[PD] getting Pd to run on Raspberry Pi

[Miller Puckette]

To Pd list,

I got my Pi in the mail today (only 2 or 3 weeks after ordering it I think -
they're starting to catch up with the enormous demand!)  and can report what
I found works and what not.

I loaded the Raspian distro (the one everyone suggests) and following the
blog on :
http://log.liminastudio.com/programming/running-puredata-on-the-raspberry-pi

just typed

pi at raspberrypi ~ $ sudo bash
root at raspberrypi:/home/pi# apt-get update
root at raspberrypi:/home/pi# apt-get install puredata

and then pd ... sort of... runs right away.  I had to fool with some
glitches though:

1.  Pd is version 0.43-2 which doesn't yet know to try to promote itself
to real-time if it's not root.  (This is only fixed in 0.44 which might not
make it to Raspian for quite some time.)  A rather bad workaround exists:
still as root (i.e., after the "sudo" step) type

root at raspberrypi:/home/pi# chmod 4775 /usr/bin/puredata

This might be a security risk (if anyone breaks into your Pi they might be
able to design an exploit to get Pd to promote them to root) but I can't
say I'm worried about that.  (Speaking of security though, since mine is
networked and ssh-enabled, I did change the password :)

I'm having no trouble running Pd via ssh from another computer.  The pi
user is already in group audio in the Raspian distro.  X windows
forwarding is also turned on by default - so the only detail you have
to check is whether your ssh is rinngin with -X set on the originating
machine.  I did notice that one can't run pd inside sudo and still get 
X forwarding; you have to stay as the pi user.

Then about audio.  I can indeed verify that the built-in analogue audio is
wretched -- in two ways.  First, it's not even functional (and LOUDLY non-
functional :) until you se the latency to at least 40 msec (it's 25 by
default in Pd on linux.)  Second, it's bad anyway.  To my ears the alsa
play program and Pd's test-tone sound equally bad.  There's a modulation that
is sort of as if a sample out of every 1024 is getting dropped or doubled.
You can hear the modulation change as the pitch of the test tone goes up and
down.

So I pulled out my cheapo USB headset from Gigaware and, with input turned
off, was able to play the test tone with latency down to 10 msec, and got a
clean-sounding tone out.  Then, getting overconfident, I turned on audio
input and although the sound came through it has hiccups and my interface
froze (had to unplug the headphones to get Pd to bail :)

So things are sort of working but there's still some tuning to do... I'll
probably have more to report over the weekend.

cheers
Miller