Move generic shorthand serialization checks out of specific routines

Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

This patch will make it easier to fix serialization of unset, and also
fix variables references in shorthands (bug 612634).

BUG=612363, 471917
Committed: https://crrev.com/9f45f2b736d341572758154cc9c37614ac4a013c
Cr-Commit-Position: refs/heads/master@{#395295}

[Timothy Loh, 2016-05-18 07:19:30]

Description was changed from

==========
Move generic shorthand serialization checks out of specific routines

TODO: write description
---------------------This line is 72 characters long--------------------

BUG=612363
==========

to

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

BUG=612363
==========

[Timothy Loh, 2016-05-18 07:19:31]

timloh@chromium.org changed reviewers:
+ alancutter@chromium.org

[Timothy Loh, 2016-05-18 07:22:41]

Description was changed from

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

BUG=612363
==========

to

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

This patch will make it easier to fix serialization of unset, and also
fix variables references in shorthands (bug 612634).

BUG=612363
==========

[Timothy Loh, 2016-05-18 07:28:15]

Sort of a big patch. I tried splitting this up but as mentioned in the
description any sort of split seemed to make behavior broken in certain ways.

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/LayoutTe...
File
third_party/WebKit/LayoutTests/editing/style/push-down-inline-styles-expected.txt
(right):

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/LayoutTe...
third_party/WebKit/LayoutTests/editing/style/push-down-inline-styles-expected.txt:16:
PASS italic converted <span style="font-style: italic; text-decoration:
line-through;"><div>hello</div></span> to <span style="text-decoration-line:
line-through; text-decoration-style: initial; text-decoration-color:
initial;"><div>hello</div></span>
The text-decoration shorthand is not in the getPropertyValue switch so it's not
serialized as a shorthand.

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/LayoutTe...
File third_party/WebKit/LayoutTests/fast/css/cssText-shorthand.html (right):

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/LayoutTe...
third_party/WebKit/LayoutTests/fast/css/cssText-shorthand.html:19:
'border-bottom-color: initial; border-image: initial; border-left-color:
initial; border-right-color: initial; border-style: initial; border-top-color:
red; border-width: 1px;'],
border-color is not in the allowInitialInShorthand switch so it's expanded here.
The 4-value shorthands only output 'initial' for longhands when specified as
'initial', so this seems reasonable to me.

[alancutter (OOO until 2018), 2016-05-18 07:32:54]

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/css/StylePropertySerializer.cpp (right):

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.cpp:313: // "intiial"
instead of the initial values for certain shorthands, so
s/"intiial"/"initial"/?

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/css/StylePropertySerializer.h (right):

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.h:63:
GenericCheckResult genericShorthandChecks(const StylePropertyShorthand&, String&
result) const;
The word generic isn't very clear to me, I start thinking templates.
The comment suggests it could be called shorthandSerializesAsCSSWideKeyword().

[alancutter (OOO until 2018), 2016-05-18 07:47:36]

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/css/StylePropertySerializer.cpp (right):

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.cpp:314: // these are
special-cased here.
This sounds like it should be a TODO to remove this special casing. Also doesn't
Firefox use the initial values instead of "initial", if so it's also an interop
bug.

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.cpp:352:
DCHECK_LE(longhandCount, 17);
Out of bounds checking should probably be in release, not just debug.

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.cpp:390: bool
allowImplicitInitial = allowInitial ||
allowImplicitInitialInShorthand(shorthand.id());
TODO remove these allowances?

[Timothy Loh, 2016-05-18 08:09:44]

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/LayoutTe...
File third_party/WebKit/LayoutTests/fast/css/cssText-shorthand.html (right):

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/LayoutTe...
third_party/WebKit/LayoutTests/fast/css/cssText-shorthand.html:19:
'border-bottom-color: initial; border-image: initial; border-left-color:
initial; border-right-color: initial; border-style: initial; border-top-color:
red; border-width: 1px;'],
On 2016/05/18 07:28:15, Timothy Loh wrote:
> border-color is not in the allowInitialInShorthand switch so it's expanded
here.
> The 4-value shorthands only output 'initial' for longhands when specified as
> 'initial', so this seems reasonable to me.

err clarifying this, it would get serialized as border-color: red initial
initial, which is invalid.

