[vorbis-dev] Rhubarber (advanced peeler)

[Segher Boessenkool]

Christopher T King wrote:
> Fixed a typo that causes all packets to be annihilated. Works good
> otherwise!

No, it wasn't a typo; it was an April's Fools joke.

Everything in this "peeler" works fine: it is the latest
revision of my (non-lookahead, no history) peeler tool
framework, and I encourage everyone to use it for
experimentation leading to a usable peeler.  The main
problem for peelers is the ordering of (residue) data
that current libvorbis encode generates -- cutting the
end off of packets acts like a lowpass when peeling just
a little, and very rapidly leads to much bigger degeneration.
I'm hoping rebook (the successor to rehuff) will make
this better.

> diff -ur apr-1.0/apr.c apr-1.0-new/apr.c

This directory should have been a dead giveaway :)

> --- apr-1.0/apr.c Mon Mar 31 21:48:02 2003
> +++ apr-1.0-new/apr.c Wed Apr  2 19:32:21 2003
> @@ -46,7 +46,7 @@
> 
>  void peel(ogg_packet *op)
>  {
> -       op->bytes = compute_optimal_factor(op);
> +       op->bytes *= compute_optimal_factor(op);
>  }

compute_optimal_factor() is complete rubbish!  But it
returns a value >= 0.0 and < 1.0, so it _looks_ right,
but the = (instead of *=, what indeed would be what
the compute_optimal_factor() name suggests) leads to
op->bytes being set to 0 always.  Fun, eh?

<p>Sorry if I got anyones hopes up prematurely (well, I'm
not _really_ sorry, I'm just mean ;) ), I _will_ keep
working on peelability, and some day we will see a Real
Peeler,

<p>Segher

<p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.