Remove StyleEngine.h include from RenderObject.h

Remove StyleEngine.h include from RenderObject.h

Achieves it by adding a header file, RenderObjectInlines.h, for use by
the compilation units that calls the inline methods.

Targets that depend on StyleEngine.h now cut in roughly half, from 678
to 278. Top ten list of other header files that are affected:

Source/core/dom/StyleEngine.h: 678 -> 278
Source/core/dom/DocumentStyleSheetCollection.h: 679 -> 279
Source/core/dom/StyleSheetCollection.h: 680 -> 280
Source/core/dom/TreeScopeStyleSheetCollection.h: 679 -> 279
Source/core/dom/StyleSheetScopingNodeList.h: 680 -> 280
Source/core/css/TreeBoundaryCrossingRules.h: 689 -> 297
Source/core/css/resolver/ScopedStyleTree.h: 690 -> 298
Source/core/css/SiblingTraversalStrategies.h: 691 -> 299
Source/core/css/SelectorFilter.h: 691 -> 299
Source/core/css/resolver/MatchRequest.h: 690 -> 298

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=177376

[davve, 2014-06-25 14:27:09]

Depends on https://codereview.chromium.org/355843002/ (and possibly
<https://codereview.chromium.org/351033003>).

Crude stats generated by calling:

ninja -C ../../out/Debug -t deps | grep "Source/core" | sort | uniq -c | sort -n

and examining the result before and after the patch.

[abarth-chromium, 2014-06-25 19:49:59]

Again, I'm not the right reviewer for this change.

[pdr., 2014-06-25 19:59:46]

-abarth
+esprehn

[esprehn, 2014-06-25 20:09:21]

https://codereview.chromium.org/358603002/diff/1/Source/core/rendering/Render...
File Source/core/rendering/RenderObjectStyle.h (right):

https://codereview.chromium.org/358603002/diff/1/Source/core/rendering/Render...
Source/core/rendering/RenderObjectStyle.h:15: return
document().styleEngine()->usesFirstLineRules() ? cachedFirstLineStyle() :
style();
I suspect neither of these really needs to be inline, can you just move them to
RenderObject.cpp instead?

[davve, 2014-06-26 07:00:33]

https://codereview.chromium.org/358603002/diff/1/Source/core/rendering/Render...
File Source/core/rendering/RenderObjectStyle.h (right):

https://codereview.chromium.org/358603002/diff/1/Source/core/rendering/Render...
Source/core/rendering/RenderObjectStyle.h:15: return
document().styleEngine()->usesFirstLineRules() ? cachedFirstLineStyle() :
style();
On 2014/06/25 20:09:21, esprehn wrote:
> I suspect neither of these really needs to be inline, can you just move them
to
> RenderObject.cpp instead?

Sure.

But as is made more visible in https://codereview.chromium.org/355843002/
style(bool) (renamed to styleOrFirstLineStyle) is currently used in several
inlined functions:

InlineBox::verticalAlign()
InlineFlowBox::borderLogical{Left,Right}
requiresLineBoxForContent
BreakingContext::handleText

This made me hesitant to outline them, because I'm not sure how I would measure
the performance impact. But perhaps we can trust the performance bots on that
and do something like this in as a fall-back plan?

[davve, 2014-06-26 13:30:45]

https://codereview.chromium.org/358603002/diff/1/Source/core/rendering/Render...
File Source/core/rendering/RenderObjectStyle.h (right):

https://codereview.chromium.org/358603002/diff/1/Source/core/rendering/Render...
Source/core/rendering/RenderObjectStyle.h:15: return
document().styleEngine()->usesFirstLineRules() ? cachedFirstLineStyle() :
style();
On 2014/06/26 07:00:33, David Vest wrote:
> On 2014/06/25 20:09:21, esprehn wrote:
> > I suspect neither of these really needs to be inline, can you just move them
> to
> > RenderObject.cpp instead?
> 
> Sure.
> 

I do seem get performance regressions when I try this locally.
Layout/line-layout.html seems to become around 3-4% slower.
Layout/line-layout-line-height.html is affected as well.

I tried to reproduce it on the bots on
https://codereview.chromium.org/359603002/ but have probably mistyped
the configuration or something. I can't read the output of the failure
for some reason.

[esprehn, 2014-06-27 08:18:09]

On 2014/06/26 at 13:30:45, davve wrote:
>
https://codereview.chromium.org/358603002/diff/1/Source/core/rendering/Render...
> File Source/core/rendering/RenderObjectStyle.h (right):
> 
>
https://codereview.chromium.org/358603002/diff/1/Source/core/rendering/Render...
> Source/core/rendering/RenderObjectStyle.h:15: return
document().styleEngine()->usesFirstLineRules() ? cachedFirstLineStyle() :
style();
> On 2014/06/26 07:00:33, David Vest wrote:
> > On 2014/06/25 20:09:21, esprehn wrote:
> > > I suspect neither of these really needs to be inline, can you just move
them
> > to
> > > RenderObject.cpp instead?
> > 
> > Sure.
> > 
> 
> I do seem get performance regressions when I try this locally.
> Layout/line-layout.html seems to become around 3-4% slower.
> Layout/line-layout-line-height.html is affected as well.
> 
> I tried to reproduce it on the bots on
> https://codereview.chromium.org/359603002/ but have probably mistyped
> the configuration or something. I can't read the output of the failure
> for some reason.

Sigh, lets name it RenderObjectInlines.h then.

[davve, 2014-06-27 14:29:54]

On 2014/06/27 08:18:09, esprehn wrote:
> Sigh, lets name it RenderObjectInlines.h then.

Done. Also made this into a independent CL. What to do about style(bool) (if
anything) can be handled separately.

[esprehn, 2014-07-01 19:29:12]

lgtm

[davve, 2014-07-02 08:37:35]

The CQ bit was checked by davve@opera.com

[commit-bot: I haz the power, 2014-07-02 08:37:45]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/davve@opera.com/358603002/40001

[commit-bot: I haz the power, 2014-07-02 13:13:45]

Change committed as 177376