On Sun, 2004-06-06 at 21:00, Jack O'Quin wrote: > Fernando Pablo Lopez-Lezcano <[EMAIL PROTECTED]> writes: > > > > Note that I have *not* been running Nando's patch so far (but I will). > > > My hypothesis is that it is only required for the NPTL library. But, > > > I don't really know that for sure yet. > > > > Well, depends on what problems were you seeing. The patch tries to fix > > the realtime priority of the client threads. I think you said at some > > point you were seeing that... > > I was. My two main problems were: (1) failure to acquire SCHED_FIFO > and (2) frequent segfaults when calling pthread_cancel(). I have not > seen either since I started testing with LD_ASSUME_KERNEL=2.4.19 set. > > Today, I've been testing with NPTL (using the env variable). That > seems to work, too, on Debian. This is still without your patch, just > JACK 0.98.3 from CVS.
Hmmm, I'm testing with 0.98.0 (with the optimization problem patched away). Maybe something in 0.98.3 is very different? > Is there something different about the Red Hat NPTL implementation? Not that I know of, but I don't know much about it at all. I have not seen segfaults at all, just xruns (and weird situations where I can get the machine to almost lock up - everything slows down enormously, killing jack stops the problem). I just sent a (hopefully) cleaner patch to jackit-devel for the SCHED_FIFO problem, and the results of some test runs I did today in different conditions. -- Fernando
