Closed Bug 582467 Opened 14 years ago Closed 14 years ago

Fullscreen video runs through CAIRO_SURFACE images

Categories

(Core :: Audio/Video, defect)

x86
Windows 7
defect
Not set
normal

Tracking

()

RESOLVED FIXED
Tracking Status
blocking2.0 --- beta3+

People

(Reporter: bas.schouten, Assigned: roc)

References

Details

Attachments

(1 file)

In the latest builds on windows it appears fullscreen video is being displayed through CAIRO_SURFACE images, note YCbCr images. This means we at the very least do not get the advantage of hardware YUV conversion.
I wonder if we should add some hooks so we can test whether we're going through fallback paths.
Assignee: nobody → roc
blocking2.0: --- → beta3+
I looked into this. While a fullscreen video is playing, I don't see a problem. I don't see us hitting ImageContainer::GetCurrentAsSurface from nsVideoFrame::BuildLayer, and everything's as fast is it should be. While a fullscreen video is paused, we *are* calling GetCurrentAsSurface because we're making the video layer "inactive" and rendering it through a ThebesLayer. We should stop doing this for accelerated backends, per discussion with vlad. I can provide a patch for that. While playing a <video> element in an HTML page in-window, again I don't see a problem. And again, when it's paused there is a problem, as for fullscreen, same fix. While playing a standalone video file in-window, there is a problem. We find a layer manager, but it's for the widget of the document loaded before we started loading the standalone video file. This is because the new widget hasn't been constructed yet... Bug 130078 will fix that. Bas, exactly which situation are you concerned about?
The fix I mentioned.
Attachment #461473 - Flags: review?(vladimir)
Comment on attachment 461473 [details] [diff] [review] fix for video layer activity Vlad already wrote this in bug 574481
Attachment #461473 - Flags: review?(vladimir)
Comment on attachment 461473 [details] [diff] [review] fix for video layer activity Even though this duplicates a Vlad patch, we should land this for beta3 so fullscreen video sticks to accelerated paths. I already discussed this with Vlad and he wrote basically the same code so I think this can just land.
Attachment #461473 - Flags: review+
Whiteboard: [needs landing]
Comment on attachment 461473 [details] [diff] [review] fix for video layer activity Well, I suppose I should follow protocol here. Vlad, I want to land this for beta3.
Attachment #461473 - Flags: review+ → review?(vladimir)
Comment on attachment 461473 [details] [diff] [review] fix for video layer activity heh, sure
Attachment #461473 - Flags: review?(vladimir)
Attachment #461473 - Flags: review+
Attachment #461473 - Flags: approval2.0+
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
Whiteboard: [needs landing]
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: