We convert relative timeline times to
a standard interface of time. I think it's a reasonable name.

https://codereview.chromium.org/1042143005/diff/80001/Source/devtools/protoco...
Source/devtools/protocol.json:4920: { "name": "animationStartTimes", "type":
"array", "items": { "$ref": "AnimationStartTime" }, "description": "List of
updated animation start times." }
On 2015/04/03 10:36:49, pfeldman wrote:
> So why do you want to collect all start times upon setPlaybackRate and
> setCurrentTimes? When does the start time of the animation player change?

Conceptual start time is always relative to timeline time. When we manipulate
playback rate of a timeline the current time of the timeline does not change.
Old monotonic values are invalidated since they cannot be converted back to
timeline times since the timeline 'scaled' relative to monotonic time.

When seeking a timeline, we are effectively shifting the timeline relative to
monotonic time. This invalidates the start times we sent earlier.

In both cases, the original monotonic start times no longer reference the right
time in the timelines so we need to update the start times.

[dstockwell, 2015-04-07 01:27:12]

let's drop the 'normalized' terminology from the protocol. *all* absolute time
values passed to/from the frontend will be monotonic time in milliseconds

for consistency, perhaps rename as: seekTimelineTo and seekTimelineBy

[samli, 2015-04-07 02:27:31]

PTAL. Naming updated.

The front-end/protocol knows nothing about timelines/monotonic time. It is
agnostic to this and treats all animations as belonging to the same conceptual
timeline.

The synchronization code is essential to making this all function. The other
approach would be to provide the front-end with enough information to be able to
recalculate these start times itself. This unnecessarily opens up the front-end
to knowledge of timelines and keeping them all up to date. This way is robust
(since there's no funny calculations having to be made) and keeps the front-end
code clean.

[pfeldman, 2015-04-07 10:27:34]

https://codereview.chromium.org/1042143005/diff/100001/Source/devtools/protoc...
File Source/devtools/protocol.json (right):

https://codereview.chromium.org/1042143005/diff/100001/Source/devtools/protoc...
Source/devtools/protocol.json:4935: { "name": "animationStartTimes", "type":
"array", "items": { "$ref": "AnimationStartTime" }, "description": "List of
updated animation start times." }
I'd like to better understand when animation times update. Can we book some time
to chat about it offline?

[samli, 2015-04-13 02:11:37]

Abandoning in favor of https://codereview.chromium.org/1081753002