STR 1. Play any video from youtube. 2. Before it shows several video frames, there's a large spinner on the screen. Then touch the screen several times. Then there's more than one play request. The video file cannot be played and only the spinner is shown, infinitely. Moreoever, when you push home button and start video app, no video cannot be played until you kill the browser.

Confirmed 2/3 times upon watching a youtube video on 8/1/2013 b2g18 build. Strange behavior. Sotaro - Two questions: 1. Why are we spinning infinitely here with no video frames if a user taps the video a few times? 2. The behavior in the last comment indicates that only one app can use the hardware decoder (i.e. lock access to it) until the app releases access to the hardware decoder. Does that align with the application sharing implementation we did or is this a bug? Leo nominating since this could impact getting signoff from leo for preinstalling YouTube. Would appreciate getting partner input here on whether we need this or not.

(In reply to Jason Smith [:jsmith] from comment #2) > Confirmed 2/3 times upon watching a youtube video on 8/1/2013 b2g18 build. > Strange behavior. > > Sotaro - Two questions: > > 1. Why are we spinning infinitely here with no video frames if a user taps > the video a few times? I checked the problem on leo device. There were following 2 pattern. - [1] When browser app is in background, video app can playback video. - [2] When browesr app is in background, video app can not playback video. when the problem happens, in allmost all cases were [1]. In this case, there was no problem to playback video in gecko side. Video frames were rendered by calling VideoFrameContainer::SetCurrentFrame(). You can hear audio, but spinner is shown in video area. Youtube site do not stop showing spinner even when the video playback started. It seems like youtube sides bug. It seems similar to Bug 881611. [2] was happend by waiting http data arrival when I checked it. > 2. The behavior in the last comment indicates that only one app can use the > hardware decoder (i.e. lock access to it) until the app releases access to > the hardware decoder. Does that align with the application sharing > implementation we did or is this a bug? Only one application can grab the hw codec, because current target devices support only one hw codec instance. When application go to background, gecko forcibly free hw codec from the applicatoin. So other application can grab hw codec when the app is in foreground. It is implemented in Bug 871485. If it works correctly, there should be no problem about sharing the hw video codec between apps. If there is a problem about it, it is a different bug like [2].

When I check the problem, almost cases were [1], though this problem could be a combination of some bugs.

By applying the patch, I checked how youtube playback works in gecko side.

(In reply to Sotaro Ikeda [:sotaro] from comment #4) > When I check the problem, almost cases were [1], though this problem could > be a combination of some bugs. At first, youtube site's problem need to be addressed.

Donovan - Can you outreach to YouTube about this bug?

Will Mozilla QA retest please. YT thinks is is fixed.

(In reply to Karen Ward [:kward] from comment #9) > Will Mozilla QA retest please. YT thinks is is fixed. Jeni - Can you retest this?

Issue repros on Build ID: 20130802164042 Gecko: /rev/ Gaia: 402e0921f78e5e93bbffb7c67c6b32164df14692 Platform Version: 18.1 When user clicks on video and rapidly keeps clicking on video four or five times the video will play the sound but picture will show the spinner showing. In order for this to happen user has to rapidly click on the video before the picture shows.

(In reply to Jeni from comment #11) > Issue repros on > Build ID: 20130802164042 > Gecko: /rev/ > Gaia: 402e0921f78e5e93bbffb7c67c6b32164df14692 > Platform Version: 18.1 > > When user clicks on video and rapidly keeps clicking on video four or five > times the video will play the sound but picture will show the spinner > showing. In order for this to happen user has to rapidly click on the video > before the picture shows. Okay, then this implies that this isn't fixed then.

I just tested this and was still able to reproduce it -- the spinner was still displayed when it should not be. Interestingly, though, the audio track of the video played even though the video was not playing. I have mentioned it to YT and will mention it again.

This was fixed by YT last week. I tested it and am no longer able to repro the issue. Thanks!