Adjust the rounding methodology for tiled background phase

Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
incompatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

This patch also removes the redundant imageContainerSize member and accessor,
because it is always the same as the tile size.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
Committed: https://crrev.com/a75de675cbb8f9f594fabc68236a4547975ad7fe
Cr-Commit-Position: refs/heads/master@{#380972}

[Stephen Chennney, 2016-03-10 19:52:05]

I need to add TestExpectations, but first I want to try one other scheme, and
see what tests fail on what bots.

[Stephen Chennney, 2016-03-10 19:52:12]

The CQ bit was checked by schenney@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-03-10 19:53:09]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1780893004/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1780893004/1

[Stephen Chennney, 2016-03-10 20:03:30]

Description was changed from

==========
Adjust the rounding methodology for tiled background phase

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. The solution is
to use the same integer snappaing method for space and phase.
Experimentation suggests that rounding both is the best strategy.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticably better. Other results move the background by a pixel
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any scheme will fail at some zoom level
due to the fact we are rounding tile size, dest rect etc to integers.

R=trchen@chromium.org
BUG=593861
==========

to

==========
Adjust the rounding methodology for tiled background phase

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. The solution is
to use the same integer snappaing method for space and phase.
Experimentation suggests that rounding both is the best strategy.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticably better. Other results move the background by a pixel
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any scheme will fail at some zoom level
due to the fact we are rounding tile size, dest rect etc to integers.

R=trchen@chromium.org
BUG=593861
==========

[Stephen Chennney, 2016-03-10 20:03:30]

schenney@chromium.org changed reviewers:
+ leviw@chromium.org

[Stephen Chennney, 2016-03-10 20:21:14]

Better not to round the results at all. New patch coming once I figure out what
to do about the ref tests.

[Stephen Chennney, 2016-03-10 20:25:01]

Description was changed from

==========
Adjust the rounding methodology for tiled background phase

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. The solution is
to use the same integer snappaing method for space and phase.
Experimentation suggests that rounding both is the best strategy.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticably better. Other results move the background by a pixel
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any scheme will fail at some zoom level
due to the fact we are rounding tile size, dest rect etc to integers.

R=trchen@chromium.org
BUG=593861
==========

to

==========
Adjust the rounding methodology for tiled background phase

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. But Skia requires
floats for phase and space anyway, so better not to round at all.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
drawing rects to integers in a way that differs from phase and space.

R=trchen@chromium.org
BUG=593861
==========

[commit-bot: I haz the power, 2016-03-10 20:33:24]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-03-10 20:33:25]

Dry run: Try jobs failed on following builders:
  linux_chromium_asan_rel_ng on tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)
  linux_chromium_chromeos_rel_ng on tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[trchen, 2016-03-11 00:31:16]

The math doesn't feel right to me. The unsnapped phase value depends on
fmodf(leftEdgeOfFirstTile, unsnappedTileSize), which equals to 

leftEdgeOfFirstTile - q * unsnappedTileSize
where q = floorf(leftEdgeOfFirstTile / unsnappedTileSize)

Then we snapped the tile size, shouldn't the phase be adjusted by q times of the
difference?

In other words, I think we should delay the computation of the phase only after
other parameters have been snapped.

By the way you forgot to mention FillLayer::hasRepeatXY in your CL description.
:)

[Stephen Chennney, 2016-03-11 15:37:42]

On 2016/03/11 00:31:16, trchen wrote:
> The math doesn't feel right to me. The unsnapped phase value depends on
> fmodf(leftEdgeOfFirstTile, unsnappedTileSize), which equals to 
> 
> leftEdgeOfFirstTile - q * unsnappedTileSize
> where q = floorf(leftEdgeOfFirstTile / unsnappedTileSize)
> 
> Then we snapped the tile size, shouldn't the phase be adjusted by q times of
the
> difference?
> 
> In other words, I think we should delay the computation of the phase only
after
> other parameters have been snapped.

It's actually worse than that for tiling with space. In the simple case of
tiling a rectangle with the top-left tile in the corner, we want the phase to be
0, but because of the way we compute the space, we end up setting the phase to
tileSize + space. The error I was seeing comes from rounding phase differently
to "tilesize + space". So maybe the right fix is to make sure that we adjust
phase by the sum of the rounding offsets on tileSize and space. I'll try that.

Specifically, in the background-repeat-with-background-color test, considering
width only:
rect width = 110
tile size = 24
space = 4.6...
phase = 28.6...

So when we floor phase but round space we get
space = 5
phase = 28

That means pixel 28 of an infinite tile+space tiling ends up at 0 in the output,
and pixel 28 is space.

> By the way you forgot to mention FillLayer::hasRepeatXY in your CL
description.
> :)

I meant to have that in a separate patch.

[Stephen Chennney, 2016-03-11 17:35:55]

The CQ bit was checked by schenney@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-03-11 17:36:37]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1780893004/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1780893004/20001

[Stephen Chennney, 2016-03-11 17:37:05]

This version snaps as we go, and removes a couple of methods and verifies that
we have a redundant variable. It's not to land, but I want to run it through the
trybots so you can see what changes in tests.

This is the snapping behavior that I think we should land.

