Unfortunately, this is *not* an issue with a particular external codec,
although it could be a problem with an internal firmware codec. SQB
uses its own codec in its firmware to decode mp3 and flac (and soon ogg
vorbis/wma). You are just streaming the data file to the SQB from
slimserver when you are playing an mp3 file. The SQB is actually
decoding, so the lame version on your computer is not relevant. I have
been experiencing sporadic but similar problems with song transitions
for nearly two years. My library is entirely flac based. I am using
linux.

I have seen no activity on my linux boxes when this event occurs. No
alarming messages or traces. My 3 GHz CPU Celerons (no CPUFreq or
speedstep so running full speed) are so completely bored by the whole
thing that I'm almost embarrassed to be checking if the system might be
overloaded. Memory is untaxed. Power management is disabled. Network
bandwidth remains unchallenged. All systems are wired with new
switches, routers and cables rated at CAT6.

I do not use crossfade. I have verified my flac files using -t option
to flac encoder supplied with slimserver. All performance statistics
are in reasonable ranges. I have posted here in the past. I have
captured the --d_select, --d_source, and --d_slimproto logs of the
condition both prior to the stop and after the stop. You should add
these flags to your debug efforts. You will see that the SQB and
slimserver end up in a slimproto handshake that both agree is ok, but
excludes any other protocol messages and leaves the SQB "playing"
silence until some IR event wakes it up.

The difficulty here is identifying whether it is slimproto in the
server that is broken, or slimproto in the firmware. My analysis has
shown that at the end of the track that plays forever, there is a small
chunk (less than 1KB) remaining in the SQB buffer. I am settling on the
opinion that a final UDP packet is lost or fumbled someplace, and that
throws the protocol out of whack (UDP has no guarantee of delivery). A
simple enough fix might be to check if the currently playing track has
passed its duration with no final chunk received. The problem with
synchronizing the protocol (provided this is even how it works) is that
the SQB no longer respects slimproto based commands (SQB does not
respond to web interface or CLI commands) and requires direct IR
activity to correct the situation (a further indication that the
protocol has been violated by one of the partners).

Something I noticed from your list of tracks produced by the Now
Playing plugin is that the request subsystem seems to be generating two
identical callbacks for the same event (newsong event?). It seems to be
doing it pretty regularly. I have seen that in the past, but have not
seen a consistent correlation with an immediate transition state
problem. Still, I've often wondered why the request subsystem would
generate duplicate callbacks for notification requests. It kind of
raises the question if it is possible for the request subsystem to
generate *no* event.

Some things that I have done to help minimize this problem are 1)
static IP address on the server and on the SQB, 2) removal of all other
player references from slimserver(.conf/.pref), 3) disabling unneeded
formats, and 4) disabling unused plugins. Factory reset using power and
plus button is a must after installation. Again, what seems to work is
to make the server more available during song transition (so it does
not miss that final UDP packet?).

I go through a lot of SQBen as an A/V installer. I have many clients
with varying levels of slimserver and SQB2/3. They are in all cases
very happy (thank you slimdevices developers and friends). Instead of
focusing on the behavior of one SQB, I have tried to focus on the
behavior of many SQBen in a variety of configurations. I appreciate
your dedicated effort to solve this problem.

I am getting regular stops on a pair of squeezeboxes that were briefly
connected to the squeezenetwork. These SQB3 were ordered recently,
removed from the box, attached to squeezenetwork, tested with some
radio streams and then permanently wired to a network with slimserver
6.5.1. Then trouble started with both of them failing to transition
properly with 6 hours of activity. An obvious test will be to open two
more SQBen to replace the first two (with no squeezenetwork or radio
stream testing) and see if there is change.

Philip, you are not alone with these transition problems. I believe
your question about the versions of the codecs in the firmware is
relevant. I hope that someone responds with that information. When
trying to track down a problem as elusive as this one, even the
smallest clue might be the one that leads to the answer. Please
continue your efforts to identify the problem.

Richard


-- 
RichardG
------------------------------------------------------------------------
RichardG's Profile: http://forums.slimdevices.com/member.php?userid=801
View this thread: http://forums.slimdevices.com/showthread.php?t=33044

_______________________________________________
beta mailing list
[email protected]
http://lists.slimdevices.com/lists/listinfo/beta

Reply via email to