Re: [PR] avcodec/s302m: enable non-PCM decoding (PR #24075)

On 2026-08-11 08:08 pm, Nicolas Gaullier via ffmpeg-devel wrote:
> On 8/11/26 12:25, Gyan Doshi via ffmpeg-devel wrote:
>> PR #24075 opened by Gyan Doshi (GyanD)
>>
>> Set up framework for non-PCM decoding in-place and add support for 
>> Dolby-E decoding.
>>
>> Useful for direct transcoding of non-PCM audio in live inputs.
>>
>>    ```fate-samples
>>    s302m/s302_20bits_DolbyE_16bits.ts
>>    s302m/s302_20bits_DolbyE_20bits.ts
>>    s302m/s302_24bits_DolbyE_20bits.ts
>>    ```
>>
>> Originally posted to the ML in early 2024.
>
> Some encoders (at least for live) support SMPTE-302 / Dolby E 
> Pass-Through, which means an mpegts stream can switch dynamically 
> between LPCM and Dolby E (yes, this is harsh when trying to decode).

If the user is aware that the stream may be volatile like that, then 
they can opt to not decode the payload.

> AFAIK, it is a conformant Dolby use case. It should be specified if it 
> is supported (from reading the code, I assume it is not), but most 
> importantly if it could be supported later with an additional patch 
> (ie. be future-proof).

It is not. With the current patch, the decoder supports a uniform 
payload where each payload packet is expected to be fully contained 
within its carrier packet. It should be possible to accommodate payload 
that spans carrier packets. And for dynamic codecs, it depends on 
whether the Dolby-E and LPCM portions both decode to the same audio 
format. If yes, that too should be possible in theory.

Will update the doc to fully reflect the current limitations.

> Also, previously, "nested decoders" were not welcome. IMO, Dolby E is 
> very particular... anyway, is it now ok for an exception here ?

That objection was by one reviewer who believed such setups were fragile 
and that it would fall on him to maintain or repair the functionality. 
As mentioned then, there were already multiple nested decoders in the 
tree since some years. No issues related to nestedness had been reported 
for them and I have not seen any reported since.

Regards,
Gyan

_______________________________________________
ffmpeg-devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]