[mythtv] Cutlist editor is broken in 0.21 'bleeding' builds from atrpms

[John Pilkington]

David Matthews wrote:
> John Pilkington wrote:
>> This is essentially a repost from 1 February, but after a new myth build 
>> and under a new kernel. The symptoms are unchanged.  Recordings are from 
>> DVB-T in the UK.
>>
>>> The cutlist editor is now broken.  It was OK in 
>>> r12545, one of Axel's earlier 'bleeding' releases for FC5. 
>>>
>>> After a few adjustments of step size and skipping backwards and forwards
>>> the editor suddenly becomes unresponsive and mythfrontend CPU usage as
>>> shown by 'top' reaches 95% or more.
>>>
>>> Frontend log shows:
>>>
>>> 007-01-31 14:35:30.242 TV: Changing from None to WatchingPreRecorded
>>> 2007-01-31 14:35:30.265 The realtime priority setting is not enabled.
>>> 2007-01-31 14:35:30.579 Video timing method: SGI OpenGL
>>> [mpeg2video @ 0x5ee91e4]current_picture not initalized
>>> 2007-01-31 14:35:39.019 AFD Error: Unknown decoding error
>>> [mpeg2video @ 0x5ee91e4]current_picture not initalized
>>> 2007-01-31 14:35:39.528 AFD Error: Unknown decoding error
>>> [mpeg2video @ 0x5ee91e4]current_picture not initalized
>>> 2007-01-31 14:35:41.345 AFD Error: Unknown decoding error
> 
> I upgraded to the latest svn recently and started seeing these symptoms 
> when fast forwarding a recording or skipping forward.  After a few of 
> these log messages the frontend would freeze and the only option was to 
> kill it.  I traced the problem to the ffmpeg update in 12645 and some 
> code that was added to libs/libavcodec/mpeg12.c .  I ifdef'd it out and 
> it fixed the problem.  It doesn't seem to have had any adverse effects.
> 
> I don't know enough about the interaction between mpeg12.c and the 
> higher levels to know the right way of dealing with this so I'm not 
> proposing this as a patch to myth.  I suspect the problem has to do with 
> mpeg streams with errors.  Due to having a slightly noisy signal there 
> are always a few mpeg errors reported in my log.
> 
It's certainly quite possible that the basic problem is a corrupt mpeg 
stream.  I just successfully edited the most recent recording that gave 
trouble, after passing it through a lossless transcode.  Of course, that 
has a habit of erroring out too, but mencoder will usually work - more 
slowly!

Thanks for the comments.

John P