Reland: CC: Adjust tiling creation triggers during pinch-zoom.

Reland: CC: Adjust tiling creation triggers during pinch-zoom.

We used to create tilings only when exceeding a certain scale
difference from the ideal scale. This meant that in the worst
case our only available high-res tiling was 2X the required
resolution.

This patch adjusts the creation of new tilings, to ensure we
always have a tiling option that is less-than-or-equal-to
the current ideal scale.

BUG=307206
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=237179

[epenner, 2013-11-21 23:50:53]

Ptal. Same as before but I added an extra test for the crash case. The fix for
the crash is already landed, but I verified it crashes without that fix.

[epenner, 2013-11-21 23:53:53]

https://codereview.chromium.org/81453002/diff/30001/cc/layers/picture_layer_i...
File cc/layers/picture_layer_impl.cc (right):

https://codereview.chromium.org/81453002/diff/30001/cc/layers/picture_layer_i...
cc/layers/picture_layer_impl.cc:930: float desired_contents_scale =
Actually, I think this might be quite good enough. Pinch-zoom itself will never
cause > 2X scale factors very quickly, but when a layer enters pinch-zoom, it
could be off by a drastic factor (due to our logic in other cases not always
adjusting the scale).

So I'm thinking this should be a loop that converges on the right scale, or
perhaps just use the ideal scale if it's too different.

[enne (OOO), 2013-11-21 23:58:51]

Thanks for uploading this as two patches.  :)

https://codereview.chromium.org/81453002/diff/30001/cc/layers/picture_layer_i...
File cc/layers/picture_layer_impl.cc (right):

https://codereview.chromium.org/81453002/diff/30001/cc/layers/picture_layer_i...
cc/layers/picture_layer_impl.cc:860: else if (!low_res && previous_low_res)
Could you also check that previous_low_res != high_res too?

It seems possible that previous_low_res == high_res and low_res is NULL because
this layer doesn't support low_res at that scale.

(Alternatively, maybe we should just set high_res's resolution after all of
this, and then it would be less fragile.)

[enne (OOO), 2013-11-22 00:01:02]

https://codereview.chromium.org/81453002/diff/30001/cc/layers/picture_layer_i...
File cc/layers/picture_layer_impl.cc (right):

https://codereview.chromium.org/81453002/diff/30001/cc/layers/picture_layer_i...
cc/layers/picture_layer_impl.cc:930: float desired_contents_scale =
On 2013/11/21 23:53:53, epenner wrote:
> Actually, I think this might be quite good enough. Pinch-zoom itself will
never
> cause > 2X scale factors very quickly, but when a layer enters pinch-zoom, it
> could be off by a drastic factor (due to our logic in other cases not always
> adjusting the scale).
> 
> So I'm thinking this should be a loop that converges on the right scale, or
> perhaps just use the ideal scale if it's too different.

Yeah.  If you want to make this more robust in a follow-up patch, I'd be all for
it.  I think this might be good enough most of the time, but it does seem like
there are edge cases that would not be handled well here.

[epenner, 2013-11-22 00:07:38]

https://codereview.chromium.org/81453002/diff/30001/cc/layers/picture_layer_i...
File cc/layers/picture_layer_impl.cc (right):

https://codereview.chromium.org/81453002/diff/30001/cc/layers/picture_layer_i...
cc/layers/picture_layer_impl.cc:860: else if (!low_res && previous_low_res)
Sorry, this isn't rebased yet but I handled that (that was the bug). 

The new code is:

if (!low_res)
  low_res = previous_low_res;
if (low_res && low_res != high_res)
  low_res->set_resolution(LOW_RESOLUTION);

[epennerAtGoogle, 2013-11-22 01:08:28]

I rebased to make it more clear. Still no changes here except
- The extra test
- Set high-res after low (to be extra-double-sure we have one)

[epenner, 2013-11-25 20:24:39]

Rebased. Any thing else needed for this one?

[enne (OOO), 2013-11-25 20:28:20]

lgtm

[commit-bot: I haz the power, 2013-11-25 20:29:56]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/epenner@chromium.org/81453002/190001

[commit-bot: I haz the power, 2013-11-25 20:52:52]

Failed to get patchset properties (patchset not found?)

[commit-bot: I haz the power, 2013-11-25 20:55:51]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/epenner@chromium.org/81453002/210001

[commit-bot: I haz the power, 2013-11-25 23:35:54]

Change committed as 237179