Bug 81740 – [chromium] Transfer wheel fling via WebCompositorInputHandlerClient

• Component: New Bugs
• Status: RESOLVED FIXED

[James Robinson]

[chromium] Transfer wheel fling via WebCompositorInputHandlerClient

[James Robinson]

Created attachment 132970 [details]
Patch

[Nat Duca]

Did the cc input handler changes not make it into this patch?

[Nat Duca]

(In reply to comment #2)
> Did the cc input handler changes not make it into this patch?

Oh, I'm retarded. Just saw the notes on the previous bug.

[Nat Duca]

Created attachment 132986 [details]
Patch

[WebKit Review Bot]

Please wait for approval from [e-mail], [e-mail], [e-mail], [e-mail] or [e-mail] before submitting, as this patch contains changes to the Chromium public API. See also https://trac.webkit.org/wiki/ChromiumWebKitAPI.

[WebKit Review Bot]

Comment on attachment 132986 [details]
Patch

Attachment 132986 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/12070273

New failing tests:
fast/dom/error-to-string-stack-overflow.html

[Adrienne Walker]

(In reply to comment #6)
> (From update of attachment 132986 [details])
> Attachment 132986 [details] did not pass chromium-ews (chromium-xvfb):
> Output: http://queues.webkit.org/results/12070273
> 
> New failing tests:
> fast/dom/error-to-string-stack-overflow.html

That looks like flake.

This all looks fine to me.  jamesr, do you have any feedback on Nat's modification of your original patch? Otherwise, I'm inclined to R+.

[James Robinson]

Comment on attachment 132986 [details]
Patch

I think everything looks fine.  I expect this will have more merge conflicts on landing, though.

[Adrienne Walker]

*** Bug 81479 has been marked as a duplicate of this bug. ***

[James Robinson]

(In reply to comment #9)
> *** Bug 81479 has been marked as a duplicate of this bug. ***

They aren't strictly speaking dupes. https://bugs.webkit.org/show_bug.cgi?id=81479 is a hookup path that works purely in WebKit land, so it's a one-patch fix.  This patch depends on chromium-side changes that aren't fully baked yet.  So if we want something that works quickly then I think https://bugs.webkit.org/show_bug.cgi?id=81479 is the way to go, despite being uglier.

[Nat Duca]

Yeah, I actually will probably go back to the original patch or a slight variation on it. This patch relies on compositor ids being == routing ids, which isn't the case. We could make it that case but that itself is clunky. I think its fine to go through the proxy. Maybe if this becomes more commonplace, we can figure out the right magic to make this less pipey.

[Nat Duca]

Reopening to attach new patch.

[Nat Duca]

Created attachment 133629 [details]
Patch

[Nat Duca]

Comment on attachment 133629 [details]
Patch

Darnit, wrong bug #

[James Robinson]

Reopening to attach new patch.

[James Robinson]

Created attachment 133913 [details]
Patch

[James Robinson]

Depends on http://codereview.chromium.org/9802006/.

I think this hookup is fairly clean. It does depend on https://bugs.webkit.org/show_bug.cgi?id=82154 to get the right clock on the main thread (since the start time passed through is monotonic).

Nat pointed out that it might be useful to have a generic "run this task on the main thread" interface from the WebCompositorInputHandler or similar. I think we should consider adding it as soon as we have a 2nd caller that would use it.

[Adrienne Walker]

Comment on attachment 133913 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=133913&action=review

The WebKit side looks good to me.  A few minor comments:

> Source/WebCore/platform/ActivePlatformGestureAnimation.cpp:73
> +{
> +}

Can you put a trace event here, like the other constructor?

> Source/WebKit/chromium/tests/WebCompositorInputHandlerImplTest.cpp:397
> +    // *) cumulativeScroll depends on the curve, but since we've animated in the -X direction the X value should be < 0

These sign flips are all really hard to follow.  The delta being positive means animating in the negative direction, and the X value being < 0 means the cumulative scroll is > 0.  Is there any way to make these comments or variable names more readable?

[Nat Duca]

(In reply to comment #18)
> (From update of attachment 133913 [details])
I'm fine with this approach in the interests of getting this landed.

[James Robinson]

(In reply to comment #18)
> (From update of attachment 133913 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=133913&action=review
> 
> The WebKit side looks good to me.  A few minor comments:
> 
> > Source/WebCore/platform/ActivePlatformGestureAnimation.cpp:73
> > +{
> > +}
> 
> Can you put a trace event here, like the other constructor?

Done

> 
> > Source/WebKit/chromium/tests/WebCompositorInputHandlerImplTest.cpp:397
> > +    // *) cumulativeScroll depends on the curve, but since we've animated in the -X direction the X value should be < 0
> 
> These sign flips are all really hard to follow.  The delta being positive means animating in the negative direction, and the X value being < 0 means the cumulative scroll is > 0.  Is there any way to make these comments or variable names more readable?

I agree - at least now the WebCompositorInputHandlerImpl code is consistent with what WebViewImpl does. I can't think of much else really brilliant to do, tho.

[James Robinson]

Depends on https://bugs.webkit.org/show_bug.cgi?id=82154 to pass the monotonic clock into m_gestureAnimations->animate in WebViewImpl::updateAnimations(). This patch will still compile, but the transferred animations will have a bad timebase before https://bugs.webkit.org/show_bug.cgi?id=82154 lands.

[James Robinson]

Created attachment 134213 [details]
Patch for landing

[WebKit Review Bot]

Comment on attachment 134213 [details]
Patch for landing

Clearing flags on attachment: 134213

Committed r112364: <http://trac.webkit.org/changeset/112364>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.