ogg dev - Nov 2008 - oggz-validate does not seem to check for correctly framed headers

While testing the new vcut, some of the early results had badly framed
headers.  I actually didn't notice it at first, because I was testing
with oggz-validate, but ogginfo does identify the problem:

"WARNING: Vorbis stream X does not have headers correctly framed.
Terminal header page contains additional packets or has non-zero
granulepos"

Conrad (or someone else) may want to change the tool to identify these
files.  There's a sample attached.

-Ivo
-------------- next part --------------
A non-text attachment was scrubbed...
Name: vcut-test-tones.ogg
Type: application/ogg
Size: 14364 bytes
Desc: not available
Url :
http://lists.xiph.org/pipermail/ogg-dev/attachments/20081110/edc79e5d/attachment.ogg

2008/11/11 Ivo Emanuel Gon?alves <justivo at
gmail.com>:
> While testing the new vcut, some of the early results had badly framed
> headers.  I actually didn't notice it at first, because I was testing
> with oggz-validate, but ogginfo does identify the problem:
>
> "WARNING: Vorbis stream X does not have headers correctly framed.
> Terminal header page contains additional packets or has non-zero
> granulepos"
>
> Conrad (or someone else) may want to change the tool to identify these
> files.  There's a sample attached.
Hi Ivo,

good idea, and thanks for also supplying an example file!

How did we get here? Originally oggz-validate just checked the basics
of Ogg framing, which are independent of the codec. These are the
things that liboggz's oggz_write() will enforce - things like BOS and
EOS placement, and that granulepos is non-decreasing. Later we started
adding checks for codec-specific constraints, like that Theora BOS
must come before the Vorbis BOS. Perhaps there are more codec-specific
constraints we can check for.

liboggz already knows how many headers each track should have, so it
should be fairly straightforward to add a check for this that works
for all codecs (assuming that the constraint -- that header and
non-header packets must not share a page -- applies to all Ogg media
types).

Conrad.

2008/11/11 Conrad Parker <conrad at
metadecks.org>:
> 2008/11/11 Ivo Emanuel Gon?alves <justivo at gmail.com>:
>> While testing the new vcut, some of the early results had badly framed
>> headers.  I actually didn't notice it at first, because I was
testing
>> with oggz-validate, but ogginfo does identify the problem:
>>
>> "WARNING: Vorbis stream X does not have headers correctly framed.
>> Terminal header page contains additional packets or has non-zero
>> granulepos"
>>
>> Conrad (or someone else) may want to change the tool to identify these
>> files.  There's a sample attached.
>
> Hi Ivo,
>
> good idea, and thanks for also supplying an example file!
>
> liboggz already knows how many headers each track should have, so it
> should be fairly straightforward to add a check for this that works
> for all codecs (assuming that the constraint -- that header and
> non-header packets must not share a page -- applies to all Ogg media
> types).
I've added the following checks in changeset:3778:

  Terminal header page has non-zero granulepos
  Terminal header page contains non-header packet
  Terminal header page contains non-header segment

It should work for all media types.

cheers,

Conrad.