Alice0775 White Reporter
	Description • 9 years ago

This happens on e10s only.
This is similar to Bug 1151319, but regression range is different.

Steps to reproduce:
1. Open http://game.goo.ne.jp/choi/title/solitaire/index.html in window[A]
2. Open new window[B] and about:home

Actual results:
The Flash is continuously flashing

Expected results:
Not flashing


#1 Regression window:
The Flash completely disappears
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=c6d4680f1f5a&tochange=1968388d42c2
Triggered by: 3eb6f6c017ce	Jim Mathies — Bug 1133237 - When transitioning from a shadow layer tree that has plugins to a tree that does not, make sure the old plugins get hidden properly. r=roc


#2 Regression window:
The Flash is continuously flickering
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=4e6629bce17b&tochange=fdf4671db324
Triggered by: ddd5517ce7d2	Jim Mathies — Bug 1138181 - Be more aggressive in updating plugin geometry in the compositor, avoids filtering out important offset updates that don't involve remote layer tree updates. r=roc

	Jim Mathies [:jimm] Assignee
	Comment 1 • 9 years ago

The needinfos keep these out of our triage lists.

	Jim Mathies [:jimm] Assignee
	Comment 3 • 9 years ago

I'm not able to reproduce this having the two windows open and sitting next to each other on the desktop.

	Jim Mathies [:jimm] Assignee
	Comment 4 • 9 years ago

Oh, I think I just got this briefly, looking.

	Olli Pettay [:smaug][bugs@pettay.fi]
	Comment 5 • 9 years ago

I see this all the time on a bit slower laptop (Win7).

	Jim Mathies [:jimm] Assignee
	Comment 6 • 9 years ago

The widget visibility routines were being asked to set all plugins not included with layer updates to hidden. Since we have one compositor per top level window each compositor was trying to the plugin windows owned by other top level windows to hidden.

This is fixes the problem - basically hand the parent widget in to the visibility routines and check it against the parent of the widget before calling Show.

	(No longer employed by Mozilla) Aaron Klotz
	Comment 8 • 9 years ago

Comment on attachment 8623280 [details] [diff] [review]
patch v.1

Review of attachment 8623280 [details] [diff] [review]:
-----------------------------------------------------------------

::: gfx/layers/ipc/CompositorChild.cpp
@@ +270,5 @@
>        NS_WARNING("Unexpected, plugin id not found!");
>        continue;
>      }
> +    if (!parent) {
> +      parent = widget->GetParent();

What if widget changes and has a different parent than the one saved here? Is that possible?

	(No longer employed by Mozilla) Aaron Klotz
	Comment 9 • 9 years ago

(In reply to Aaron Klotz [:aklotz] (please use needinfo) from comment #8)
> Comment on attachment 8623280 [details] [diff] [review]
> patch v.1
> 
> Review of attachment 8623280 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: gfx/layers/ipc/CompositorChild.cpp
> @@ +270,5 @@
> >        NS_WARNING("Unexpected, plugin id not found!");
> >        continue;
> >      }
> > +    if (!parent) {
> > +      parent = widget->GetParent();
> 
> What if widget changes and has a different parent than the one saved here?
> Is that possible?

	Jim Mathies [:jimm] Assignee
	Comment 10 • 9 years ago

(In reply to Aaron Klotz [:aklotz] (please use needinfo) from comment #8)
> Comment on attachment 8623280 [details] [diff] [review]
> patch v.1
> 
> Review of attachment 8623280 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: gfx/layers/ipc/CompositorChild.cpp
> @@ +270,5 @@
> >        NS_WARNING("Unexpected, plugin id not found!");
> >        continue;
> >      }
> > +    if (!parent) {
> > +      parent = widget->GetParent();
> 
> What if widget changes and has a different parent than the one saved here?
> Is that possible?

Plugin updates here are grouped by top level window since compositors are tied to top level windows. So every one of these plugins should have the same parent. If the parent changes at some point (tab drag out for example) it shouldn't effect this code at all since the parents should all get updated.

	(No longer employed by Mozilla) Aaron Klotz
	Comment 11 • 9 years ago

Comment on attachment 8623280 [details] [diff] [review]
patch v.1

Review of attachment 8623280 [details] [diff] [review]:
-----------------------------------------------------------------

OK, thanks for the info. LGTM.

	Jim Mathies [:jimm] Assignee
	Comment 12 • 9 years ago

try push:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=6f0e4d914cd7

	Pulsebot
	Comment 13 • 9 years ago

https://hg.mozilla.org/integration/mozilla-inbound/rev/24c16a06692e

	Ryan VanderMeulen [:RyanVM]
	Comment 14 • 9 years ago

https://hg.mozilla.org/mozilla-central/rev/24c16a06692e