[FFmpeg-devel] [PATCH] movenc: Force setting of "Samples per packet" field in SoundDescription V1 extended info
nichot20 at yahoo.com
Mon Oct 8 09:15:11 CEST 2012
On 05/10/12 19:26, Michael Niedermayer wrote:
> On Fri, Oct 05, 2012 at 06:23:11PM +0200, Michael Niedermayer wrote:
>> On Fri, Oct 05, 2012 at 09:42:52AM +0100, Tim Nicholson wrote:
>>> Apple spec is for this field to be always one for uncompressed PCM,
>>> currently this is only forced for V2 Sound descriptions and left at
>>> track->enc->frame_size for V1.
>>> Curiously this seems to come out at "1" for 'le' flavours anyway but
>>> 'be' returns "0".
>> why does the frame_size field differ between le and be ?
>> that doesnt feel correct at all
> the difference is caused simply by the LE fate test not using mov but
> rather wav
Ahh. Hadn't spotted that.
Actually the inconsistency that concerns me is that for V1 descriptors
the value is simply written from 'track->enc->frame_size', but for
version 2 it is calculated using 'get_samples_per_packet' and which
already forces the result of '1' for lpcm. AFAIK this atom is common to
both V1 and version 2 descriptors so I am wondering why the difference,
and is 'track->enc->frame_size' ever a valid figure?
> ill wait a day or 2 in case someone wants to test with common non free
> tools if they like the more correct value, before commit
I am currently running a lot of stuff through 'Baton' which seems
particularly picky, Hence the current rash of clean ups.
More information about the ffmpeg-devel