[vlc-devel] Re: Strange descriptorLength fields in SL over TS

[Nico Sabbi]

Laurent Aimar wrote:

>On Mon, Mar 14, 2005, Nico Sabbi wrote:
>  
>
>>Hi,
>>yesterday, testing my SL demuxer with VLC, I found very strange
>>descriptorLength values in the IOD:
>>the length of the outermost descriptor was 67 bytes, while the length
>>of the first descriptor contained therein was 49000+.
>>Since my demuxer is very new I controlled the code, but it seems to
>>be correct, basically a plain uvlc reader:
>>
>>value = 0;
>>i = 0;
>>while(1)
>>{
>>   value |= buf[i] & 0x7f:
>>   if(buf[i] & 0x80)
>>   {
>>       value <<= 7;
>>       i++;   
>>   }
>>   else
>>      break;
>>}
>>
>>I also made a dump of the input bytes, and they were sequences of
>>0xff 0xff 0xf7 ... ... (<0x80)
>>and  so on. I don't have the dump right now, but I can post it tonight 
>>if needed.
>>
>>Is it possible there's some bug in the uvlc coder, or that the IOD 
>>generated by VLC
>>has some unusual coding?
>>    
>>
> The iod generated by vlc hasn't (really )been tested, so it has
>probably bugs (anyway vlc generates invalid TS with mpeg4 audio)
>
>  
>

yes, this is another thing that didn't sound good yesterday, and you 
know what happens
when you pass wrong data to libfaad :(

BTW vlc's demuxer doesn't seem to know that object_type==0x21 && 
stream_type==0x4
in the DecoderDescriptor identifies H264.

Do you have any sample of SL packetized data besides the one made by VLC 
itself?

Thanks,
    Nico

-- 
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html