[Icecast] theora streaming problem...

[Joern Nettingsmeier]

[oddsock, this concerns your theora guide, so i'm cc:ing you]

Karl Heyes wrote:
> On Mon, 2005-02-21 at 22:04, Joern Nettingsmeier wrote:
> 
>>hi everyone!
>>
>>
>>i'm playing around with theora streaming for the linux audio conference 
>>2005 (http://lac.zkm.de).
> 
> 
> nice,
> 
> 
>>i checked out svn.xiph.org/trunk, built all ogg, vorbis, speex and 
>>theora related libs from scratch, then pulled icecast-2.2.0.
> 
> 
> 
>>three issues:
>>
>>
>>[1]
>>
>>icecast runs fine as always, but when i do
>>cat myvideo.ogg | oggfwd myhost 8000 mypasswd /test.ogg, the stream dies 
>>after a few seconds. (the icecast log is appended below.)
>>
>>the same command sequence with an ogg vorbis audio file instead of the 
>>theora video works without problems.
> 
> 
> that would indicate libshout timing thing, make sure it's using the
> right libshout. After a quick check, it looks like libshout 2.1 will be
> ok

dead on. thanks. i pulled icecast/branches/kh/libshout, is that correct? 
at least it works for me. it's not exactly obvious where libshout 2.1 is...


>>[2]
>>
>>i tried to follow oddsock's howto at 
>>http://www.oddsock.org/guides/video.php, but ezstream does not compile 
>>for me. it bails out with
> 
> ...
> 
>>probably something obvious, but i could not find where SHOUT_FORMAT_OGG 
>>is defined... libshout is installed, and it's not in there.
> 
> 
> you probably have libshout 2.0 installed

right. libshout from -kh fixed that too.
oddsock, could you add a little hint to your theora guide to make sure 
everyone gets the correct libshout?

>>[3]
>>
>>when i transcode an mpeg2 file to theora, it's barely faster than 
>>realtime on an athlon64 4000+. is this to be expected, or did i forget 
>>some optimizations? here's the ffmpeg output, so that you can see the 
>>stream properties.
> 
> 
> AFAIK the theora lib is not optimised yet, there is some work going on
> in the theora-mmx branch to improve that but check that you have a
> recent gcc producing code for athlon 64 as well.

i'm using -march=k8, which according to the gcc manpage implies all simd 
streaming optimizations the chip is capable of...


many thanks,

jörn