Selectors in styles in shadowRoots should match in all of the host's shadowRoots.

Selectors in styles in shadowRoots should match in all of the host's shadowRoots.

Styles in shadow trees hosted by the same shadow host has the same scoping node, i.e. the shadow host.
So the styles can match elements in all the shadow trees, i.e.

(1) the styles should be checked when any element in a shadow tree hosted by the shadow host,
(2) the styles should have the same cascade scope,
(3) the style in older shadow tree has lower priority than the style in younger shadow tree (cascade order).

The document for this patch is:
https://docs.google.com/document/d/1qMsnHn9-dhkcDvoI-B1BR2yQ0RdWpWoZm6M5mU5gYgc/edit?usp=sharing

BUG=311717
TEST=fast/dom/shadow/style-in-shadow-matches-elements-in-all-sibling-shadows.html

[tasak, 2013-12-12 09:13:12]

I will write a document to explain this patch.

[tasak, 2013-12-16 09:13:01]

I added google docs link (about this patch) to the description.

[tasak, 2014-01-16 04:59:45]

Would you review this CL?

[dglazkov, 2014-01-16 17:31:31]

lgtm

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
File Source/core/css/resolver/ScopedStyleResolver.h (right):

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
Source/core/css/resolver/ScopedStyleResolver.h:85: // Always ++cascadeScope if
style scoped.
"++cascadeScope" lacks context here. It describes the effect, rather than the
cause.

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
File Source/core/css/resolver/ScopedStyleTree.cpp (right):

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
Source/core/css/resolver/ScopedStyleTree.cpp:69: // To make younger shadow root
a child of its older shadow root.
To make younger shadow root _act_ as a child of the older shadow root?

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
Source/core/css/resolver/ScopedStyleTree.cpp:165: if (isShadowHost(scopingNode))
{
There's a definite "style tree" (needs better name?) traversal logic that could
be split out, just like we have node traversal algorithms. But this can be done
later.

[tasak, 2014-01-20 03:50:05]

Thank you for reviewing.

I'm now looking at spec discussion
https://www.w3.org/Bugs/Public/show_bug.cgi?id=23636.

If the result is "ok", I will land this patch. If not, I will close this issue
and change crbug status to "WontFix".

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
File Source/core/css/resolver/ScopedStyleResolver.h (right):

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
Source/core/css/resolver/ScopedStyleResolver.h:85: // Always ++cascadeScope if
style scoped.
On 2014/01/16 17:31:31, Dimitri Glazkov wrote:
> "++cascadeScope" lacks context here. It describes the effect, rather than the
> cause.

Done.

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
File Source/core/css/resolver/ScopedStyleTree.cpp (right):

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
Source/core/css/resolver/ScopedStyleTree.cpp:69: // To make younger shadow root
a child of its older shadow root.
On 2014/01/16 17:31:31, Dimitri Glazkov wrote:
> To make younger shadow root _act_ as a child of the older shadow root?

Yes. Done.

https://codereview.chromium.org/102953007/diff/90001/Source/core/css/resolver...
Source/core/css/resolver/ScopedStyleTree.cpp:165: if (isShadowHost(scopingNode))
{
On 2014/01/16 17:31:31, Dimitri Glazkov wrote:
> There's a definite "style tree" (needs better name?) traversal logic that
could
> be split out, just like we have node traversal algorithms. But this can be
done
> later.

I see.
I will done in another patch.

[dglazkov, 2014-01-21 16:46:25]

Yes, Elliott and Scott ended up resolving this as wontfix. Thank you for your
work :)