[mythtv] Call for testing branch devel/ffmpeg-resync (mpegts-harmonize)

[John Pilkington]

On 29/07/2022 18:17, Scott Theisen wrote:
> On 7/29/22 09:35, Klaas de Waal wrote:
>> This is the commit that causes a failure on BBC streams with 
>> the devel/ffmpeg-resync branch:
>>
>> 1bba2b59941d575ade7206d0ba52a556f5d3e672 is the first bad commit
>> commit 1bba2b59941d575ade7206d0ba52a556f5d3e672
>> Author: Scott Theisen <scott.the.elm at gmail.com>
>> Date:   Fri Mar 18 00:34:50 2022 -0400
>>
>> The problem is not in the MHEG processing but in the presence of the 
>> streams that contain the additional MHEG information.
>> With MHEG processing disabled it crashes in the same way.
>>
>> Details:
>> - BBC streams received from Astra-2 satellite (a.k.a. Freesat).
>> - When playing a recording, the failure is on stopping playback.
>> - When using Live TV the failure is immediate, probably because Live 
>> TV does a start / stop / start sequence when starting playback.
>>
>> Could just be that your new reference counting code is not very good 
>> at counting....
>>
>>
> 
> I think that is the problem John was having.  add_section_stream() 
> reuses the SectionContext if it already exists, but this causes the same 
> pointer to be added to multiple streams' `priv_data`, causing crashes 
> when it is then double freed.  My patch prevented this double freeing, 
> but the DSMCC_B code is never executed, so it doesn't work.
> 
> However, with my new MHEG sample from John, I can't bring up the menu 
> for the interactive TV after only 6 commits, so I am working through 
> that commit splitting it into smaller pieces to find what broke MHEG.
> 
> Regards,
> 
> Scott

Largely as an excercise-for-the-student, I now have a build from Scott's 
pull request, labelled 32Pre4276, git f27fd9c, installed and running. 
The MHEG menu works on the recent clip, and also on an HD clip from last 
year where the cutting had retained the full set of streams.  Both show 
today's date...

Cheers,

John