[FFmpeg-devel] [PATCH 1/3] lavf: add AV_DISPOSITION_TIMED_THUMBNAILS

[Rodger Combs]

> On Jan 8, 2016, at 18:30, Marton Balint <cus at passwd.hu> wrote:
> 
> On Fri, 8 Jan 2016, Rodger Combs wrote:
> 
> [...]
>> diff --git a/libavformat/avformat.h b/libavformat/avformat.h
>> index 95a645b..eaf6270 100644
>> --- a/libavformat/avformat.h
>> +++ b/libavformat/avformat.h
>> @@ -824,11 +824,17 @@ typedef struct AVIndexEntry {
>> #define AV_DISPOSITION_CLEAN_EFFECTS     0x0200  /**< stream without voice */
>> /**
>> * The stream is stored in the file as an attached picture/"cover art" (e.g.
>> - * APIC frame in ID3v2). The single packet associated with it will be returned
>> - * among the first few packets read from the file unless seeking takes place.
>> - * It can also be accessed at any time in AVStream.attached_pic.
>> + * APIC frame in ID3v2). The first (usually only) packet associated with it
>> + * will be returned among the first few packets read from the file unless
>> + * seeking takes place. It can also be accessed at any time in
>> + * AVStream.attached_pic.
> 
> Maybe you should clarify that if the stream contains multiple packets, what will happen to AVStream.attached_pic. Will it contain all packets? Or only the first? Or some random?
This does explain that (though it's awkwardly-worded so I can see why it'd be confusing). It contains the first packet.

> 
>> */
>> #define AV_DISPOSITION_ATTACHED_PIC      0x0400
>> +/**
>> + * The stream is sparse, and contains thumbnail images, often corresponding
>> + * to chapter markers. Only ever used with AV_DISPOSITION_ATTACHED_PIC.
>> + */
>> +#define AV_DISPOSITION_TIMED_THUMBNAILS  0x0800
> 
> I don't if it is better to use this flag along with attached pic instead of keeping it distinct from it.
> 
> Changing the semantics of attached pics streams (multiple packets, timed packets) seems quite a substantial change, so if it were up to me, I'd rather keep this new flag distinct.
> 
> Where do you benefit if you use ATTACHED_PIC for this as well?
My goal there is to get existing consumers not to treat them as "normal" video streams, but I'd be OK with making this distinct instead.

> 
> Thanks,
> Marton
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel