[Jack-Devel] JACK1's Future?

[Josh Junon]

Hi; creator of Captain Jack here.

The problem with the new AudioServerPlugin model is that, as Stéphane quite
accurately pointed out, all of the plugins are run in the system bootstrap
within the coreaudiod process. Because of this, things like UNIX pipes,
FIFOs, and most other forms of IPC are blocked off. However, there is a
blanket permission on network functionality, which is what I'm using to do
all of the IPC between the plugin (coreaudiod) and the launchctl daemon.

There were two other problems that were faced: XPC, Apple's "we can do it
better" IPC framework, requires a main loop to plug into to handle the
events since it's asynchronous. Whereas XPC would allow us direct
communication with the daemon, there's nowhere to put the main loop;
sticking it just anywhere causes coreaudiod to freeze since the implemented
functionality in an AudioServerPlugin is via callback functions - there is
no 'main' entry point.

Lastly, the reason why the AudioServerPlugin doesn't use libjack directly
is because whatever form of IPC libjack requires doesn't work within the
system bootstrap context. I'm sure there's a way to use JACK's networking
functionality (I haven't dived too far into that) but I believe it would
require configuration of the AudioServerPlugin itself - something I wanted
to avoid for experience concerns.

Hence why captain jack looks like the following:

        AudioServerPlugin <==TCP/loopback==> LaunchDaemon <=> JACK

Believe me, it's not ideal.

However, this achieves two things: first, a loopback audio device isn't
incredibly helpful since we have to handle both input and output. Plus,
it's easily misconfigured to cause feedback loops. Secondly, Apple did
something right in that you get a PID of whatever program has a handle to
the audio device, included when they connect, disconnect, and exchange
audio buffers. Pass that PID to the daemon, get the process information,
and hand those off to JACK as individual ports. From there you could route
audio from just about any application to any other application, not just
the "master" in/out channels.

Hopefully I've redeemed myself just a little on Captain Jack's design.
Would love to see JACK working on OS/X so I can finish the project.

- Josh (Qix-)

On Tue, Sep 13, 2016 at 7:23 AM Stéphane Letz <letz at grame.fr> wrote:

> A note about JackRouter : this week-end I tried to better understand how
> the new AudioServerPlugIn model is working (this replaces the old
> AudioHardwarePlugIn model that was used for JackRouter). It happens that
> all AudioServerPlugIn plugins now run in the context of the « coreaudiod »
> OS X audio server (AudioHardwarePlugIn were running in the context of the
> loading CoreAudio application) . AudioServerPlugIn then appears as audio
> devices to be used by CoreAudio application.
>
> Using this model, the guys at Rogue Amoeba where able to develop the
> LoopBack application/Plugin that somewhat allows Audio IAC in a way similar
> to JACK (but without a central server AFAICS)
>
> https://rogueamoeba.com/loopback/
>
> So I was wondering if the JACK server itself could be loaded in coreaudiod
> process with this AudioServerPlugIn model, then be accessible by CoreAudio
> applications with a redesigned JackRouter plugin. A question that is not
> clear yet is how the JACK/AudioServerPlugIn would access the real audio
> hardware from coreaudiod context. The AudioServerPlugIn documentation says
> that the is not possible, since the regular CoreAudio API is not usable
> from there... but obviously Apple is able to do that for their own purpose
> (since  coreaudiod is actually the connection between the CoreAudio
> applications *and* the hardware), and the Rogue AmoebaLoopBack stuff do a
> similar thing.
>
> Technically JACK2 server code was designed to be loadable in this kind of
> use case with the libjackserver library… Could be tried at least, and could
> possibly be a better design compared to Captain Jack as far as I
> understand  Captain Jack model...
>
> Stéphane
>
>
>
>
> > Le 13 sept. 2016 à 15:55, Paul Davis <paul at linuxaudiosystems.com> a
> écrit :
> >
> > Once I get my current burst effort on Ardour's (broken) MIDI looping
> done, I'll make it a priority to do my final JACK release and hand over to
> Filipe.
> >
> > The background on Captain Jack's design sounds horrendous (thanks,
> Apple).
> >
> > On Thu, Sep 8, 2016 at 2:30 PM, Josh de Kock <josh at itanimul.li> wrote:
> > Hi,
> >
> > Back in January/February at the start of this year, Paul announced that
> he would be stepping down, and he requested that someone would stand for
> being the new maintainer.
> >
> > As far as I know, this didn't happen, up until Filipe Coelho(?) offered
> to maintain JACK1. After speaking with Paul around March, he said that once
> Ardour 5.0 was released he would do the final release of JACK1 and then
> hand over maintainership.
> >
> > There are a few pull requests on the GitHub, some of which are critical
> to the core functionality of JACK1, like the patch for OSX which begins to
> modernise JACK1's usages of the CoreAudio APIs (while still maintaining
> backward compatibility).
> >
> > I'd like to help the JACK community in getting OSX support
> up-to-scratch. There is a promising project called CaptainJack (
> https://github.com/Qix-/CaptainJack) which aims to be a modern
> replacement for JackOSX, which is a vital part of the JACK ecosystem within
> OSX. While it is just a start, it shows there is at least some interest.
> >
> > My question to the community is: Should an effort be put in to revive
> JACK1?
> >
> > And to Paul; I know it's only been a little over a month since 5.0 was
> released, but is it possible you could do this 'final release' of JACK1
> soon so we can get the ball rolling again, or at least give some sort of
> time-frame on when this can be achieved.
> >
> > I personally think there should be an effort, the main issue at this
> time is how the clients are sorted, and there is a patch for this. I would
> be willing to step forward to fix all the issues related to OSX, and
> reattempt integration of Fons' topological sort patch. In addition to this,
> JACK1 is a good code-base, albeit slightly outdated, which makes a great
> reference implementation for JACK3.
> >
> > --
> >
> > Josh
> > _______________________________________________
> > Jack-Devel mailing list
> > Jack-Devel at lists.jackaudio.org
> > http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org
> >
> > _______________________________________________
> > Jack-Devel mailing list
> > Jack-Devel at lists.jackaudio.org
> > http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org
>
> _______________________________________________
> Jack-Devel mailing list
> Jack-Devel at lists.jackaudio.org
> http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org
>
-- 
PGP signature: 03A0 B7D0 432E 1514
You may send me encrypted email using my key: https://keybase.io/qix/key.asc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.linuxaudio.org/archives/jackaudio/attachments/20160914/c34e612d/attachment.html>