From the Beowulf release notes. I just tested the second method in daedalus, and it still works.

I think someone would have to get brave and clone the heads git repo to build it with live-sdk.

The right way to do it would have been to un-check xfce and check kde at the tasksel window. Then you would end up with kde, sddm, elogind instead of consolekit and slim. Installing two desktops can be tricky, because some of them have conflicting dependencies.

You can probably fix it by installing elogind and libpam elogind and getting rid of consolekit, slim and any parts of xfce that want to go with it. It might take a few tries and it might be faster to start over.

Maybe

aptitude purge task-xfce-desktop
aptitude reinstall task-kde-desktop

or the equivalent with apt or apt-get if you prefer.

You are correct that it is possible in Expert install to install the system without creating a root password or a user account. This can be done by using the arrow keys to skip over that task in the menu. If you just press ENTER every time you get to the menu, it automatically takes you to the next item, and you won't miss anything.

About the menu items "Install" and "Expert install": That's the way the installer has been since Debian Etch. Changing the words would probably confuse thousands of people.

If you are unfortunate enough to boot to a system that you installed without root password or user account, there is a simple fix:
- At the boot menu, choose Advanced options and then boot into Rescue mode.
- Press ENTER to get a root console.
- Set a root password and create a user account with the following commands and on-screen instructions:

passwd
adduser <your-login-name>

If you don't want to wait for packages to download during the installation, use the DVD and do not select a mirror. The install will only take a few minutes. I can sympathize with you on this point. The netinstall is too slow for my taste, and I have 100Mb/s down speed here.

If you don't want the installer to stop and ask you questions, you can do an Automatic install. It's in the Advanced options of the installer boot menu, and there are other things you must do for it to work. I don't know the details about it, but that's what people who need to install dozens or more of systems use.

The primary goal of Devuan has always been to provide Debian without systemd. We try to change as little as possible, and we encourage people to make derivative distros if they want to provide something different.

If you want to make a live-iso from a running live-iso without installing it to hard drive, you can do it. You must edit the config file to set the work_dir and snapshot_dir to some storage device, like a mounted partition on a hard disk or a usb stick. You could also do it by booting the .iso file in a virtual machine and setting the work_dir and snapshot_dir to be on a virtual disk.

Don't boot the system to RAM unless you have a lot of extra. You'll need it fo update the package cache and to hold any changes or additions you make to the system.

And you'll have to make all those changes again if you reboot. You already discovered that. I find it easier to install the system on a virtual disk and making the changes there. Then I can make a new snapshot iso any time I want with minimal effort.

https://refracta.org/docs/readme.refractasnapshot.txt

Whenever debian testing goes into freeze is when I start making isos for the next refracta release. Last couple of releases I was late getting started. Looks like this time I'm early. Here are some xfce isos based on devuan daedalus. I started with the refracta-12 no-X isos and added the usual software. I checked the package list against refracta-11, and it looks like everything is there.

I have no idea what works or does not work. Well, I know they boot (in qemu) and get an ethernet connection. And the browsers work. (ff-esr and tor) But this is way early, and something must be broken. If not, wait a week and upgrade, and maybe something will break.

Ceres is also in sources.list and pinned to a priority of 10 in case you want to get something sooner. OK, it's pinned in both isos, but the ceres line didn't make it into the amd4's sources.list. Next time...

Have fun, and don't fall off the edge.
https://get.refracta.org/files/testing/

To fix that...

gpg --keyserver keyserver.ubuntu.com  --refresh-keys

pub   rsa4096 2017-10-07 [SC] [expires: 2022-04-26]
      67F5013216271E85C251E480A73823D3094C5620
uid           [ultimate] fsmithred (aka fsr) <fsmithred@gmail.com>

The following things did not work:
- boot installer iso in rescue mode and reinstall grub
- boot installer iso in rescue mode, open a shell in the installed system and run and or all of the following in any order, both before and after removing grub-efi-amd64-signed.
    dpkg-reconfigure grub-efi-amd64
    grub-install
    grub-install --bootloader-id=devuan
    update-initramfs -u
    update-initramfs -u -k 5.15.0-1-amd64
- boot installer iso in rescue mode and install grub to removable media path

This did not create /boot/efi/EFI/devuan

grub-install --bootloader-id=devuan

Last commands in root's history I executed in a chroot before it started working:

cd /boot/efi/EFI
cp -r debian devuan
grub-install --bootloader-id=devuan
dpkg-reconfigure grub-efi-amd64 (and choose yes to force removable media path
update-initramfs -u -k 5.15.0-1-amd64

Edit: I keep forgetting to mention that I can't get qemu to boot from the iso when it's attached to the disk image file.  This does work for me with any of my other disk image files. I have to boot the iso from grub command line.

Confirmed. I got the same behavior in qemu. Chrooting the installed system from a live-iso and installing cryptsetup and cryptsetup-initramfs solved it.

Note: I used the server iso and installed without mirror. cryptsetup (and cryptsetup-bin, I think) is in the iso, but it didn't get installed in the system, even though I chose encrypted lvm, but cryptsetup-initramfs was missing.

Fixed in ceres and will move into daedalus soon.
https://bugs.devuan.org/cgi/bugreport.cgi?bug=635

Here are the main changes. I also changed calls for update-notifier.sh to update-notifier2.sh instead. I'm too lazy to see which files they were in. It makes more sense to change the name of the script (drop the "2" since there's only one of them.)

Note the change in the she-bang. We may get stuck with usr-merge in the not too distant future, and /bin/bash will go away. It won't go far, but using '/usr/bin/env bash' will survive the change.

/etc/default/update-notifier

# Set the default action for clicking on the notification icon.
# Valid choices: 
# "none" to only see a list of available upgrades.
# "terminal" to run the upgrade in a terminal
# "gui" to run the upgrade in synaptic package manager.

FRONTEND="terminal"

# Change the terminal if you don't have xterm.
# End your terminal command with the option that runs a command inside the terminal.
# Example:
# TERMINAL_COMMAND="/usr/bin/xfce4-terminal"
# TERMINAL_OPTIONS="-e"

TERMINAL_COMMAND="/usr/bin/xterm"
TERMINAL_OPTIONS="-fa mono -fs 12 -e"

/usr/local/bin/upgrade.sh

#!/usr/bin/env bash

source /etc/default/update-notifier

killall yad

if [ "$FRONTEND" = terminal ] ; then
	pkexec env DISPLAY=$DISPLAY XAUTHORITY=$XAUTHORITY "${TERMINAL_COMMAND} ${TERMINAL_OPTIONS}" 'apt-get dist-upgrade'
elif [ "$FRONTEND" = gui ] ; then
	synaptic-pkexec --dist-upgrade-mode --non-interactive
fi

exit 0

/usr/local/bin/update-notifier2.sh

#!/usr/bin/env bash

list=$(apt-get -s dist-upgrade | awk '/^Inst/ { print $2 }')

count=$(echo "$list" | wc -l)

if [ -n "$list" ] ; then

	notify-send --expire-time=20000 --icon="/usr/share/pixmaps/updates.svg" \
"Available updates: $count

$list"

	yad --notification \
		--image="/usr/share/pixmaps/updates.svg" \
		--icon-size=22 \
		--command="/usr/local/bin/upgrade.sh" \
		--text="Available updates: $count
	_____________________
	
$list" &

fi

exit 0

Done.