[mythtv-users] 0.25-fixes 'Video Frame Buffering Failed TooMany Times' whenever a new program starts (Live TV)

Jim Stichnoth stichnot at gmail.com
Sun Apr 22 17:35:23 UTC 2012


On Sun, Apr 22, 2012 at 9:11 AM, Jim Stichnoth <stichnot at gmail.com> wrote:
> On Sun, Apr 22, 2012 at 8:47 AM, Warpme <warpme at o2.pl> wrote:
>> FYI
>> I backported all recent (0349883 & 7c7852f) master LiveTV commits to
>> 0.25-fixes and unfortunately livetv works much worse for me compared to
>> current vanilla 0.25-fixes.
>> With recent commits from master almost every transition ends in main menu,
>> while current 0.25-fixes ends let say 1 on 10.
>> With current 0.25-fixes I also observed another trigger causing livetv
>> "return main menu": pressing info on just switched new livetv segment. Doing
>> this causes hang on livetv playback every 2-4 tries.
>> Sorry passing not nice feedback....
>>
>> My sys has separate BE & FE. FE are disk-less PXE booted over GE network. BE
>> part is ext3.
>
> Thanks for the feedback, now I have another person to bug for feedback
> on future changes. :)
>
> On my system running Master, transitions have pretty much stopped
> failing with the "video frame buffering failed" messages.  However,
> there is often a 15-20 second freeze between transitions which I'm now
> looking into, so I don't consider it fixed by any means.  Behavior
> tends to be slightly different depending on whether I'm viewing ATSC
> OTA or HDPVR content.  I suspect that has to do with bitrate -- lower
> bitrate recordings may be more likely to show the bugs as they are
> more likely to cause false EOF conditions.
>
> FWIW, my FE/BE setup is similar to yours, except the file systems are ext4.
>
> Progress is unfortunately slow, as I have limited windows of time to
> play around with the production backend, and then it takes time to
> study frontend and backend logs that are 20-30 MB each...

For those people still seeing the problem in Master, I would
appreciate hearing if the following makes any difference.  In
libs/libmythtv/ringbuffer.cpp, find the single instance of
"remotefile->SetTimeout(true);" (in the RingBuffer::run() method) and
comment it out.  Only the frontend needs to be restarted with this
change.  I'm still seeing problems with this mechanism even after the
related commits to Master.

Jim


More information about the mythtv-users mailing list