I'm still having some problems with this. I'm using six instances of Rotter, and patching them one per physical input with esjit. All seems to go well for a while, usually; even though xruns stack up the recording still happens. However, once in a while the files being written by Rotter get mangled, resulting in there being a file of size 800k or so rather than the circa 50 megabytes that is expected. I then have to kill the Rotter instances, and jackd, and start over. I've tried this on Ubuntu Studio 11.04 and the latest Arch, with an Edirol FA-101, on a P4 3ghz with a second sata drive as an audio volume. Most vexatious is that Rotter doesn't report any issues even with the --verbose flag, and neither does jackd (other than xruns of course). Roberto and Lieven mention integer overflow; is this something that can be dealt with, or is it a hardware limitation?

On Mon, May 23, 2011 at 4:06 PM, Paul Davis <paul@linuxaudiosystems.com> wrote:

On Mon, May 23, 2011 at 6:56 PM, Eric Steinberg
<eric.steinberg@gmail.com> wrote:
> Hi folks,
> I'm trying to build an audio logger, that can record from six different
> sources to separate files. I've been using the program rotter, and I
> thought it was working but have discovered that jackd crashed and recording
> was interrupted. Unfortunately this was not reflected by qjackctl, which is
> what I use to launch jackd. The qjackctl display showed that jackd was
> running, right down to the flashing "RT", but when I tried to launch
> meterbridge it complained that the jack server was not running. The
> instances of rotter that I launched were still running, but were making
> files of just a few bytes, with no audio in them. Is this a bug in
> qjackctl? I am using Arch, on a Pentium 4, and using a firewire interface
> (Edirol FA-101).

you should use:

ps aux | grep jackd

to establish whether jack has "crashed".

my guess is that had not crashed, but was no longer accepting new
clients and was otherwise hosed.