a media-hub thread ocasionally hangs when resetting pipeline

[Alberto Aguirre]

After around 50 or so iterations (on average, but happens randomly) of creating/destroying a mediaplayer-app instance to play video using the script here: http://pastebin.ubuntu.com/8534316/ the media-hub server will hang.

When receiving the client death notification, the gstreamer engine playbin attempts to reset the pipeline with a call to gst_element_set_state(pipeline, GST_STATE_NULL) wich never completes - preventing the client disconnected signal from broadcasting and consequently the player instance is never destroyed (and remains in play state).

When another instance is started (say another mediaplayer-app instance) the media-hub server attempts to pause all other playing instances including the one waiting for the NULL state transition - this hangs as the pause transition can never complete.

Examining the thread stack, it seems the Android stagefright framework has sent a flush message to the codec, waiting until the codec responds with a flush done signal, which never happens.

If the pause request (as described above) is ignored then the next instance plays with no issue. It's almost as if the framework forgot to cleanup its pending ack request (like flush) when another instance was created.

[Jim Hodapp (jhodapp)]

Does this still happen on the latest RTM image?

[Launchpad Janitor (janitor)]

[Expired for media-hub (Ubuntu) because there has been no activity for 60 days.]