Remove all members in ScopedStyleTree for preparing class removal

Remove all members in ScopedStyleTree for preparing class removal

As a part of ScopedStyleTree removal (issue 392018),
remove all private members in ScopedStyleTree.

|m_cache| is no longer necessary as ScopedStyleResolver
lookup from an element no longer require ascending the
DOM tree up, thanks to <style scoped> removal.

Unfortunately |m_authorStyles| cannot be removed completely
- StyleResolver still has to clear ScopedStyleResolvers
which are associated with TreeScopes, and thus moved from
ScopedStyleTree to StyleResolver (and renamed
|m_scopedStyleResolvers| to match what it contains).

BUG=392018
TEST=no layout test failures

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=179051

[kochi, 2014-07-25 04:55:37]

Takashi, could you review?

[esprehn, 2014-07-25 05:01:00]

https://codereview.chromium.org/418833002/diff/180001/Source/core/css/resolve...
File Source/core/css/resolver/StyleResolver.cpp (right):

https://codereview.chromium.org/418833002/diff/180001/Source/core/css/resolve...
Source/core/css/resolver/StyleResolver.cpp:379:
const_cast<TreeScope&>((*it)->treeScope()).clearScopedStyleResolver();
This needs to be done for Oilpan builds too, otherwise we have incorrect
ScopedStyleResolvers hanging off TreeScopes. At the very least we'll need to
guard any access to them with a check that the primary StyleResolver is valid.
It won't crash in Oilpan, but it'll give wrong values back.

[kochi, 2014-07-25 06:02:15]

(+cc haraken FYI)

Takashi and I debugged the destructor, and found a path that
should explicitly clear ScopedStyleResolver but does not.

Now StyleEngine is responsible for the scope of all ScopedStyleResolver's,
it should be fine not to clear ScopedStyleResolvers
in ~StyleResolver().

Also moved explicit clearing of Document's ScopedStyleResolver in
StyleSheetContents.cpp, which
was added in haraken's oilpan fix at:
https://codereview.chromium.org/419513002/
to StyleEngine.

PTAL

[tasak, 2014-07-25 06:02:16]

https://codereview.chromium.org/418833002/diff/180001/Source/core/css/resolve...
File Source/core/css/resolver/StyleResolver.cpp (right):

https://codereview.chromium.org/418833002/diff/180001/Source/core/css/resolve...
Source/core/css/resolver/StyleResolver.cpp:379:
const_cast<TreeScope&>((*it)->treeScope()).clearScopedStyleResolver();
On 2014/07/25 05:00:59, esprehn wrote:
> This needs to be done for Oilpan builds too, otherwise we have incorrect
> ScopedStyleResolvers hanging off TreeScopes. At the very least we'll need to
> guard any access to them with a check that the primary StyleResolver is valid.
> It won't crash in Oilpan, but it'll give wrong values back.

I chatted with kochi-san.  I think, this clean-up should be done in
StyleEngine::clearResolver, because only  StyleEngine can destory StyleResolver.

[kochi, 2014-07-25 06:03:50]

+cc haraken FYI

[haraken, 2014-07-25 06:03:56]

Thanks for fixing this! I'm hoping this will fix all the failures happening on
oilpan bots :)

[tasak, 2014-07-25 06:07:59]

lgtm.

https://codereview.chromium.org/418833002/diff/200001/Source/core/css/resolve...
File Source/core/css/resolver/StyleResolver.h (right):

https://codereview.chromium.org/418833002/diff/200001/Source/core/css/resolve...
Source/core/css/resolver/StyleResolver.h:301: // FIXME: Probably this should
move to StyleEngine eventually.
NIT: I agree that this should be moved to StyleEngine.

Now only collectFeaturesTo needs this. So I think, at the second step of the
design doc:
https://docs.google.com/a/google.com/document/d/1Z13sUZoVP6swGMp2VUIyhOe9-STW...,
we can move.

[esprehn, 2014-07-28 17:40:53]

Removing the cache does mean we need to traverse up tree scopes for every
element, you could construct a scenario where that would become very expensive,
but for now lets start simple.

lgtm

[esprehn, 2014-07-28 17:40:59]

The CQ bit was checked by esprehn@chromium.org

[commit-bot: I haz the power, 2014-07-28 17:41:46]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/kochi@chromium.org/418833002/200001

[commit-bot: I haz the power, 2014-07-28 18:08:39]

Change committed as 179051

[kochi, 2014-07-29 02:00:17]

Thanks for the review!

bug filed:
https://code.google.com/p/chromium/issues/detail?id=398283

for potential perf regression.