[Stephen Chennney, 2016-03-11 17:39:01]

https://codereview.chromium.org/1780893004/diff/20001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/fast/backgrounds/background-position-zoomed.html
(right):

https://codereview.chromium.org/1780893004/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/fast/backgrounds/background-position-zoomed.html:6:
height: 26px;
I might not need this now. Have to check.

https://codereview.chromium.org/1780893004/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/paint/BackgroundImageGeometry.h (right):

https://codereview.chromium.org/1780893004/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/paint/BackgroundImageGeometry.h:32: LayoutSize
tileSize() const { ASSERT(m_imageContainerSize == m_tileSize); return
m_tileSize; }
I'll remove these before landing, and remove the notion of imageContainerSize,
as it is redundant.

[commit-bot: I haz the power, 2016-03-11 18:41:29]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-03-11 18:41:30]

Dry run: Try jobs failed on following builders:
  linux_chromium_rel_ng on tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[Stephen Chennney, 2016-03-11 20:02:25]

The CQ bit was checked by schenney@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-03-11 20:02:56]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1780893004/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1780893004/40001

[Stephen Chennney, 2016-03-11 20:08:53]

Description was changed from

==========
Adjust the rounding methodology for tiled background phase

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. But Skia requires
floats for phase and space anyway, so better not to round at all.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
drawing rects to integers in a way that differs from phase and space.

R=trchen@chromium.org
BUG=593861
==========

to

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
compatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
==========

[Stephen Chennney, 2016-03-11 20:09:36]

Ready for review now. Take a look at the layout test results from the second
patch to see how the results have changed.

[Stephen Chennney, 2016-03-11 20:11:29]

Description was changed from

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
compatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
==========

to

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
compatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

This patch also removes the redundant imageContainerSize member and accessor,
because it is always the same as the tile size.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
==========

[Stephen Chennney, 2016-03-11 20:27:42]

Patchset #4 (id:60001) has been deleted

[Stephen Chennney, 2016-03-11 21:01:21]

The CQ bit was checked by schenney@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-03-11 21:02:31]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1780893004/100001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1780893004/100001

[leviw_travelin_and_unemployed, 2016-03-11 21:47:32]

LGTM.

Thanks for looking into this!

[trchen, 2016-03-11 22:24:22]

On 2016/03/11 21:47:32, leviw wrote:
> LGTM.
> 
> Thanks for looking into this!

lgtm too. Thanks for making this thing saner!
I hoped we can see repeat:space align the bottom right edges, but alas, that is
really difficult.

[commit-bot: I haz the power, 2016-03-11 23:33:56]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-03-11 23:33:57]

Dry run: This issue passed the CQ dry run.

[Stephen Chennney, 2016-03-14 13:13:31]

Description was changed from

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
compatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

This patch also removes the redundant imageContainerSize member and accessor,
because it is always the same as the tile size.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
==========

to

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
incompatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

This patch also removes the redundant imageContainerSize member and accessor,
because it is always the same as the tile size.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
==========

[Stephen Chennney, 2016-03-14 13:17:46]

The CQ bit was checked by schenney@chromium.org

[commit-bot: I haz the power, 2016-03-14 13:17:58]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1780893004/100001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1780893004/100001

[commit-bot: I haz the power, 2016-03-14 14:34:26]

Description was changed from

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
incompatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

This patch also removes the redundant imageContainerSize member and accessor,
because it is always the same as the tile size.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
==========

to

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
incompatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

This patch also removes the redundant imageContainerSize member and accessor,
because it is always the same as the tile size.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
==========

[commit-bot: I haz the power, 2016-03-14 14:34:27]

Committed patchset #5 (id:100001)

[commit-bot: I haz the power, 2016-03-14 14:36:05]

Description was changed from

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
incompatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

This patch also removes the redundant imageContainerSize member and accessor,
because it is always the same as the tile size.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861
==========

to

==========
Adjust the pixel snapping methodology for tiled backgrounds

The BackgroundImageGeometry::pixelSnapGeometry method was using rounding
for tile spacing and flooring for phase, resulting in one pixel gaps in
coverage on the top/left of a repeat: space tiled image. This results in
incompatible rounding for phase, space and tilesize, the combination of which
leads to sub-optimal tiling for background-repeat: space; content, including
missing coverage at the edge of a tiled area.

The new test explicitly shows this issue. Existing results for
svg/as-background-image/background-repeat.html,
ietestcenter/css3/bordersbackgrounds/background-repeat-space-padding-box.htm
and fast/images/color-profile-mask-image-svg.html are
noticeably better. Other results move the background by a pixel or less
with no negative impact on the quality of the result.

The test fast/backgrounds/background-position-zoomed.html was relying on
the floor behavior due to the test zooming odd numbers by 0.5. A minor test
change fixes the problem. Note that any test relying on pulling exact pixels
from the image will fail at some zoom level due to the fact we are rounding
to integers.

This patch also removes the redundant imageContainerSize member and accessor,
because it is always the same as the tile size.

R=trchen@chromium.org, leviw@chromium.org
BUG=593861

Committed: https://crrev.com/a75de675cbb8f9f594fabc68236a4547975ad7fe
Cr-Commit-Position: refs/heads/master@{#380972}
==========

[commit-bot: I haz the power, 2016-03-14 14:36:07]

Patchset 5 (id:??) landed as
https://crrev.com/a75de675cbb8f9f594fabc68236a4547975ad7fe
Cr-Commit-Position: refs/heads/master@{#380972}