Linux

Here (archive.org) is an excellent explanation.

Run the following as root:

apk add cups cups-filter
# lpadmin is required to administer CUPS.
# I don't know whether lp is required.
adduser "$user" lp lpadmin # no relog required
rc-service start cupsd
# lp and lpr are now usable with either -d <dst> or a default destination

At first (2025-06-19), I didn’t realize I had to install cups-filter. I was unable to print a test page.

To make things quick, set a default destination. Then, for example, you can position point on a file in Dired and press P RET. Invoke the following with a suitable printer-name to set the default destination within Emacs for the duration of the session.

(setenv "LPDEST" printer-name)

Canon isn’t a make explicitly recognized by CUPS. I added a Canon printer as a “Generic/PCL Laser Printer” and everything was fine.

The builtin local service runs scripts in /etc/local.d/ when it starts and stops.

On Alpine, see file:///etc/local.d/README.

Thanks to Chuan Ji’s post about the fontconfig configuration. This worked immediately with PGTK Emacs on Wayland, which is promising.

Simply don’t set any default families in Emacs. I set only a size, and Emacs loaded the config-specified default font at the proper size.

Nmap does this if it has root permissions:

doas nmap -sn 10.0.0.0/24

Extract the IP for a given MAC using this shell invocation:

doas nmap -sn 10.0.0.0/24 | \
    grep -B 2 $mac | \
    sed -n '/report for/ s/^.*\(\([0-9]+\.\)\{3\}[0-9]+\)$/\1/p'

The console accepts escape sequences for this purpose. Use printf(1) with the %b specifier, which interprets escape codes in its argument.

Here’s an example with the ’everforest dark medium’ color palette:

printf %b '\e]P02f383e' \
       '\e]P1e67e80' \
       '\e]P2a7c080' \
       '\e]P3dbbc7f' \
       '\e]P47fbbb3' \
       '\e]P5d699b6' \
       '\e]P683c092' \
       '\e]P7d3c6aa' \
       '\e]P84b565c' \
       '\e]P9e67e80' \
       '\e]Paa7c080' \
       '\e]Pbdbbc7f' \
       '\e]Pc7fbbb3' \
       '\e]Pdd699b6' \
       '\e]Pe83c092' \
       '\e]Pfd3c6aa'

• light fonts from the ZAP group

[2025-10-11 Sat]

Phew, this one took a while to figure out.

1. Pick a TTY with no process attached, not even getty. Call it /dev/ttyN.

2. Change its permissions so audio can come through. I think it uses the XDG runtime directory to do this, but I’m really not sure:

   # chown <user> /dev/ttyN
   

   This is the only step requiring root privileges. You’ll probably have to do it after every boot.

   Or, if you’re on a single-user system and will do this often, invoke chown as a local.d script.

3. Start the process. MPV can run in a framebuffer because it’s based on mplayer2, which also can. I believe VLC has similar functionality. ffplay seems like the kind of thing that would also, but without good user interface. (Gotta draw the line somewhere.)

   $ openvt -s -c N mpv --vo=drm path/to/video.mp4
   

   • -s switches to /dev/ttyN. MPV displays on the screen regardless of which tty is focused. Therefore, switch to it with -s so input immediately applies to MPV. Otherwise, input will keep going to the focused (now obscured) tty and doing nothing to MPV.
   • -c N specifies /dev/ttyN as the tty to run the process on. In step (1), we chose /dev/ttyN because it was not the controlling tty for any process. If it had been, there would now be two processes listening for input on the same tty, and results would be unpredictable and probably not what you want.

The video should be running and audible! Use the familiar Alt+Fn, Alt+Left, Alt+Right to switch between ttys.

[2024-05-25 Sat] In an effort to get displays working through USB-C hubs on Enki, I’m installing the linux-firmware meta package, which depends on all firmware. That’s pretty large, apparently, but oh well. I’m removing from /etc/apk/world explicit dependencies on:

• linux-firmware-i915
• linux-firmware-intel
• linux-firmware-mediatek
• linux-firmware-other
• linux-firmware-rtl_bt

This is in response to a dmesg warning that some firmware failed to load. I don’t have the knowledge right now to discern anything else.

