[vlc-devel] [PATCH 03/13] es_out: add the ES_OUT_SET_FIRST_PCR control

Steve Lhomme robux4 at ycbcr.xyz
Fri Aug 23 07:21:24 CEST 2019


On 2019-08-22 17:07, Rémi Denis-Courmont wrote:
> Le torstaina 22. elokuuta 2019, 12.00.25 EEST Thomas Guillem a écrit :
>> On Wed, Aug 21, 2019, at 19:16, Rémi Denis-Courmont wrote:
>>> Le keskiviikkona 21. elokuuta 2019, 17.13.54 EEST Thomas Guillem a écrit :
>>>> In order to check if a module need this control, one should check the
>>>> DEMUX_GET_TIME implementation for a subtraction with a first_pcr member.
>>>
>>> That does not sound right. AFAIK, some demuxers use 0 and some 1 as their
>>> start timestamp, but this seems to assume always 0.
>>
>> Ah yes, I could set it to 1 by default then.
>>
>> Is having a possible off by 10^-6 us conceivable in the case the demuxer
>> origin is not 1 ?
> 
> I don't know. We'd better fix VLC_TS_0 to be zero (and remove it) as was
> originally planned by Laurent.
> 
> spatch should be able to spot the remaining bogus PTS/DTS zero test.

What is "spatch" ?

>>> As for the TS demuxer, I think its current time has nothing to do with the
>>> PCR whatsoever.
>>
>> Ah ?
>>
>> In demux/mpeg/ts.c:
>>      case DEMUX_GET_TIME:
>>           ...
>>          *va_arg( args, vlc_tick_t * ) = FROM_SCALE(i_pcr -
>> p_pmt->pcr.i_first);
>>
>> Ah you saying the FIRST_PCR naming is not correct ? What do you propose
>> then?
> 
> No, I'm saying that that snippet from the TS demuxer is not systematically
> used, as this patch assumes.
> 
> -- 
> Реми Дёни-Курмон
> http://www.remlab.net/
> 
> 
> 
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel
> 


More information about the vlc-devel mailing list