RELENG_7 jerky mouse and skipping sound (still a problem -BETA3)

[Kip Macy]

On Nov 19, 2007 9:53 AM, Anish Mistry <mistry.7 at osu.edu> wrote:
> On Wednesday 07 November 2007, Anish Mistry wrote:
> > On Monday 05 November 2007, [LoN]Kamikaze wrote:
> > > Marc Fonvieille wrote:
> > > > On Thu, Oct 18, 2007 at 05:53:47PM +0200, [LoN]Kamikaze wrote:
> > > >> Anish Mistry wrote:
> > > >>> On Thursday 18 October 2007, Marc Fonvieille wrote:
> > > >>>> On Wed, Oct 17, 2007 at 12:28:30PM -0400, Anish Mistry wrote:
> > > >>>>> I just updated to RELENG_7 from 6.2 and I'm running into
> > > >>>>> some really annoying issues with jerky mouse movement and
> > > >>>>> skipping sound.  This seems to be similar to:
> > > >>>>> Re: SCHED_4BSD in RELENG_7 disturbs workflow
> > > >>>>> This happens both with 4BSD and ULE.
> > > >>>>>
> > > >>>>> I seems to happen when I'm compiling ports and a new
> > > >>>>> cc/bzip2/sh process fires off (I'm just watching top), I'll
> > > >>>>> get the skip/freezeup.
> > > >>>>
> > > >>>> [...]
> > > >>>>
> > > >>>> Using ULE and UP kernel (i.e. without SMP etc.) helped a bit
> > > >>>> the things but it's still very annoying to use firefox
> > > >>>> during ports build.  I see this lag/freeze on all boxes I
> > > >>>> use with 7.0, but it's true that with a fast machine people
> > > >>>> can ignore the problem, it's less obvious than with a 1GHz
> > > >>>> box for example.
> > > >>>
> > > >>> Yeah, I'm still seeing this behavior.  Does anyone have
> > > >>> suggestions on debugging?
> > > >>>
> > > >>> Thanks,
> > > >>
> > > >> I did post the solution in this thread.
> > > >
> > > > It has nothing to do with the mouse.
> > >
> > > Does the problem persist for you? It's gone for me, even with
> > > moused.
> >
> > Yes, the problem seems to have been fixed.  I'm back to
> > kern.hz=1000 and removed FULL_PREEMPTION.  No skipping.
> It looks like I spoke too soon.  I've just tried to compile miro and
> as it was compiling the boost-python dependency I noticed the problem
> again.  Switching kern.hz="100" seems to fix the problem.  Can any of
> the developers in this area reproduce the issue?  It's pretty easy to
> reproduce on my 1.33Ghz Athlon.
>
There is an ithread priority inversion bug that might be causing this.
The fix for that should be going in shortly.

 -Kip