2015-12-04, 02:56
Update, running another test after a clean boot using the same settings. Live TV looked good for about 2 hours, then it started skipping frames. The audio stayed in sync, but the picture was really jerky with the skipped frames. CPU usage remained normal for a few minutes, then climbed through the roof. So it would appear that the frame skipping doesn't result from high CPU usage, but leads to it eventually. The sound continued to play okay. I'm going to let it play for a while longer and see if the sound stops like it did earlier. I wasn't watching it as closely earlier. After a few more minutes (10 or 15), the audio began to come and go and was out of sync.
I checked the CPU frequency and it was 900MHz (I thought maybe it might be throttling back to 600MHz for some reason and causing this, but no such luck). No messages in the log file for over 2 hours and nothing in the dmesg output. I don't know what else to check. In htop, the VIRT memory stayed pretty much at 564M, but the RES increased from 101M to 103M after the problem appeared. There are basically four PIDs running and the lowest numbered PID is the one consuming all the CPU. All of them are /usr/lib/kodi/kodi.bin.
I stopped the live TV playback and restarted it on the same channel (720p). htop showed pretty much normal values, though the VIRT had dropped to 560M and RES increased to 121M. Should the RES value increase like that?
EDIT: Stopping playback again and restarting the same channel lowered VIRT to 542M and RES went down to 105M. So it doesn't appear that start/stop are leading to some type of leak as far as I can tell from looking at htop output.
On a side note, I used the Bitvise SSH client (tunnelier) to access the Pi and still had the occasional "freeze" when running htop. Whatever is causing that doesn't seem to be related to the client program. I don't notice any such "freezing" when doing speed tests, file copies etc across the network. Everything seems to be okay, and when htop stops updating for a few seconds, other things continue to work just fine.
I checked the CPU frequency and it was 900MHz (I thought maybe it might be throttling back to 600MHz for some reason and causing this, but no such luck). No messages in the log file for over 2 hours and nothing in the dmesg output. I don't know what else to check. In htop, the VIRT memory stayed pretty much at 564M, but the RES increased from 101M to 103M after the problem appeared. There are basically four PIDs running and the lowest numbered PID is the one consuming all the CPU. All of them are /usr/lib/kodi/kodi.bin.
I stopped the live TV playback and restarted it on the same channel (720p). htop showed pretty much normal values, though the VIRT had dropped to 560M and RES increased to 121M. Should the RES value increase like that?
EDIT: Stopping playback again and restarting the same channel lowered VIRT to 542M and RES went down to 105M. So it doesn't appear that start/stop are leading to some type of leak as far as I can tell from looking at htop output.
On a side note, I used the Bitvise SSH client (tunnelier) to access the Pi and still had the occasional "freeze" when running htop. Whatever is causing that doesn't seem to be related to the client program. I don't notice any such "freezing" when doing speed tests, file copies etc across the network. Everything seems to be okay, and when htop stops updating for a few seconds, other things continue to work just fine.
