flac dev - Oct 2004 - liboggflac bug (flac.exe) : wrong header field

Just playing around with my new flac code, and did some tests with stuff 
from the reference tools, and i noticed that one of the fields in the first 
header is wrong. (The number of headers field is 1 less than it should be)

The spec doc says...

" A two-byte, big-endian binary number signifying the number of header 
(non-audio) packets, including this one. "

Here's a partial dump from a file created by flac.exe

Page Location : 0 to 79
Ver No      : 0
Head Flags  : 2
Granule Pos : 0
Serial No   : 7660
Seq No      : 0
Checksum    : 1670249472
Num Segs    : 1
------------------------
Head Size   : 28
Data Size   : 51
Page Size   : 79
Num Packets : 1
------ Packet  0 (51 bytes) -------
00000000: 7F 46 4C 41 43 01 00 00 03 66 4C 61 43 00 00 00
........

The 9th byte in says there are 3 headers (including this one...)
=================================Page Location : 79 to 151
Ver No      : 0
Head Flags  : 0
Granule Pos : 0
Serial No   : 7660
Seq No      : 1
Checksum    : 12445430
Num Segs    : 1
------------------------
Head Size   : 28
Data Size   : 44
Page Size   : 72
Num Packets : 1
------ Packet  0 (44 bytes) -------
00000000: 04 00 00 28 20 00 00 00 72 65 66 65 72 65 6E 63
.....
This is the comment header packet (number 2)
==================================Page Location : 151 to 580
Ver No      : 0
Head Flags  : 0
Granule Pos : 0
Serial No   : 7660
Seq No      : 2
Checksum    : 3279053945
Num Segs    : 2
------------------------
Head Size   : 29
Data Size   : 400
Page Size   : 429
Num Packets : 1
------ Packet  0 (400 bytes) -------
00000000: 03 00 01 8C 00 00 00 00 00 00 00 00 00 00 00 00
......

This is a seek table header packet (number 3) and notice the last header 
flag is not set.
======================================
Page Location : 580 to 4724
Ver No      : 0
Head Flags  : 0
Granule Pos : 0
Serial No   : 7660
Seq No      : 3
Checksum    : 1559455392
Num Segs    : 17
------------------------
Head Size   : 44
Data Size   : 4100
Page Size   : 4144
Num Packets : 1
------ Packet  0 (4100 bytes) -------
00000000: 81 00 10 00 00 00 00 00 00 00 00 00 00 00 00 00

This is a padding header (number 4) and the last header flag is set on this 
one.

Just thought you should know as it may mean that a decoder needs to detect 
the vendor string to identify this version and add one to the field, 
assuming people have created files with this already.

Or you could take the microsoft route... and just change the spec to say 
"The number of header *not including this one*" :)

Zen.

--- illiminable <ogg@illiminable.com> wrote:
> Just playing around with my new flac code, and did some tests with
> stuff 
> from the reference tools, and i noticed that one of the fields in the
> first 
> header is wrong. (The number of headers field is 1 less than it
> should be)
> 
> The spec doc says...
> 
> " A two-byte, big-endian binary number signifying the number of
> header 
> (non-audio) packets, including this one. "
the spec is wrong in this case (if you look in the old html
there is a little note to myself in comments but it got
overlooked for the release).

I updated the page to say 'not including this one'.

the 'unknown' value is still OK to be zero since legal
streams must have at least one more header packet (the
vorbis comment).

I put this packet count in because I wasn't sure if it would
be required to know this info in advance, since the number
of vorbis header packets is always known in advance.  it may
never actually be used.

Josh



		
__________________________________
Do you Yahoo!?
Yahoo! Mail Address AutoComplete - You start. We finish.
http://promotions.yahoo.com/new_mail

----- Original Message ----- 
From: "Josh Coalson" <xflac@yahoo.com>
To: "illiminable" <ogg@illiminable.com>;
<flac-dev@xiph.org>
Sent: Thursday, October 14, 2004 1:01 AM
Subject: Re: [Flac-dev] liboggflac bug (flac.exe) : wrong header field

>> The spec doc says...
>>
>> " A two-byte, big-endian binary number signifying the number of
>> header
>> (non-audio) packets, including this one. "
>
> the spec is wrong in this case (if you look in the old html
> there is a little note to myself in comments but it got
> overlooked for the release).
>
I thought it was probably something like that...
> I updated the page to say 'not including this one'.
>
> the 'unknown' value is still OK to be zero since legal
> streams must have at least one more header packet (the
> vorbis comment).
>
Yeah... i know the zero value is ok... but i wanted to fill it in to make 
sure it's ok for all players. And also when it converts the old flac to the 
new flac, it fills in that field.

All working now... can handle both formats... original ogg flac and the new 
ogg flac.

Thanks,

Zen.