cc: Workaround resourceless draw crash

cc: Workaround resourceless draw crash

It's possible for resourceless draw to raster a layer with no recording,
which crashes in DisplayListRasterSource::RasterCommon. Workaround this
by not appending picture quad if there is no recording for that layer.

Small patch for merging back to release.

BUG=526402
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/d16632eaf7354c9f277e7a7ee4ef666ce186d562
Cr-Commit-Position: refs/heads/master@{#346760}

[boliu, 2015-09-01 19:07:51]

boliu@chromium.org changed reviewers:
+ danakj@chromium.org, enne@chromium.org

[boliu, 2015-09-01 19:07:52]

small workaround for merging back, no tests..

/me off to lunch too :p

[danakj, 2015-09-01 20:16:26]

danakj@chromium.org changed reviewers:
+ vmpstr@chromium.org

[danakj, 2015-09-01 20:16:26]

+vmpstr

this looks sketchy at first glance. but vmpstr will know all.

[enne (OOO), 2015-09-01 20:17:26]

lgtm

[vmpstr, 2015-09-01 20:26:38]

https://codereview.chromium.org/1326713002/diff/1/cc/layers/picture_layer_imp...
File cc/layers/picture_layer_impl.cc (right):

https://codereview.chromium.org/1326713002/diff/1/cc/layers/picture_layer_imp...
cc/layers/picture_layer_impl.cc:185: if (visible_geometry_rect.IsEmpty() ||
!raster_source_->HasRecordings())
Can you leave a comment here explaining why this can be the case? And maybe
early out before AppendDebugBorderQuad (or put a MissingTileBorderColor on it
instead or something)

[enne (OOO), 2015-09-01 20:30:26]

There are other problems here, such as the source may have recordings, but not
actually have the visible rect in its visible viewport.  This is just a
workaround, and I'm going to try to make a better fix.

[vmpstr, 2015-09-01 20:40:54]

On 2015/09/01 20:30:26, enne wrote:
> There are other problems here, such as the source may have recordings, but not
> actually have the visible rect in its visible viewport.  This is just a
> workaround, and I'm going to try to make a better fix.

Ah ok, lgtm as a workaround.

[danakj, 2015-09-01 20:42:30]

https://codereview.chromium.org/1326713002/diff/1/cc/layers/picture_layer_imp...
File cc/layers/picture_layer_impl.cc (right):

https://codereview.chromium.org/1326713002/diff/1/cc/layers/picture_layer_imp...
cc/layers/picture_layer_impl.cc:185: if (visible_geometry_rect.IsEmpty() ||
!raster_source_->HasRecordings())
On 2015/09/01 20:26:38, vmpstr wrote:
> Can you leave a comment here explaining why this can be the case? And maybe
> early out before AppendDebugBorderQuad (or put a MissingTileBorderColor on it
> instead or something)

If it's a workaround, can you leave a TODO pointing to a bug stating as such?

[boliu, 2015-09-01 20:50:41]

Added TODO comment for enne

[boliu, 2015-09-01 20:50:53]

The CQ bit was checked by boliu@chromium.org

[boliu, 2015-09-01 20:50:53]

The patchset sent to the CQ was uploaded after l-g-t-m from vmpstr@chromium.org,
enne@chromium.org
Link to the patchset: https://codereview.chromium.org/1326713002/#ps20001
(title: "add todo")

[commit-bot: I haz the power, 2015-09-01 20:51:40]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1326713002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1326713002/20001

[danakj, 2015-09-01 20:52:39]

Thanks LGTM 2

[commit-bot: I haz the power, 2015-09-01 22:37:06]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-09-01 22:37:48]

Patchset 2 (id:??) landed as
https://crrev.com/d16632eaf7354c9f277e7a7ee4ef666ce186d562
Cr-Commit-Position: refs/heads/master@{#346760}