Re: [Lmuse-developer] CVS commit muse + start/stop playback

[Mathias L. <mat...@br...>]

> > Yes, indeed. Is the problem how to make the application aware of which
> > controllers the softsynth actually uses? Some kind of descriptor for all
> > controller "services"?
>
> That already implemented. MusE "asks" the Synthi for a list of usable
> controllers. This is mainly for the controller editor gui where the
> available controllers now have a name. For external devices the controller
> names and some properties are configurable in an instrument description
> file. MusE can keep track of controller changes but cannot know what you
> send with sysex events. The "sysex" thing is available to for restoring
> synth state and can be used for things MusE must not know about (e.g. sound
> font names).

Right, my bad again, confusing sysexes and controllers. Still wondering if it 
would be possible to create "sysex-services" storing descriptors over 
different states and parameters for those in an easy way. But I guess it 
would be tricky and create too much overhead...

> Another ugly thing is comunnication between synth and synth gui.
> When the gui is startet at any point it has to ask the synth for current
> status. This bidirectional communication is routed through MusE which i
> dislike.
>
> Maybe its better to move the gui + gui/synth communication behind the
> MESS interface so that MusE has only to deal with communication to the
> synthi. The synthi infrastructure should be build as a library usable for
> all syntis.

Hmmm.... I haven't really gotten a hang of that part enough to have a real 
clue... As far as I understand it, the gui is running in a separate process, 
and communicates with the synth via MusE (via midi-events handled by the 
audio/midi thread) through pipes connected to the gui-process' standard 
output. Are you thinking of moving the handling of the communication out of 
the audiothread, or something else? Or have I misunderstood the whole concept 
here? :-)

/Mathias