[vorbis-dev] Updated libvorbisfile patch

[olivier at infolibre.org]

On Thu, Nov 01, 2001 at 02:29:24PM -0800, rillian wrote:
> Maybe this if obvious, but all this is why multiplexed groups can only 
> begin and end at chaining boundaries. At the beginning of the stream, 
> just read pages until you find one w/o the b_o_s flag set; now you have 
> all the substream ids that should appear in that segment. You can 
> likewise count back from the e_o_s pages. This doesn't really solve the 
> bisection issue, but perhaps thinking about it this way will help.
> 
> Note that what we want libvorbisfile to do is play the first vorbis 
> substream it finds in each chaining segment, not the just the first 
> substream.

Well, if I understood correctly what you said the attached patch does
implement all this ;)
It does the b_o_s searching at beginning of the bisection function, and
doesn't much alter its functionning. It allows for 3 consecutive pages
w/o b_o_s before it forgive. This scanning cause a small overhead.

Do you think it would be an acceptable fix for this bug ? or is it just
way too ugly?

Olivier.


<HR NOSHADE>
<UL>
<LI>text/plain attachment: vorbisfile-multiplex-2.diff
</UL>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: vorbisfile-multiplex-2.diff
Type: application/octet-stream
Size: 8634 bytes
Desc: not available
Url : http://lists.xiph.org/pipermail/vorbis-dev/attachments/20011102/c345c934/vorbisfile-multiplex-2-0001.obj