Support author Shadow DOM for INPUT elements

Support author Shadow DOM for INPUT elements.

When the first author ShadowRoot is added to an input element, we need to
disable all of UI-specific behavior.

 * Add Element::didAddShadowRoot callback, and HTMLInputElmeent::
  didAddShadowRoot replaces m_inputTypeView with one which disables
  UI-specific behavior, and re-attach a renderer.

 * InputType::destroyShadowSubTree: We should not remove the content of
  author ShadowRoots.

We don't need to take care of removal of author ShadowRoots.  It never
happens.

BUG=274205
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=157510

[tkent, 2013-09-05 06:14:34]

Finally the patch is reviewable.

[tkent, 2013-09-05 06:17:09]

ideally, we should check all of INPUT element behavior with INPUT elements +
author ShadowRoots. But it is not practical.

[esprehn, 2013-09-09 01:18:33]

On 2013/09/05 06:17:09, tkent wrote:
> ideally, we should check all of INPUT element behavior with INPUT elements +
> author ShadowRoots. But it is not practical.

You don't need the lazyReattach in the didAddShadowRoot callback. ElementShadow
already does that when you add a new root.

[tkent, 2013-09-09 02:32:49]

On 2013/09/09 01:18:33, esprehn wrote:
> You don't need the lazyReattach in the didAddShadowRoot callback.
ElementShadow
> already does that when you add a new root.

Thanks.  I'll remove it.

[Hajime Morrita, 2013-09-09 05:02:22]

Wow, it's finally coming! I haven't expected this to happen such early :-)

Let's have a set of tests for

- dynamically updating @type attribute, especially from/to hidden to non-hidden
- shadows with <shadow> and/or <content>

[tkent, 2013-09-10 03:50:54]

On 2013/09/09 05:02:22, morrita1 wrote:
> Wow, it's finally coming! I haven't expected this to happen such early :-)
> 
> Let's have a set of tests for
> 
> - dynamically updating @type attribute, especially from/to hidden to
non-hidden
> - shadows with <shadow> and/or <content>

Good catch! I found we needed to update InputType::destroyShadowSubTree.

    // It's ok to clear contents of all other ShadowRoots because they must have
    // been created by InputFieldPasswordGeneratorButtonElement, and we don't
allow adding
    // AuthorShadowRoot to HTMLInputElement.
    // FIXME: Remove the PasswordGeneratorButtonElement's shadow root and then
remove this loop.
    while ((root = root->youngerShadowRoot())) {
        root->removeChildren();
        root->appendChild(HTMLShadowElement::create(shadowTag,
element()->document()));
    }

[tkent, 2013-09-10 05:18:15]

I have updated the patch.

On 2013/09/09 05:02:22, morrita1 wrote:
> - dynamically updating @type attribute, especially from/to hidden to
non-hidden

Added shadowdom-for-input-type-change.html.

> - shadows with <shadow> and/or <content>

Updated shadow-element-inactive.html covers <shadow>, and updated
shadowdom-for-input-appearance.html covners <content>.

[Hajime Morrita, 2013-09-10 06:35:58]

LGTM. Let's give it a try to see how fuzzer beats this.

https://codereview.chromium.org/21165005/diff/26001/Source/core/dom/Element.h
File Source/core/dom/Element.h (right):

https://codereview.chromium.org/21165005/diff/26001/Source/core/dom/Element.h...
Source/core/dom/Element.h:447: virtual void didAddShadowRoot(ShadowRoot&);
It isn't great to have both didAddShadowRoot() and didUserAgentShadowRoot(). 
Can we unify these somehow?

[Hajime Morrita, 2013-09-10 06:35:59]

LGTM. Let's give it a try to see how fuzzer beats this.

[Hajime Morrita, 2013-09-10 06:36:00]

LGTM. Let's give it a try to see how fuzzer beats this.

[tkent, 2013-09-10 06:42:34]

https://codereview.chromium.org/21165005/diff/26001/Source/core/dom/Element.h
File Source/core/dom/Element.h (right):

https://codereview.chromium.org/21165005/diff/26001/Source/core/dom/Element.h...
Source/core/dom/Element.h:447: virtual void didAddShadowRoot(ShadowRoot&);
On 2013/09/10 06:35:59, morrita1 wrote:
> It isn't great to have both didAddShadowRoot() and didUserAgentShadowRoot(). 
> Can we unify these somehow?

We can. We need to update all of existing didAddUserAgentShadowRoot
implementations.

[commit-bot: I haz the power, 2013-09-10 07:38:52]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/tkent@chromium.org/21165005/34001

[commit-bot: I haz the power, 2013-09-10 07:44:07]

Change committed as 157510