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/css/StylePropertySerializer.cpp (right):

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.cpp:313: // "intiial"
instead of the initial values for certain shorthands, so
On 2016/05/18 07:32:54, alancutter wrote:
> s/"intiial"/"initial"/?

Done.

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.cpp:314: // these are
special-cased here.
On 2016/05/18 07:47:36, alancutter wrote:
> This sounds like it should be a TODO to remove this special casing. Also
doesn't
> Firefox use the initial values instead of "initial", if so it's also an
interop
> bug.

Added a TODO. I don't think I've seen any bugs filed on serialization
differences around this area, and IMO this is low enough priority to fix that a
TODO should be sufficient.

IIRC FF uses initial values but has lots of -moz stuff everywhere (e.g.
border:10px sets border-color to -moz-use-text-color).

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.cpp:352:
DCHECK_LE(longhandCount, 17);
On 2016/05/18 07:47:36, alancutter wrote:
> Out of bounds checking should probably be in release, not just debug.

Why? This doesn't need to be in official builds. Running layout tests will catch
if you add a shorthand with too many longhands.

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.cpp:390: bool
allowImplicitInitial = allowInitial ||
allowImplicitInitialInShorthand(shorthand.id());
On 2016/05/18 07:47:36, alancutter wrote:
> TODO remove these allowances?

implied by above TODOs

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/css/StylePropertySerializer.h (right):

https://codereview.chromium.org/1988013003/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/StylePropertySerializer.h:63:
GenericCheckResult genericShorthandChecks(const StylePropertyShorthand&, String&
result) const;
On 2016/05/18 07:32:54, alancutter wrote:
> The word generic isn't very clear to me, I start thinking templates.
> The comment suggests it could be called shorthandSerializesAsCSSWideKeyword().

This checks more than css-wide keywords though. The cases where this returns
Invalid means we know already that serialization will fail, and so the specific
shorthand serialization logic can assume that all properties are set and not to
css-wide keywords, and so on. I renamed generic->common, wdyt?

[alancutter (OOO until 2018), 2016-05-20 01:57:12]

lgtm with nit.

https://codereview.chromium.org/1988013003/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/css/StylePropertySerializer.h (right):

https://codereview.chromium.org/1988013003/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/css/StylePropertySerializer.h:63:
CommonCheckResult commonShorthandChecks(const StylePropertyShorthand&, String&
result) const;
This could return String instead of needing a separate enum, the calling code
can check for isNull() for the equivalent of Unknown.

[Timothy Loh, 2016-05-20 06:10:54]

https://codereview.chromium.org/1988013003/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/css/StylePropertySerializer.h (right):

https://codereview.chromium.org/1988013003/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/css/StylePropertySerializer.h:63:
CommonCheckResult commonShorthandChecks(const StylePropertyShorthand&, String&
result) const;
On 2016/05/20 01:57:12, alancutter wrote:
> This could return String instead of needing a separate enum, the calling code
> can check for isNull() for the equivalent of Unknown.

Done.

[Timothy Loh, 2016-05-20 06:12:04]

Description was changed from

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

This patch will make it easier to fix serialization of unset, and also
fix variables references in shorthands (bug 612634).

BUG=612363
==========

to

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

This patch will make it easier to fix serialization of unset, and also
fix variables references in shorthands (bug 612634).

BUG=612363, 471917
==========

[Timothy Loh, 2016-05-20 06:12:18]

The CQ bit was checked by timloh@chromium.org

[Timothy Loh, 2016-05-20 06:12:19]

The patchset sent to the CQ was uploaded after l-g-t-m from
alancutter@chromium.org
Link to the patchset: https://codereview.chromium.org/1988013003/#ps40001
(title: "use string return type")

[commit-bot: I haz the power, 2016-05-20 06:12:32]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1988013003/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1988013003/40001

[commit-bot: I haz the power, 2016-05-20 07:25:36]

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

[commit-bot: I haz the power, 2016-05-20 07:25:37]

Try jobs failed on following builders:
  win_chromium_rel_ng on tryserver.chromium.win (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_...)

