[vorbis] divide by zero and CVS version stability

[Michael Smith]

On Wednesday 16 July 2003 00:06, Hauke Duden wrote:
> Hello everyone,
>
> I recently stumbled upon an issue in the encoder of the v1.0 SDK. At the
> end of each encoding session a divide by zero (D=0) happens in
> bark_noise_hybridmp (psy.c:642). This causes a crash on Win9x systems
> (WinNT seems to handle this fine by using infinite as the result). My
> encoding code is almost identical to the encoder_example, so I'm reasonably
> sure that this is not a problem on the calling side.
>

This can happen with some input files in 1.0 (it doesn't happen with all - or 
even most - files). We haven't had any previous reports of this causing 
actual crashes on win32 before, though, which is a little strange. 

> After finding this error I tried the current CVS snapshot version and it
> seems to work fine. So I guess this is a known bug that has been fixed.
>

Yes, that's correct.

> My question now is: can the current CVS version be considered stable? If
> not, is there another post-1.0 version that I could use? I'm on a deadline
> and am a bit reluctant to include the most recent (less tested) code in our
> application. So it would be nice if you guys could help me sleep better by
> telling me which code is save to use.
>
> Also, if I use a more recent encoder, should I also use updated decoder/ogg
> stuff or will the v1.0 libraries still be compatible?

The current CVS version is very stable and well tested. It's definately 
recommended that you use it in preference to 1.0. libvorbis CVS head is 
bugfixes-only over 1.0, so there should be no problems at all.

The decoder is entirely compatible, I don't think there have been any decoder 
changes in cvs since 1.0 - so you can safely use it, or keep the 1.0 release, 
it should make no difference.

Mike

--- >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-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.