This may be related to bug 877032. After the Youtube app crashed (I think), I tried to relaunch it and got nothing but a white screen. This is a persistent state; going back to the homescreen and tapping Youtube again brings me back to the white screen. Interestingly, b2g-ps gives me: APPLICATION USER PID PPID VSIZE RSS WCHAN PC NAME b2g root 627 1 196956 77120 ffffffff 40102330 S /system/b2g/b2g Homescreen app_713 713 627 79204 26884 ffffffff 400d8330 S /system/b2g/plugin-container (Preallocated a app_861 861 627 69836 21204 ffffffff 400b8330 S /system/b2g/plugin-container consistently. So it looks like we failed to properly launch the app, assuming it's pid 861 in this case.

Long-press brings up the app thumbnail and it's named "Youtube". So how could we get into a state where Gaia knows the name is "Youtube" but b2g-ps doesn't? This is stuff I know nothing about.

Flicking the "Youtube" app away after a long-tap, and then relaunching Youtube from the homescreen, brings me back to the same state. Looks like process 861 got reused again.

Er, except when I kill 861 from the shell, nothing is affected. So this "Youtube" app is running in either the "b2g" process or the "Homescreen" process. Or possibly it's not running in any process?

OK, so if I close or flick away the "Youtube" app in the card view, it disappears from the card view, but if I reenter card view again, it's back even though I didn't relaunch it! Something's in a bad state in Gaia I guess.

The card-view not properly closing the app sounds unrelated to the YouTube app itself. We should simply be killing the YouTube process. If that's not happening then it's bad, but not something that should block bug 877024. So removing the blocking of that bug.

We are killing the Youtube process, but the card is still around.

https://github.com/mozilla-b2g/gaia/blob/master/apps/system/js/window_manager.js#L1850 is where kill happens in gaia. I suspect this is a gaia bug. When the process of youtube app is killed, we should got mozbrowserclose event at https://github.com/mozilla-b2g/gaia/blob/master/apps/system/js/window_manager.js#L1421 and do the following stuff(removeChild in kill())

(In reply to Robert O'Callahan (:roc) (Mozilla Corporation) from comment #6) > We are killing the Youtube process, but the card is still around. Hmm...this is now starting sound similar to bug 875199.

I can reproduce a form of this bug: just open Youtube, press the home button to go back to the homescreen, open Youtube again (it reloads), and repeat several times. Eventually you get stuck on a white page with no loading happening. I get this in logcat near when this happens: E/GeckoConsole( 811): Content JS LOG at app://youtube.gaiamobile.org/youtube.js:22 in anonymous: mozbrowserlocationchange uri: https://m.youtube.com/?tsp=1&player=html5&monetize=1#/watch?v=FCKDov_eXqk&feature=g-logo-xit E/GeckoConsole( 811): Content JS LOG at app://youtube.gaiamobile.org/youtube.js:25 in anonymous: mozvisibilitychange mozHidden=false and then shortly I/Gecko ( 811): [Child 811] ###!!! ASSERTION: C planes with different sizes: 'Error', file /home/roc/mozilla-b2g18v1/content/media/nsBuiltinDecoderReader.cpp, line 167 I/Gecko ( 811): [Child 811] WARNING: Unable to create VideoData: file /home/roc/mozilla-b2g18v1/content/media/omx/nsMediaOmxReader.cpp, line 210 Not sure if that has anything to do with it. It's weird though. In this case, the process is still around.

And now I can't reproduce it again :-(.

Given that we are giving up on YouTube for v1.0.1, we might not need to worry about this issue since it sounds like something that is triggered by the wrapper-app. Of course, it still sounds like there's a bug hiding here somewhere, which would be nice to fix. But unless this is happening elsewhere, we have plenty of other bugs with higher impact to spend our time on.