On 12/4/23 21:23, Yann Seznec wrote:
I’ll answer both of these questions at once…to be more clear, the most obvious test of the Pi issues I was having is to launch “humanCaller_test.pd” from this project: https://github.com/yannseznec/humanCall https://github.com/yannseznec/humanCall
It is a granular patch that will just play a stream of grains on startup.
When I was encountering issues, this patch would stop working after somewhere between 30-90 seconds, and print many instances of the error “restart alsa output alsa xrun recovery apparently failed”.
I am no longer having this issue, after compiling Pd 0.54.1 myself on the 32 bit Pi OS 11.
since I was just playing along with an RPi4 i thought i'd give it a try.
hardware: Raspberry Pi 4 (Model B, 4GB RAM; iirc!) using the onboard soundcard operating system: IEMberry graphics: running without monitors, but with a full Desktop connected to via VNC Pd: Pd-0.53-1, as packaged sound: ALSA
IEMberry is our in-house RPi4 distribution, it's basically a Raspberry Pi OS (32bit) with loads of audio and development tools preinstalled. https://iemberry.iem.at/ i'm using iemberry11.20231017 (as can be currently downloaded on the webpage), which is based on Raspberry Pi OS "Bullseye" 11 (the "with desktop and recommended software" flavour). The image was generated Mid-October, so it obviously did *not* use "Raspberry Pi OS (Legacy)" as can be found on the RPi homepage right now, as the latter was released 2023-12-05. Instead, the base image was the last release of Rasbian available back then, which turns out to be have been released 2023-05-03).
IEMberry provides only very few packages by itself, and instead relies on the Debian packages. It also has Debian/backports and the official Raspberry Pi OS repositories enabled (with Debian being preferred).
IEMberry comes with Pd from Debian/bullseye-backports pre-installed:
$ pd -version
Pd-0.53-1 ("") compiled for Debian (0.53.1+ds-2~bpo11+1) on 2023/01/31
at 6:55:52 UTC
in order to get your patch running, I did:
whatever reason it preferred the "HDMI 1" output, even though no monitor is connected; probably some VNC thing)
i've run the "humanCaller_test.pd" patch for about half an hour without any problems, using ALSA as the backend.
seeing that everything works, i've stopped Pd, launched jackd (again making sure to output to the headphones instead of the non-existant monitor) and tried with JACK as backend. after another 10 minutes I decided that everything works great.
gmadsr IOhannes