Remove flags for nonFastScrollable and touchEventHandler regions getting changed

Remove flags for nonFastScrollable and touchEventHandler regions getting changed

Currently, two flags are controlling whether we push these regions to layerImpl from
layer and with implSidePainting, we end up not pushing the regions at all since the
flags are flipped back after pusing them once. Removing both flags to make sure we
get the regions carried over.

BUG=168929
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=176024

[jamesr, 2013-01-09 23:02:19]

This means doing a region->region copy every frame, which can be expensive if
the region is expensive.  Why can't we track if it's changed properly?

[enne (OOO), 2013-01-09 23:04:39]

On 2013/01/09 23:02:19, jamesr wrote:
> This means doing a region->region copy every frame, which can be expensive if
> the region is expensive.  Why can't we track if it's changed properly?

Because we're syncing to the pending tree, which doesn't have that information.

[jamesr1, 2013-01-09 23:29:49]

On Wed, Jan 9, 2013 at 3:04 PM, <enne@chromium.org> wrote:

> On 2013/01/09 23:02:19, jamesr wrote:
>
>> This means doing a region->region copy every frame, which can be
>> expensive if
>> the region is expensive.  Why can't we track if it's changed properly?
>>
>
> Because we're syncing to the pending tree, which doesn't have that
> information.
>

This tracking is only tracking main-thread state - why would it need to
care about pending / active trees?

>
>
https://codereview.chromium.**org/11829041/<https://codereview.chromium.org/1...
>

[enne (OOO), 2013-01-09 23:36:35]

If the pending tree is created fresh, this Layer's corresponding LayerImpl does
not have any region set on it.  If this is not pushed every time from the main
thread, then it will not have it.

[Yusuf, 2013-01-09 23:47:26]

Would it be feasible to have a different pushPropertiesTo, for a pending tree
that is getting created fresh, that pushes all properties and keep using these
flags for all other situations?



On 2013/01/09 23:36:35, enne wrote:
> If the pending tree is created fresh, this Layer's corresponding LayerImpl
does
> not have any region set on it.  If this is not pushed every time from the main
> thread, then it will not have it.

[enne (OOO), 2013-01-09 23:49:05]

Sorry, the pending tree is always created fresh at the moment.  I should have
been more clear.

I talked with jamesr, who sounds like he's going to add per-layer dirty bits
rather than these per-property ones.  In the meantime, let's land this patch and
fix the touch bug.

lgtm

[danakj1, 2013-01-09 23:50:07]

im pretty sure there was a bug to add a "ever got pushed to" impl flag on
layer impl that would fix this (and clean up something in tree synchonizer
iirc)

On Wednesday, January 9, 2013, wrote:

> Would it be feasible to have a different pushPropertiesTo, for a pending
> tree
> that is getting created fresh, that pushes all properties and keep using
> these
> flags for all other situations?
>
>
>
> On 2013/01/09 23:36:35, enne wrote:
>
>> If the pending tree is created fresh, this Layer's corresponding LayerImpl
>>
> does
>
>> not have any region set on it.  If this is not pushed every time from the
>> main
>> thread, then it will not have it.
>>
>
>
>
>
https://codereview.chromium.**org/11829041/<https://codereview.chromium.org/1...
>
> --
>
>
>

[danakj1, 2013-01-09 23:52:39]

also impl flag would avoid regressing existing paths

On Wednesday, January 9, 2013, Dana Jansens wrote:

> im pretty sure there was a bug to add a "ever got pushed to" impl flag on
> layer impl that would fix this (and clean up something in tree synchonizer
> iirc)
>
> On Wednesday, January 9, 2013, wrote:
>
>> Would it be feasible to have a different pushPropertiesTo, for a pending
>> tree
>> that is getting created fresh, that pushes all properties and keep using
>> these
>> flags for all other situations?
>>
>>
>>
>> On 2013/01/09 23:36:35, enne wrote:
>>
>>> If the pending tree is created fresh, this Layer's corresponding
>>> LayerImpl
>>>
>> does
>>
>>> not have any region set on it.  If this is not pushed every time from
>>> the main
>>> thread, then it will not have it.
>>>
>>
>>
>>
>>
https://codereview.chromium.**org/11829041/<https://codereview.chromium.org/1...
>>
>> --
>>
>>
>>

[jamesr1, 2013-01-10 00:02:20]

Yeah, I'm workin' on that now.

On Wed, Jan 9, 2013 at 3:50 PM, Dana Jansens <danakj@google.com> wrote:

> im pretty sure there was a bug to add a "ever got pushed to" impl flag on
> layer impl that would fix this (and clean up something in tree synchonizer
> iirc)
>
> On Wednesday, January 9, 2013, wrote:
>
>> Would it be feasible to have a different pushPropertiesTo, for a pending
>> tree
>> that is getting created fresh, that pushes all properties and keep using
>> these
>> flags for all other situations?
>>
>>
>>
>> On 2013/01/09 23:36:35, enne wrote:
>>
>>> If the pending tree is created fresh, this Layer's corresponding
>>> LayerImpl
>>>
>> does
>>
>>> not have any region set on it.  If this is not pushed every time from
>>> the main
>>> thread, then it will not have it.
>>>
>>
>>
>>
>>
https://codereview.chromium.**org/11829041/<https://codereview.chromium.org/1...
>>
>> --
>>
>>
>>

[commit-bot: I haz the power, 2013-01-10 00:20:57]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/yusufo@chromium.org/11829041/1

[commit-bot: I haz the power, 2013-01-10 04:45:54]

Change committed as 176024