[FFmpeg-devel] [PATCH 1/4] mxfdec: Index table based seeking

Georg Lippitsch georg.lippitsch at gmx.at
Tue Sep 27 21:52:36 CEST 2011

Am 27.09.2011, 11:08 Uhr, schrieb Tomas Härdin <tomas.hardin at codemill.se>:

> I took the liberty of refactoring and fixing a few bugs in your patch,
> in addition to rebasing my partition parsing code on top of it and
> adding some patches to support the V210 sample at
> http://titan.codemill.se/~tomhar/samples/multiindex/

Great news, thank you. I have a bit more time to work on this now, so  
hopefully I can provide some useful comments within the next days.

> Finally, a few questions about the original patch:
> What's the deal with n_delta and i in the original patch? Shouldn't at
> least track_id be used instead of i?

Actually I never found out how to correctly link between tracks and  
n_delta, so this is a hack for doing this.
SMPTE377 says "There shall be one Delta Entry for each and every indexed  
Element within the Essence Container.", but IIUC this does not imply that  
every track is actually indexed. If a track is not indexed at all, seeking  
will fail, but I've not figured out how to handle this correctly.
Or did I understand this wrong? Probably, simply using track_id is the  
right way to do it.

> The "if (i < mxf->system_item + st->index) n_delta++;" logic also looks
> strange - is its purpose to offset past the system item if present?

Yes. But as above, I never really understood this track/n_delta logic, so  
this is probably wrong as well.


More information about the ffmpeg-devel mailing list