[LAU] Important test:SOLUTION !!!

[Paul Davis]

On Sat, 2007-09-15 at 21:18 -0700, Ken Restivo wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On Sun, Sep 16, 2007 at 01:07:32PM +1000, Geoff Beasley wrote:
> > On Sunday 16 September 2007 09:45, Florian Schmidt wrote:
> > 
> > >
> > > With the -rt kernels, have you also tuned your priority [jack/irq's] setup?
> > >
> > > Flo
> > done and dusted !
> > 
> > no, the actual culprit is qjackctl/jack ... if you 
> > enable "auto-refresh-connections" whilst re-recording tone as in my exaple 
> > sit back and watch the droputs ! jack surely shouldn't click when 
> > adding/subtrackting channels....but what do i know ...

jackd will always stand a good chance of causing a glitch whenever
anything is done that requires taking a lock on the "graph" (the set of
ports it is currently managing). this includes adding/removing ports,
but it also includes checking current connections.

i don't know why geoff has this option selected - its not on by default,
and for sure, every time you check the current graph state you stand a
good, though not guaranteed, chance of causing an xrun. qjackctl will
refresh its view of things every time jackd notifies it of a graph
change, which is basically what everybody wants. there will be a click
then too, but since this only happens when adding/removing ports or
clients, its not really out of the realm of the expected.

i've never even seen this option till i went to look for it, and sure
enough on my system, which runs at 128 frames/period without an RT
kernel and almost no xruns, it causes audible clicks every time the
auto-refresh happens. its a pointless option as far as i can see, and
should probably be removed/made insensitive unless qjc knows that
jackdmp is in use (which i think (but am not sure) is free of this
problem).

--p