[FFmpeg-devel] [PATCH] Correct header parsing for CDATA files

[Reimar Döffinger]

Hello,
On Mon, Nov 05, 2007 at 06:55:09PM +0900, Jonathan Wilson wrote:
> > seperate patch and please check a few more bytes, this is too easily true
> > in random data
> There are no other bytes that are constant enough in .cdata files to use to 
> identify valid ones.
> If anyone has any other ideas, please feel free to let me know. (all known 
> cdata files have the .cdata extention btw)

Only from looking at the wiki: how about byte1 & 0xe3 == 0?
What about total_samples < total size?
What about the relationship between the two "total samples" entries?

[...]
> > and the list of codec_tag <-> codec_id mappings should be exported in
> > AVInputFormat.codec_tag and av_codec_get_id() (or another function)
> > could then be used to map tag->id
> Looking at the code, I dont see codec_tag as being the right thing here, 
> codec_tag (and the places that seem to use it and reference it) are based 
> on the idea that a codec_tag is a FourCC (i.e. 4 alphanumeric characters 
> which can then be converted to upper case).

No it is not. codec_tag is just a 32 bit integer. Any interpretation
beyond that depends on the demuxer.

Greetings,
Reimar D?ffinger