Move Document.fullscreenElement to DocumentOrShadowRoot

third_party/WebKit/Source/core/dom/DocumentOrShadowRoot.idl

Index: third_party/WebKit/Source/core/dom/DocumentOrShadowRoot.idl
diff --git a/third_party/WebKit/Source/core/dom/DocumentOrShadowRoot.idl b/third_party/WebKit/Source/core/dom/DocumentOrShadowRoot.idl
index 34a53a8f428b461d2c2bee625a0a4e3c973a7763..4d7cbf131b43bd5f5aabfbdb1bf6ce4f4c10a329 100644
--- a/third_party/WebKit/Source/core/dom/DocumentOrShadowRoot.idl
+++ b/third_party/WebKit/Source/core/dom/DocumentOrShadowRoot.idl
@@ -21,4 +21,11 @@
     // PointerLock API
     // https://w3c.github.io/pointerlock/
     readonly attribute Element? pointerLockElement;
+    // Fullscreen API

[foolip, 2016/09/20 13:44:21]

Does our IDL compiler support partial interfaces for mixins? If so
DocumentOrShadowRootFullscreen.idl would be nice, I have been thinking that
maybe fullscreen could move into core/fullscreen or modules/fullscreen after the
merge into top layer.

If not, that's OK too.

[kochi, 2016/10/07 14:26:42]

Confirmed with our IDL compiler people and partial interface for mixin does not
work yet.
So let me go in this file.

[foolip, 2016/10/10 12:13:40]

Very well. Is there a bug for that?

[kochi, 2016/10/17 11:14:27]

Filed http://crbug.com/656517

+    // https://fullscreen.spec.whatwg.org/
+    // Mozilla version
+    [MeasureAs=PrefixedDocumentCurrentFullScreenElement, ImplementedAs=currentFullScreenElement] readonly attribute Element webkitCurrentFullScreenElement;

[foolip, 2016/09/20 13:44:21]

I was unclear/misleading with "important that we keep the prefixed and
unprefixed API in sync". I think that the document.fullscreenElement and
.webkitFullscreenElement should be plain aliases, but for ShadowRoot I see two
unfortunate options:

1. Add prefixed attributes, as in this CL.
2. Don't add them, and make it impossible to find the real fullscreen element.

I am working on unprefixing, so how about gating the change in behavior of the
prefixed attributes on the FullscreenUnprefixed flag? That increases the compat
risk of shipping FullscreenUnprefixed, but the other two options seem worse.

[kochi, 2016/10/07 14:26:42]

I may be still confused. Correct my if I misunderstand.

So for webkit-prefixed attributes, we don't add them in ShadowRoot, and leave
them
in Document, and in the implementation we do the new behavior gated by
FullscreenUnprefixed flag?

Original motivation for this was not to leak nodes inside shadow, but
as it is already shipped, changing this has some compat risk.

So effectively we do not change .webkitCurrentFullScreenElement and
.webkitFullscreenElement, and only make changes in .fullscreenElement,
as it is specced in whatwg fullscreen spec.

[foolip, 2016/10/10 12:13:40]

Yes, I'm suggesting no web exposed changes at all right now, but rather to
change the implementation of the (internal, non-static)
Fullscreen::fullscreenElement() and ::currentFullScreenElement() to return the
retargeted element if RuntimeEnabledFeatures::fullscreenUnprefixedEnabled().

[kochi, 2016/10/17 11:14:27]

It does not work, as Fullscreen::fullscreenElement() is also called
by other non-binding users (e.g. PaintLayerCompositor.cpp).
So we cannot rely on RuntimeEnabledFeatures to determine the retargeting
behavior in internal fullscreenElement() API.

So I still left the logic for retargeting or not in *ForBinding* functions.

+    // W3C version
+    [RuntimeEnabled=FullscreenUnprefixed] readonly attribute Element? fullscreenElement;
+    [MeasureAs=PrefixedDocumentFullscreenElement, ImplementedAs=fullscreenElement] readonly attribute Element webkitFullscreenElement;
 };