Allow perfect scroll bubbling between outer/inner viewports

Allow perfect scroll bubbling between outer/inner viewports

The current compositor scroll bubbling logic is not a respecter of layer
types. That is, when bubbling, it will always try to truncate and/or
project the remaining scroll delta before bubbling to the parent scroll
layer. Such truncation is inappropriate when bubbling between outer and
inner viewport layers, as their movement should appear seamless to the
user. Make an exception for this case, allowing perfect bubbling.

BUG=429326
Committed: https://crrev.com/a7173dd75978963be1f007a7d18bf9bb4dbdb1c5
Cr-Commit-Position: refs/heads/master@{#304898}

[jdduke (slow), 2014-11-18 21:18:20]

jdduke@chromium.org changed reviewers:
+ bokan@chromium.org

[jdduke (slow), 2014-11-18 21:18:20]

bokan@, does this also address crbug.com/420208?

[jdduke (slow), 2014-11-18 21:29:55]

https://codereview.chromium.org/737883002/diff/1/cc/trees/layer_tree_host_imp...
File cc/trees/layer_tree_host_impl.cc (right):

https://codereview.chromium.org/737883002/diff/1/cc/trees/layer_tree_host_imp...
cc/trees/layer_tree_host_impl.cc:2644: did_lock_scrolling_layer_ = true;
Hmm, this is wrong and needs some work.

[jdduke (slow), 2014-11-18 21:34:16]

https://codereview.chromium.org/737883002/diff/1/cc/trees/layer_tree_host_imp...
File cc/trees/layer_tree_host_impl.cc (right):

https://codereview.chromium.org/737883002/diff/1/cc/trees/layer_tree_host_imp...
cc/trees/layer_tree_host_impl.cc:2644: did_lock_scrolling_layer_ = true;
On 2014/11/18 21:29:55, jdduke wrote:
> Hmm, this is wrong and needs some work.

Restored the old logic here, not really related to the diagonal case.

[bokan, 2014-11-19 00:58:34]

lgtm, thanks! (I'll check if this fixes crbug.com/420208 when I get back to the
office tomorrow).

https://codereview.chromium.org/737883002/diff/20001/cc/trees/layer_tree_host...
File cc/trees/layer_tree_host_impl.cc (right):

https://codereview.chromium.org/737883002/diff/20001/cc/trees/layer_tree_host...
cc/trees/layer_tree_host_impl.cc:2658: pending_delta = gfx::Vector2dF();
I know you didn't write this, but do we need to clear the pending_delta if we're
breaking out of the loop? Looks like it isn't referenced outside  the loop.

[bokan, 2014-11-19 14:37:36]

On 2014/11/19 00:58:34, bokan wrote:
> lgtm, thanks! (I'll check if this fixes crbug.com/420208 when I get back to
the
> office tomorrow).
> 
Tried it out, it does indeed address that. Feels so much better!

[jdduke (slow), 2014-11-19 16:20:14]

jdduke@chromium.org changed reviewers:
+ aelias@chromium.org

[jdduke (slow), 2014-11-19 16:20:14]

+aelias for owner review.

https://codereview.chromium.org/737883002/diff/20001/cc/trees/layer_tree_host...
File cc/trees/layer_tree_host_impl.cc (right):

https://codereview.chromium.org/737883002/diff/20001/cc/trees/layer_tree_host...
cc/trees/layer_tree_host_impl.cc:2658: pending_delta = gfx::Vector2dF();
On 2014/11/19 00:58:33, bokan wrote:
> I know you didn't write this, but do we need to clear the pending_delta if
we're
> breaking out of the loop? Looks like it isn't referenced outside  the loop.

I'm guessing there was some trailing code that used it that has since been
deleted? I'd be inclined to remove the zero-ing code, but since the variable is
visible outside the for loop I think it's fine as-is. A shame for loop
statements don't allow different variable type declarations.

[aelias_OOO_until_Jul13, 2014-11-19 19:33:32]

lgtm

[jdduke (slow), 2014-11-19 20:37:43]

The CQ bit was checked by jdduke@chromium.org

[commit-bot: I haz the power, 2014-11-19 20:39:14]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/737883002/20001

[commit-bot: I haz the power, 2014-11-19 22:05:37]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2014-11-19 22:06:14]

Patchset 2 (id:??) landed as
https://crrev.com/a7173dd75978963be1f007a7d18bf9bb4dbdb1c5
Cr-Commit-Position: refs/heads/master@{#304898}