Bug 32059 – [Qt] Plugins: Force windowless mode when there is no native window handle

• Status: RESOLVED
• Resolution: FIXED
• Component: Plug-ins

[Girish Ramakrishnan]

When flash is instantiated in windowed mode, it requires a native window handle and will not paint to an offscreen pixmap on Windows and X11.

A solution on X11 is to implement composited plugins (https://bugs.webkit.org/show_bug.cgi?id=31232). On Windows, presumably this can be done by creating a fake window and grabbing it's contents.

All that seems to be a lot of work compared :-) For flash atleast, we can force it into windowless mode by injecting wmode transparent/opaque.

[Girish Ramakrishnan]

Created attachment 44136 [details]
Inject wmode=opaque

* opaque is injected for comatibility with windowed mode (as opposed to transparent)

[WebKit Review Bot]

Attachment 44136 [details] did not pass style-queue:

Failed to run "WebKitTools/Scripts/check-webkit-style" exit_code: 1
WebCore/plugins/PluginView.cpp:73:  Alphabetical sorting problem.  [build/include_order] [4]
Total errors found: 1

[Kenneth Rohde Christiansen]

This was my original idea :-) I think Antonio said that Fennec did that, but I know it is not doing that anymore. Would be nice to find out why

[Girish Ramakrishnan]

(In reply to comment #3)
> This was my original idea :-) I think Antonio said that Fennec did that, but I
> know it is not doing that anymore. Would be nice to find out why

I left a comment in #31232 about that (pasted below)

Per my understanding, the approach of using xcomposite was chosen by fennec,
only because they didn't have flash 10 on ARM (flash 10 added support for
windowless mode). 

See https://bugzilla.mozilla.org/show_bug.cgi?id=442109#c10 and
https://bugzilla.mozilla.org/show_bug.cgi?id=442109#c11

[Girish Ramakrishnan]

Simon, comments? zecke and kenneth seem to agree.

[Simon Hausmann]

Comment on attachment 44136 [details]
Inject wmode=opaque

I agree with the approach, but I have two comments about the implementation:

1) The proposed patch appends the wmode parameter. Are you sure that flash will always use the last wmode parameter if it's specified? Wouldn't it be better to _replace_ the value of an existing wmode if specified and only otherwise append?

2) I think this kind of quirk doesn't belong into PluginView.cpp but rather into FrameLoaderClientQt::createPlugin. IMHO it's much more specific to the needs of Qt (with its QGraphicsView) than to the implementation of the spec or the needs of the other ports.

[Girish Ramakrishnan]

(In reply to comment #6)
> (From update of attachment 44136 [details])
> I agree with the approach, but I have two comments about the implementation:
> 
> 1) The proposed patch appends the wmode parameter. Are you sure that flash will
> always use the last wmode parameter if it's specified? Wouldn't it be better to
> _replace_ the value of an existing wmode if specified and only otherwise
> append?
> 

I have tested this on Windows and Linux. It does seem to respect the value that is provided at the end. 

> 2) I think this kind of quirk doesn't belong into PluginView.cpp but rather
> into FrameLoaderClientQt::createPlugin. IMHO it's much more specific to the
> needs of Qt (with its QGraphicsView) than to the implementation of the spec or
> the needs of the other ports.

Great point. Totally agree.

[Girish Ramakrishnan]

Created attachment 44397 [details]
Inject wmode=opaque (2)

Simon, I will commit only after testing on the Mac but if you can review in meantime that will be good :-)

[WebKit Review Bot]

style-queue ran check-webkit-style on attachment 44397 [details] without any errors.

[Simon Hausmann]

Comment on attachment 44397 [details]
Inject wmode=opaque (2)

r=me, although I would still prefer code that would replace the value of an existing wmode parameter and only append if there isn't any present. Just to make it more robust :)

Thanks Girish for working on this!

[Girish Ramakrishnan]

Created attachment 44400 [details]
Inject wmode=opaque (3)

Make simon happy :-)

[WebKit Review Bot]

style-queue ran check-webkit-style on attachment 44400 [details] without any errors.

[Simon Hausmann]

Comment on attachment 44400 [details]
Inject wmode=opaque (3)


r=me

[Girish Ramakrishnan]

Landed in r51759

[Hui Huang]

Hi Girish and Simon, I'm working on Symbian Flash player integration with QT Webkit and QGraphicsWebView. Your patch 51759 forces Symbian Flash player to run in Windowless mode by default. Since Symbian Flash player works in both window and windowless modes in QGraphicsWebView. We would like to run Flash Plugin in window mode on Symbian by default for performance and specification compliance reasons. Is it possible to disable this implementation for Symbian? Thanks, Hui.

[Viatcheslav Ostapenko]

As I understand, your commit completely disables transparent mode for QGraphicsWebView (which is also windowless). 
Do you have any reason for this?

[Viatcheslav Ostapenko]

(In reply to comment #14)
> Landed in r51759

Bug related to this change: https://bugs.webkit.org/show_bug.cgi?id=50495