Bug 89189 – [chromium] Compositor should be aware of |flipped| status of video textures per-platform

• Component: New Bugs
• Status: RESOLVED FIXED

[Ami Fischman]

Migrated from http://code.google.com/p/chromium/issues/detail?id=130106

Today CCVideoLayerImpl::appendQuads assumes all video textures are "not flipped", but the orientation (0-on-top vs. 0-on-bottom) actually varies by VDA implementation, which can be determined by the OS/platform the binary is built for.  Set |flipped| accordingly.

[Ami Fischman]

Created attachment 147773 [details]
Patch

[Dana Jansens]

Comment on attachment 147773 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=147773&action=review

LGTM otherwise.

> Source/WebCore/platform/graphics/chromium/cc/CCVideoLayerImpl.cpp:214
> +#else // OS_CHROMEOS && __ARMEL__

This comment confuses me, the first case is OS_CHROMEOS && __ARMEL__. Isn't this case non-windows and (non-chromeos or non-armel)? Since there's a comment on the flipped var already, maybe this one could just go away?

[Ami Fischman]

Created attachment 147813 [details]
Patch

[Ami Fischman]

(In reply to comment #2)
> (From update of attachment 147773 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=147773&action=review
> 
> LGTM otherwise.
> 
> > Source/WebCore/platform/graphics/chromium/cc/CCVideoLayerImpl.cpp:214
> > +#else // OS_CHROMEOS && __ARMEL__
> 
> This comment confuses me, the first case is OS_CHROMEOS && __ARMEL__. Isn't this case non-windows and (non-chromeos or non-armel)? Since there's a comment on the flipped var already, maybe this one could just go away?

Sure; I hate these comments but put them in for style-guide compliance.  Gone.

[James Robinson]

Comment on attachment 147813 [details]
Patch

Fun stuff!

[WebKit Review Bot]

Comment on attachment 147813 [details]
Patch

Clearing flags on attachment: 147813

Committed r120497: <http://trac.webkit.org/changeset/120497>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Ananta Iyengar]

The ifdef for the win32 platform in the CCVideoLayerImpl::appendQuads function is wrong.

[James Robinson]

Which way do you want the bool set on windows?  Before this patch the bool was unconditionally false, this patch (attempted) to make it be true on windows - I'm assuming because somebody (Ami?) thought that DXVA wants it to be true.  Are videos upside down?

[Ananta Iyengar]

the bool for windows should be true. I will upload a change fixing the ifdef in a bit.

[James Robinson]

OK great. Please open a new bug and attach the patch to that ("Tools/Scripts/webkit-patch upload" will automate this process for you).

[Changbin Shao]

I have made a test, the bool should be false for Windows(Sandy Bridge). Is this patch updated by yet?
(In reply to comment #10)
> the bool for windows should be true. I will upload a change fixing the ifdef in a bit.

[Ananta Iyengar]

The elif defined(OS_WINDOWS) was wrong in that CL. For webkit we need to use OS(WINDOWS)