[icecast-dev] Icecast 2 for Live365 MP3 Relaying

[Michael Smith]

Marc Mulcahy <marc at plbb.net> said:

> Hi All,
> 
> I attempted to switch my MP3 server from Icecast 1 to Icecast 2 today.  I
> run a Live365 station, Big Planet Radio, which simply relays my local
> Icecast server.
> 
> My Icecast 2 error.log file shows a lot of the following messages when the
> Live365 server, Nanocaster 2.5.6,  is connected to it:
> 
> [2003-05-14 22:09:15] DBUG format-mp3/format_mp3_write_buf_to_client Client
> had recoverable error -1
> 
> [2003-05-14 22:09:15] DBUG format-mp3/format_mp3_write_buf_to_client Client
> had recoverable error -1
> 
> [2003-05-14 22:09:35] DBUG format-mp3/format_mp3_write_buf_to_client Client
> had recoverable error -1
> 
> 
> 
> I realize this may not be near enough information, but any idea what might
> be going on here?  When listening to the Live365 station, these errors are
> apparently manifest by breaks in the audio while the Live365 Nanocaster
> attempts to resynchronize or something.

This usually means that the client wasn't keeping up well enough. The most
common reasons for that are a) network link too slow, or b) source client (you
said you wrote your own, though libshout is meant to handle that part) is
sending too fast. 

That debug message is also slightly wrong: format_mp3.c line 189 should use
sock_error() rather than ret to print the actual error code. Not sure if
that'll actually help you, though.

Note that it _is_ a debug message, and icecast (unless there's a bug in the
mp3 format handler? It's possible, that's not the best-tested of code, since
it was largely written when I had no mp3 source client available at all) will
NOT lose data because of it. It's a sign that something might be not-so-good,
but shouldn't be a problem by itself.

It's probably worth having a better look at format_mp3_write_buf_to_client
(ooh. there's a clumsy function name! :-) if you're interested, there could be
bugs there - are you using mp3 metadata? If so, that codepath is about an
order of magnitude more complex, so there's more chance of subtle errors.

Mike

--- >8 ----
List archives:  http://www.xiph.org/archives/
icecast project homepage: http://www.icecast.org/
To unsubscribe from this list, send a message to 'icecast-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.