Stop firing orientationchange events at pages that are not visible

Stop firing orientationchange events at pages that are not visible

Stop firing 'orientationchange' events to pages that are not visible as per the
specification:
https://w3c.github.io/screen-orientation/#handling-screen-orientation-changes

This avoids having background pages handle 'orientationchange' events and
wasting battery.

As indicated in the specification, screen.orientation keeps returning the old
orientation if the orientation changes while the page is not visible.

R=mounir@chromium.org, jochen@chromium.org
BUG=162827
TEST=screen_orientation/page-visibility.html

[Inactive, 2014-05-02 20:24:21]

This requires the following Chromium-side change (to improve test
infrastructure) for the layout tests to pass:
https://codereview.chromium.org/268673015/

[mlamouri (slow - plz ping), 2014-05-02 20:26:52]

I think the fullscreen check should be made by the embedder, not Blink.
Depending on the situation, the embedder might allow the screen orientation lock
to happen even if the document isn't fullscreen.

[Inactive, 2014-05-02 21:07:32]

On 2014/05/02 20:26:52, Mounir Lamouri wrote:
> I think the fullscreen check should be made by the embedder, not Blink.
> Depending on the situation, the embedder might allow the screen orientation
lock
> to happen even if the document isn't fullscreen.

As far as I understood, both Firefox and Internet Explorer only allow screen
orientation locking while in fullscreen, right? Do you have concrete plans to
behave differently for Chrome? Do you have use cases for allowing screen locking
when not in fullscreen?

[sof, 2014-05-08 21:41:17]

https://codereview.chromium.org/261983005/diff/60001/Source/modules/screen_or...
File Source/modules/screen_orientation/ScreenOrientation.cpp (right):

https://codereview.chromium.org/261983005/diff/60001/Source/modules/screen_or...
Source/modules/screen_orientation/ScreenOrientation.cpp:163:
stopListeningForFullscreenExit();
With Oilpan, this is a somewhat problematic cleanup action, as it runs the risk
of touching a now-dead DOMWindow. But can m_associatedDOMWindow (which
document() consults) be anything other than 0? If it can't, then this call has
no effect. If it can be non-zero, then I think you need to override
willDestroyGlobalObjectInFrame() instead and remove the fullscreen observer from
there.

[Inactive, 2014-05-12 17:49:32]

On 2014/05/02 21:07:32, Chris Dumez wrote:
> On 2014/05/02 20:26:52, Mounir Lamouri wrote:
> > I think the fullscreen check should be made by the embedder, not Blink.
> > Depending on the situation, the embedder might allow the screen orientation
> lock
> > to happen even if the document isn't fullscreen.
> 
> As far as I understood, both Firefox and Internet Explorer only allow screen
> orientation locking while in fullscreen, right? Do you have concrete plans to
> behave differently for Chrome? Do you have use cases for allowing screen
locking
> when not in fullscreen?

Mounir, ping?

Adam, Peter, if you have an opinion, feel free to pitch in as well.

One clarification as well, if this gets implemented on Chrome side instead. Does
this mean this would be in closed-source code for Android? I seem to remember
that Android's TestShell does not support the fullscreen API but I may be wrong.
It is not a big deal either way, I just want to understand better the
implications.