Sure enough, after rebooting, following dmesg, and plugging in the hub, I saw some unfamiliar interface drivers load. Still, though, the displays aren’t being detected through the USB-C hub.

[2025-12-28 Sun]

1. Install caps2esc. It pulls its required dependencies, including interception-tools, which it’s a module for.
2. Start the udevmon OpenRC service. This is provided by interception-tools.

3. Add the following snippet to /etc/interception/udevmon.yaml:

    - JOB: intercept -g $DEVNODE | caps2esc -m 1 | uinput -d $DEVNODE
   DEVICE:
     EVENTS:
       EV_KEY: [KEY_CAPSLOCK, KEY_ESC]
   

   See documentation of caps2esc for the meaning of the mode flag -m.

[2025-06-01 Sun 12:42] <- Groups: See

On the Alpine wiki (2024-01-08) the Sway page discourages adding users to the video group, it being unnecessary. I don’t know the reasoning.

Sections after “Steps” are in alphabet, not chronological, order.

Updating /etc/network/interfaces to read:

auto lo
iface lo inet loopback

auto eth0
iface eth0 inet static
	address 192.168.1.150/24

The address given appears on the Alpine Wiki. I’ve opted for no gateway, since I don’t quite know what that would mean. The intent is to simply ssh into the system after connecting the computers by ethernet cord.

It won’t be able to update itself, but that’s a problem for another time. I just updated it to Alpine 3.20, so it should be solid awhile.

This forum thread suggests adding a line to /etc/sysctl.conf:

net.ipv6.conf.all.disable_ipv6 = 1

I don’t know much about that file. Are these kernel parameters?

That seemed to work.

The same forum thread suggests that the net..._ipv6 key above will be unrecognizable unless the ipv6 module is loaded at init time. Fortunately, on both Enki and Thoth, /etc/modules contains an ipv6 line by default.

See Packages for Plasma. Uninstall these.

Shadow-login and pam-rundir should be enough to get going. DWM doesn’t need Seatd. Make sure to edit .xinitrc and .profile. Re-enable acpid. Switch rofi from wayland to X11.

See NetworkManager setup with KDE.

[2024-05-18 Sat 09:56] <- Setting fonts/Alpine linux

Give setfont a full filename argument.

Fonts are kept, at least, in /usr/share/consolefonts. If that directory contains, for example, ter-122b.psf.gz, invoke:

setfont ter-122b.psf.gz

For Steam games, this worked for me simply by plugging a Dualshock 3 into the USB port of Enki. It doesn’t do anything in dwm, but I assume that’s dependent on the graphical environment. For example, I’d guess GNOME would auto-detect the controller and allow some basic inputs with it. I wonder if any other apps would see it the same.

It’s especially nice that the Steam Flatpak detected the controller without any setup on my part.

These are packages that I need installed for some local application or another, but which don’t have any obvious dependents within APK.

• [ ] copy ssh and gpg keys
• [ ] copy wpa supplicant conf

The package dbus-x11 provides the executable dbus-launch, which I use in .xinitrc.

Install packages nextcloud-mysql, nextcloud-initscript.

Each of these requires nextcloud itself. I’m not entirely sure what nextcloud-initscript does.

[2024-09-29 Sun 07:55] <- On Alpine Linux

There’s an autostart patch for DWM.

~/.config/dwm/autostart.sh:

#!/bin/sh
echo 'started' "`date -Is`" >>~/.var/log/dwm-autostart
/usr/libexec/pipewire-launcher 2>&1 >~/.var/log/pipewire-launcher &
# if [ -z "$(pgrep syncthing)" ] ; then
#     syncthing --no-browser 2>&1 >~/.var/log/syncthing &
# fi
playerctld daemon &
dunst &
~/.fehbg &

[2025-07-09 Wed] I’ve successfully set up something I’ve wanted for a while.

