[LAU] kernel and midi seq

[Ken Restivo]

On Thu, Jan 01, 2009 at 04:04:56PM -0800, Fernando Lopez-Lezcano wrote:
> On Thu, 2009-01-01 at 14:13 -0800, Ken Restivo wrote:
> > On Thu, Jan 01, 2009 at 04:58:13PM -0500, R Dicaire wrote:
> > > On Thu, Jan 1, 2009 at 4:45 PM, Emiliano Grilli <emillo at libero.it> wrote:
> > > 
> > > > I only tested it briefly, but with ubuntu 8.10 regular kernel alsa
> > > > MIDI behaves OK, whereas with linux-rt I had no joy: good audio
> > > > performance but unuseable MIDI latency (with an external usb MIDI
> > > > ikeyboard)
> > > 
> > > Thanks, this is what I was afraid of as I can no longer used my usb
> > > midi interface, as well as no midi functionality
> > > in everything else.
> > > 
> > > So last RT kernel where midi worked?
> > > 
> > 
> > I think 2.6.24, or so I've been told. Alas, a kernel that old won't support 
> > my new EEE's hardware, its ACPI and sleep modes, and all its built-in gadgets.
> > 
> > Slightly different question: what's the estimate on a fix for this problem? 
> > Is anyone working on fixing it? Are we talking weeks, months, years, or 
> > "forget about it"?
> 
> Apparently a new rt kernel patch tree (a git repository) is being
> created. Apparently we could see a new rt patch in the 2.6.28
> timeframe... (no hope for 2.6.26/2.6.27, apparently)
> 
> I posted a while back about the midi problems in lkml but there was not
> that much feedback. One suggestion led to a working midi input for
> external interfaces but the internal timer seemed to still not work (and
> I was not able to follow up). And I have no idea about the core reason
> for the problem.... 
> 
> BTW, Ingo is not the only rt patch developer...
> 

Thank you.

Is JACK MIDI an effective workaround? In other words, does JACK open the raw or the alsa_seq interface? Does it exhibit this problem?

IIRC there is a command-line argument in newer jackd's to provide a software-only alsaseq-type interface to all the JACK MIDI devices, be they hardware MIDI interfaces or JACK softsynths. Would doing this obviate the need to use alsaseq at all?

Thanks again for all your work on this!

-ken