[Timothy Loh, 2016-05-23 03:24:35]

The CQ bit was checked by timloh@chromium.org

[Timothy Loh, 2016-05-23 03:24:35]

The patchset sent to the CQ was uploaded after l-g-t-m from
alancutter@chromium.org
Link to the patchset: https://codereview.chromium.org/1988013003/#ps60001
(title: "rebase")

[commit-bot: I haz the power, 2016-05-23 03:24:49]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1988013003/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1988013003/60001

[commit-bot: I haz the power, 2016-05-23 04:30:11]

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

[commit-bot: I haz the power, 2016-05-23 04:30:12]

Try jobs failed on following builders:
  linux_android_rel_ng on tryserver.chromium.android (JOB_FAILED,
https://build.chromium.org/p/tryserver.chromium.android/builders/linux_androi...)

[Timothy Loh, 2016-05-23 06:24:12]

The CQ bit was checked by timloh@chromium.org

[Timothy Loh, 2016-05-23 06:24:13]

The patchset sent to the CQ was uploaded after l-g-t-m from
alancutter@chromium.org
Link to the patchset: https://codereview.chromium.org/1988013003/#ps80001
(title: "update one more test")

[commit-bot: I haz the power, 2016-05-23 06:24:23]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1988013003/80001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1988013003/80001

[commit-bot: I haz the power, 2016-05-23 07:43:14]

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

[commit-bot: I haz the power, 2016-05-23 07:43:15]

Try jobs failed on following builders:
  win_chromium_rel_ng on tryserver.chromium.win (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_...)

[Timothy Loh, 2016-05-23 07:50:59]

The CQ bit was checked by timloh@chromium.org

[Timothy Loh, 2016-05-23 07:51:00]

The patchset sent to the CQ was uploaded after l-g-t-m from
alancutter@chromium.org
Link to the patchset: https://codereview.chromium.org/1988013003/#ps100001
(title: "actually fix test")

[commit-bot: I haz the power, 2016-05-23 07:51:21]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1988013003/100001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1988013003/100001

[commit-bot: I haz the power, 2016-05-23 08:54:23]

Description was changed from

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

This patch will make it easier to fix serialization of unset, and also
fix variables references in shorthands (bug 612634).

BUG=612363, 471917
==========

to

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

This patch will make it easier to fix serialization of unset, and also
fix variables references in shorthands (bug 612634).

BUG=612363, 471917
==========

[commit-bot: I haz the power, 2016-05-23 08:54:25]

Committed patchset #6 (id:100001)

[commit-bot: I haz the power, 2016-05-23 08:56:08]

Description was changed from

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

This patch will make it easier to fix serialization of unset, and also
fix variables references in shorthands (bug 612634).

BUG=612363, 471917
==========

to

==========
Move generic shorthand serialization checks out of specific routines

This patch moves generic shorthand serialization checks out of the
specific shorthand serialization routines into a single location.
For a shorthand to successfully serialize, all longhands must be set
and the important flag must be set consistently.

If all longhands are the same css-wide keyword (initial or inherit,
unset not yet supported properly), then the shorthand serializes to
that keyword. If css-wide keywords are otherwise used, the serialization
of a shorthand should fail, but currently we allow initial for certain
properties.

This patch also makes a couple of other changes. These changes would
make some behavior incorrect if they are left out or done separately.
- No longer skip serializing implicit initial values in cssText. This
    affects the case where a shorthand is set and other longhands are
    overridden so we can't serialize as a shorthand.
- Remove separate background serialization codepath. Currently we call
    appendBackgroundPropertyAsText() when serializing cssText if any
    background properties are present, which serializes either the
    background shorthand or the longhands individually. This is removed
    in this patch so we just treat background like any other shorthand,
    so background properties are no longer moved to the end of cssText
    and now ordered as output by the parser.

This patch will make it easier to fix serialization of unset, and also
fix variables references in shorthands (bug 612634).

BUG=612363, 471917

Committed: https://crrev.com/9f45f2b736d341572758154cc9c37614ac4a013c
Cr-Commit-Position: refs/heads/master@{#395295}
==========

[commit-bot: I haz the power, 2016-05-23 08:56:09]

Patchset 6 (id:??) landed as
https://crrev.com/9f45f2b736d341572758154cc9c37614ac4a013c
Cr-Commit-Position: refs/heads/master@{#395295}