On boot, all radio interfaces are rfkill’d, and eth0 is brought up on a hotplug basis. It doesn’t block getting to login quickly.

Here are the necessary configuration files:

auto lo
iface lo inet loopback

allow-hotplug eth0
iface eth0 inet dhcp

iface wlan0 inet dhcp

Some ifupdown implementations respect allow-hotplug as an alternative to auto that will bring up eth0 whenever an ethernet device (?) is detected. ifupdown-ng seems not to support it, so we need this:

#!/bin/sh
ifup eth0 >/dev/null 2>&1 &

Finally, the rfkill:

#!/bin/sh
rfkill block all &

And remember to enable everything:

# rc-update add local
# chmod +x /etc/local.d/50-ifup-eth0.start
# chmod +x /etc/local.d/50-rfkill.start

Follow OpenRC documentation.

I’m getting an error trying to manually start user.ty. Maybe that’s because I’m already logged in? I added user.ty to the default runlevel. Also, I can’t enable user services right now. “Runlevel sysinit doesn’t exist’”. I’ll try rebooting and seeing if these two things are resolved.

Also recall the official service script writing guide.

[2024-10-18 Fri 19:30] <- Minimal + SSH Debian

From Guix’s virtualization guide.

qemu-system-x86_64 \
    -nic user,model=virtio-net-pci \
    -enable-kvm -m 2048 \
    -device virtio-blk,drive=myhd \
    -drive if=none,file=${IMAGE},id=myhd

From Guix’s virtualization guide. Forward an address to the host. Make sure to start an ssh server on the guest.

Compared to Minimal invocation on x86₆₄:

-	-nic user,model=virtio-net-pci
+ -nic user,model=virtio-net-pci,hostfwd=tcp::10022-:22	

Yields (with qcow2 path hardcoded):

qemu-system-x86_64 \
    -nic user,model=virtio-net-pci,hostfwd=tcp::10022-:22 \
    -enable-kvm -m 2048 \
    -device virtio-blk,drive=myhd \
    -drive if=none,file='/home/ty/virtual-machines/images/debian.qcow2',id=myhd

Manage long-running processes in .profile, and launch graphical sessions directly from the command line.

if [ -z "`pgrep syncthing`" ] ; then
    syncthing --no-browser 1>>/tmp/ty-stdout-syncthing 2>>/tmp/ty-stderr-syncthing &
fi

Note the & asynchronicity. This differs from pipewire-launcher, which forks and exits quickly. I don’t think Syncthing provides its own forking flag or subcommand.

The Steam Flatpak demands a DISPLAY env var properly loaded into the DBus activation environment. I couldn’t figure out how to appease it while running a DBus session that persists across and without graphical environments. The simple solution:

dbus-run-session -- flatpak run org.valvesoftware.Steam

Audio connection under Pipewire seems totally fine even with the duplicate DBus sessions.

Installing elan and components through the Lean 4 extension seems to work well. Plus, components thus installed work even on musl-libc when installed within the VS Code Flatpak. At least, so far ([2024-10-26 Sat]).

Giving up on that. I had to restart the server constantly to see any information. Bad workflow.

On Void Linux now. Working fine.

As of Zotero 7.

Go to

Settings → General → File Renaming → Customize Filename Format...

and set the format to {{ citationKey }}. With BetterBibTex active and working, this is exactly what I want.

Got this error when launching the Flatpak out of the box: the /tmp directory is mounted with noexec, which may prevent some Minecraft versions from launching. Sure enough, I was getting crashes on all instances.

I added the following to the launcher’s top-level JVM arguments box:

-Djava.io.tmp="/home/ty/.local/tmp"

The error disappeared, and games launched to the menu screen. I haven’t tried playing yet.

The obvious next step is to try and mount ~/.local/tmp as an actual in-memory filesystem to keep things clean. Or, to mount /tmp as exec. I might be able to do that with /etc/fstab; a hot remount (below) didn’t work for me.

$ mount -o remount,exec /tmp