[mythtv-users] ffmpeg-resync

[John Pilkington]

On 28/11/2021 01:27, Peter Bennett wrote:
> 
> On 11/27/21 5:38 PM, John Pilkington wrote:
>> Now running at commit 034a15a.  For me VDPAU playback looks good
>> and the blocking is no longer present on skips, but it does have a
>> problem with decoding error at EoF.  It will segfault if left to
>> itself but ^C can catch it while the decoder makes multiple
>> attempts to Resync the Position Map.
>> 
>> John
> 
> I do not see that problem. I see the message "decoding error End of
> file (-541478725)" and several messages "OpenGLInterop: Not a valid
> hardware frame" but no segfault and nothing unusual on the playback
> screen.
> 
> I have applied the blocking fix to the devel/ffmpeg-resync-0 in case
>  anybody wants to compare results of the two ffmpeg-resync branches.
> 
> Peter
> 
> 

Rebooted and tried some more.  The reference to the Position Map happens 
when skipping to EoF within the editor (PageDown). An immediate skip 
back from that may work, but there ought to be something better.

All my 'recordings' have current PositionMaps aka seektables, and these 
have been regenerated for the test files, which I have linked to the 
filenames of preexisting recordings.

The -v playback logs have many references to AOG (AudioOutputGraph), and 
close examination shows that often the required position will be outside 
the 'available' range.  The graph is still drawn.  I tried to make the 
AOG more meaningful years ago but eventually came to the conclusion that 
for me success was less likely that total breakage.

The quirks in the editor first arose during the big rewrite around Nov 
2019 and it was hard to trace them to individual commits.  Maybe things 
will settle down sometime...

Cheers,

John