Only allow reprojecting shadows between ShadowRoots of the same type

We should never allow reprojecting the inside of author shadows, except when they're
being projected between ShadowRoots of the same type. This breaks all the tests that
were testing what happens if we allowed ShadowRoot on form controls and reproject
the inside of the form control, but we don't actually plan to ship that feature anyway.

I removed all the tests that can't be made into ref tests easily since the result is
a strangely busted form control and updated the other tests to show the behavior
when you add a ShadowRoot to something with author shadows.

This behavior was blocking landing a patch to remove fixPlaceholderRenderer
https://codereview.chromium.org/23035007/ which was further blocking making everything
lazyAttach.

BUG=234020, 253801, 276689
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=156668

[hayato, 2013-08-23 04:17:58]

Just in curious, could the older shadow root of the UA's shadow root be an
author shadow root?
Does this happen on the input form controls?

[esprehn, 2013-08-23 04:28:31]

On 2013/08/23 04:17:58, hayato wrote:
> Just in curious, could the older shadow root of the UA's shadow root be an
> author shadow root?
> Does this happen on the input form controls?

Nope, they all call ensureUserAgentShadowRoot() in the constructor or return
true from alwaysCreateUserAgentShadowRoot which means we create it when the
first author one is added. the UA shadow is always the oldest one. 

You can also see this from the implementation of Element::userAgentShadowRoot()

[hayato, 2013-08-23 05:35:06]

LGTM

I am afraid that I haven't followed the recent discussion on BUG=234020.

From the discussion, I assumed that we agreed that a shadow insertion point in
an author shadow root should not select elements from the UA shadow root.

That made Dimitri sad. Me too. :)

[esprehn, 2013-08-23 19:49:16]

On 2013/08/23 05:35:06, hayato wrote:
> LGTM
> 
> I am afraid that I haven't followed the recent discussion on BUG=234020.
> 
> From the discussion, I assumed that we agreed that a shadow insertion point in
> an author shadow root should not select elements from the UA shadow root.
> 
> That made Dimitri sad. Me too. :)

I believe the current decision is that once you add an author shadow root to an
element like an <input> it becomes a regular RenderBlock (instead of some fancy
renderer) and you can style it however you want and put another <input> down
inside if you want. You could the use the attributeChanged callback to
synchronize them if needed. There's no way I can think to make reprojection work
properly for UA shadows, ex.

<textarea> #shadow-root
   <div> #shadow-root // What does this shadow root get to see?
      <shadow>
      <content>
   </div>

Also note that the original fix that tkent added didn't really work since as
soon supportsShadowElementForUserAgentShadow() returns false for <input> (which
is necessary for security reasons) all the reprojection tests that I changed
here break since the UA shadow doesn't get a proper insertion point. So for
example a ShadowRoot on a range input that does <shadow></shadow>(after)
actually renders as (after){range slider} since the nodes are not distributed
properly and just get thrown on the end.

[commit-bot: I haz the power, 2013-08-23 20:31:48]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/esprehn@chromium.org/22852020/8001

[commit-bot: I haz the power, 2013-08-23 22:35:05]

Retried try job too often on linux_blink_rel for step(s) webkit_tests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=linux_blin...

[commit-bot: I haz the power, 2013-08-24 00:12:38]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/esprehn@chromium.org/22852020/21001

[commit-bot: I haz the power, 2013-08-24 01:39:21]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/esprehn@chromium.org/22852020/31001

[commit-bot: I haz the power, 2013-08-24 02:55:50]

Change committed as 156668