[B2-devel] Midi support in Basilisk (by hook or by crook)

[Brent B. <br...@ke...>]

Short version:
I have been trying to run Sound Process Visual Editor in emulation.  It
is a unique program that runs only on classic MacOS, and when m68k Macs
are no longer physically supportable, there will be no program that can
do what it does.

Long version:
It's truly amazing, the trip I've been on trying to do this.

Details:
This application controls the voice engine of an Ensoniq Mirage sampler,
to assist in building patches.  There are a number of reasons why you'd
want to do this:

More modern samplers do not have analog filters.  The Mirage had Curtis
chips.  The OS that Ensoniq shipped with the Mirage (which is called
MASOS, and which does have a lot of software support on various OS's) is
not really optimized for synthesis though; it's optimized to use the
sampler as a sampler.  (The difference may be more information than
should really be in this email.)  So you only have decent software
support for the Mirage if you're running an OS on it that's not much
good for synthesis.

SoundProcess is a third party (non-Ensoniq) OS that the Mirage can boot
from which provides an entirely different voice architecture than what
it normally supports.  It is optimized for people who want to use the
Mirage for synthesis instead of traditional sampling.

There is only one program on the computer side that allows you to edit
sounds on a Mirage sampler running SoundProcess (instead of the
manufacturer provided MASOS OS), and it is a third-party app called
Sound Process Visual Editor.  It is a 68k Mac application made in the
90's by an individual who has disappeared.

Basilisk:
From what I've been able to see in the forums, Midi support in Basilisk
was abandoned because it would be too complex to guarantee stable timing
for sequencer playback.

However, for programs like this, which don't aim to play music but only
send raw data to a device, stable timing isn't really even important.
As long as the bytes get in and out, you can still edit patches.  For
that matter, the behavior of Basilisk that I've seen running on my Linux
machine with a 4-core 3.2GHz AMD Phenom II X4 955 CPU is extremely
glitch free and responsive, far faster than any real Mac Quadra I've
ever been on.  I really doubt the timing would be all that terrible even
if there were no realtime or guarantees provided.  (I've seen hardware
sequencers that didn't even have any other jobs besides being sequencers
that were truly awful in that respect, just for comparison.)

Environment:
I have setup a Linux kernel module called tty0tty on my machine.  It's
very useful, as it provides four pairs of virtual serial ports connected
by virtual nullmodems.  You get /dev/tnt0 through /dev/tnt7, with each
odd numbered device node connected bidirectionally to its even numbered
next device (0<=>1, 2<=>3, 4<=>5, 6<=>7).  This allows you to connect
any program that expects a serial port to any other program that expects
a serial port, without getting any real serial ports involved.  One
interesting benefit of this that I've observed has been that when you
use these virtual ports, it seems you can set any baud rate you want on
either side, and they don't seem to have to match.  The data will simply
come across regardless, even if one side is 2400 baud and the other is
115200, which makes sense considering there really is no actual medium.

It has been very frustrating though seeing that I can do this, and have
Hayes modem commands come through the virtual nullmodem, but even the
simplest Midi strings produce "unimplemented control code" errors on
stderr, and not a single byte makes it through the nullmodem.

Prognosis:
All I really need it to do is pass bytes.  It doesn't have to do it
well, just let them through.

Is there any hope that in the future Basilisk will get any kind of Midi
support, even for applications like this where timing isn't important?