[FFmpeg-devel] [PATCH] Fix Ogg data_offset computation.

[Aaron Colwell]

Hi,

Ping. Can I get an up or down vote on this patch please. It's a quick read
with minimal impact.

Thanks,
Aaron

On Tue, Nov 16, 2010 at 9:47 AM, Aaron Colwell <acolwell at chromium.org>wrote:

> Hi,
>
> Here is a patch that fixes the data_offset computation for Ogg
> files that have the first video frame split across several pages.
>
> Here is an example where the code fails.
>
> Key:
> v - An Ogg page that contains a fragment of a video packet.
>      (May or may not have continued flag set and has a packet
>       count of 0)
> V - An Ogg page that contains the end of a fragmented packet
>      as well as other full video packets. (ie Has continued flag
>      set, granulepos set, & packet count > 0)
> A - An Ogg page that contains complete audio packets.
>
> The current code fails to compute data_offset properly if it
> encounters a file that starts with a page sequence of
> 'vvAV'.
>
> The code fails because it is assuming that the stream that
> it receives the first "non-header" packet from indicates the
> start of the stream data. In the scenario above this isn't the
> case because the first packet decoded is audio, but the
> stream data actually starts with video pages.
>
> My patch simply allows data_offset to be reset to a lower
> file offset if a file like this is encountered. It selects the
> lowest offset across all streams instead of simply selecting
> the first stream that outputs a packet.
>
> I checked with some people on the #theora IRC channel to
> make sure that this is a valid Ogg page sequence. They
> claimed that it was. I've encountered this  sequence
> in a file generated by recordMyDesktop.
>
>
> Aaron
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ogg_data_offset.patch
Type: text/x-patch
Size: 522 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20101118/4a183c1b/attachment.bin>