Re: NUV type fixes

On Wed, 2008-01-02 at 19:38 +0000, Andrew Church wrote:
> >You have my blessing. I'm been a bit distant from HEAD since I've
> >expended my holidays by fixing some things in real life, being sick (as
> >in "affected by the flu") and by improving ogg/vorbis/theora support in
> >transcode, after a direct private request (and given my own interest in
> >the area, I've happily accepted - after all, mantainership is supposed
> >to be ALSO funny, it isn't? ;) ).
> 
> Ouch--hope your flu is better! 

Yes, it is going better thanks ;)

>  I'm all for improved Ogg support myself,
> though after trying (and failing) to implement it in an environment where
> libvorbisfile isn't an option, I'm rather disturbed by the extent of
> second-system syndrome in the file format.  I think that any personal use
> I make of it is going to be raw Vorbis, just to preserve my own sanity...

I've started with best intentions, but I've ended doing a pretty dirty
hack like the one seen in libavcodec/libavformat, and, believe it or
not, this still was the cleanest thing I've been able to figure out.

I've faced a no-win situation:
1. mangle the payload of ogg packets produced to vorbis encoder (theora
created MUCH less problems), by merging/splitting with support functions
and thus violating encapsulation/layering.
2. blindly embed the *whole* ogg packet + payload in our frame payload
and go ahead like a bulldozer.

I've tried for a few days to go #1, and I've been rewarded with graceful
failures (Yes, I'm not that skilled in handling multimedia formats, and
I'm not that skilled in vorbis especially). Eventually, I've tried #2
and, hey, it worked at first try!

Later, a quick inspection revealed that other softwares/encoders often
doing like that, so either this solution isn't SO MUCH braindamaged or
vorbis/ogg interaction is a bit more rotten than expected :)

Anyway at 2008/01/01, transcode-xiph branch produces playable
ogg/vorbis/theora streams *without* using external tools, and streaming
support is on the way (being the last big missing piece fo fullfill the
request). A lot of tuning is still needed.

> At any rate, the NUV fix is now in CVS HEAD; anyone with a 64-bit system,
> please test and let me know if there are any problems.

Build test in progress (since I'm not a NUV user, I can't do much more).

Bests,

-- 
Francesco Romani // Ikitt
[ Out of memory. ~ We wish to hold the whole sky, ~ But we never will. ]