[vorbis] Compression Artifacts at -q 5. Help!

[Lawrence Wade]

At 11:29 AM 6/16/2003 +1000, you wrote:
>The problem is more complex than that - figuring out exactly which
>compilers/versions are broken, and on what platforms, and with what options,
>is quite difficult. Actually, doing anything at all with autoconf is
>difficult, but this is more difficult than most :-)

    Oh. Okay, then I can understand.

    My programming style is so much brute force and ignorance that 
Microsoft tried to hire me to write security patches for Outlook Express. I 
haven't cut my own source since I finished a horrible C++ course in 
university. Forgive me for not knowing what is and isn't realistically 
possible.

>If you want to contribute checks for "known-bad" sets of
>compiler/platform/options/etc., we'd happily add them - but it's a lot of
>work for pretty minimal (these days - _most_ people who compile their own
>software have more recent development tools) gain.

    I generally compile my own stuff because, even on a newer system where 
all binaries are readily available, I find it to be faster and easier than 
fighting through dependencies - for some reason, compiling from source just 
*works* 90+% of the time.

    Into the READMEs:

********************************
SOUND QUALITY WARNING:

This codec does some pretty big math involving Fourier transforms and other 
scary things. Some combinations of compilers, assemblers, linkers, 
libraries and processors don't play nicely together, causing subtle 
calculation errors which might or might not always be apparent during 
playback. Typical symptoms include reverb-ish artifacts most apparent 
during playback of quiet musical passages.

Combinations which are known to cause problems include:
  - gcc/egcs-2.91.66 on Intel/AMD (RedHat 6.x users take note!)
  - [add more]

Potential solutions in order of preference: upgrade the operating system, 
upgrade the compiler and libraries, or install precompiled binaries 
(including .rpm files).

Since these compiler combinations have caused bizarre mathematical, 
performance and reliability problems with many other software packages, it 
is *highly* recommended that you find a means to use another compiler.
***********************************

    That way, people have at least been warned...

    Lawrence Wade, BOFH
    www.glowingplate.com

<p><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-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.