You are not logged in.
You have to delete 'user' to make noautologin work.
No, I don't have to delete 'user', and noautologin does work with or without persistence. I've been doing it this way for about 15 years. Before devuan existed I was doing this with debian. I guess we're not doing the same thing.
I use this to make a live-usb: https://get.refracta.org/files/tools/re … ~4_all.deb
Ed. Ahh, I see I have reached the perfect post count. big_smile
I think I have a better understanding of the phrase, "The devil is in the details." ![]()
I don't get the same results as you. The noautologin option is working correctly here on two different live-isos.
The devuan desktop-live iso has user 'devuan' pre-defined, and with noautologin, boot stops at the slim login screen.
A refracta live iso has user 'user' pre-defined, and with noautologin, boot stops at the lightdm login screen.
Maybe I should call it a pre-existing user instead. That user already exists and has a home directory inside the squashfs as opposed to being created when the system boots.
See 'man live-build' and 'man live-config' for details about the options for the live system.
The default user already exists in devuan live isos. In debian live isos, the default user is created during boot if the option 'config' is included in the command line. I think if you remove that, you'll have the default user again.
The failsafe option in the boot menu has these:
noapic noapm nodma nomce nolapic nosmp nomodeset vga=normalAnd I noticed that in the grub menu created by live-sdk, the failsafe entry is missing 'nomodeset'. It's only in the syslinux menu. In refractasnapshot, 'nomodeset' is in the failsafe entries in both menus.
@tyder, see if you can narrow down which kernel parameters make a difference. TAB or e to edit the boot entry on the fly.
Here's mine.
Booted from usb on hardware. It works here.
inxi -G
Graphics:
Device-1: Intel HD Graphics 5500 driver: i915 v: kernel
Device-2: Chicony Integrated Camera driver: uvcvideo type: USB
Display: x11 server: X.org v: 1.21.1.16 driver: X: loaded: modesetting unloaded: fbdev,vesa
dri: iris gpu: i915 tty: 121x24 resolution: 1600x900
API: EGL v: 1.5 drivers: iris,swrast platforms: gbm,surfaceless,device
API: OpenGL v: 4.6 compat-v: 4.5 vendor: mesa v: 25.0.7-2 note: console (EGL sourced)
renderer: Mesa Intel HD Graphics 5500 (BDW GT2), llvmpipe (LLVM 19.1.7 256 bits)
Info: Tools: api: eglinfo,glxinfo de: xfce4-display-settings x11: xdriinfo, xdpyinfo, xprop,
xrandrfailsafe is just a name for a choice in the boot menu. Look at the boot command for that choice and you will see some kernel paramaters. One of them is 'nomodeset' which is sometimes useful when normal boot options result in a black screen.
What video cards are you using? I don't think you mentioned that.
This one (dzz's third example) shows me the boot menu and goes to black screen.
qemu-system-x86_64 -enable-kvm -m 2048 -bios /usr/share/ovmf/OVMF.fd -boot d -cdrom devuan_excalibur_6.1.1_amd64_desktop-live.isoI can get to serial console in qemu and log in. Slim is not running, xorg is running. I can kill xorg, then log into console as root and run startx. Desktop comes up. Alternatively, I can drop to console, install lightdm, start lightdm, get graphical login and desktop comes up.
This one (abbreviated fourth example - no network settings) works from boot menu to desktop.
qemu-system-x86_64 -machine accel=kvm -m 1924 -boot d -cdrom devuan_excalibur_6.1.1_amd64_desktop-live.iso -bios /usr/share/ovmf/OVMF.fd -boot once=d,menu=offOk, I'm curious to know if the excalibur-nonfree iso that I linked above works.
You could build an iso using live-sdk, but for a one-off, it's easier to install what you want in a VM and then run refractasnapshot to make the live iso. With live-sdk, if you list a package to be installed that doesn't exist, then the build fails and you have to start over again. If you're working in a VM and planning to use refractasnapshot, if you try to install a package that doesn't exist, that one apt command fails, and you still have whatever work you've done building up the system.
Here are instructions for live-sdk:
https://dev1galaxy.org/viewtopic.php?id=551
Here are instructions for refractasnapshot:
https://git.devuan.org/devuan/refractas … apshot.txt
Can you get different results using my qemu commands?
All the major desktops are available in the installer isos.
Some more detail about exactly what you tried with which iso would help. Also, if you can find any of these other relevant discussions that you mentioned and provide the links, I would greatly appreciate it. I'm doing 15 different things and couldn't find them.
I have no trouble booting the isos in qemu or virtualbox in bios mode, but I have mixed results booting them in uefi mode. I don't know the reason for that. Gathering all the evidence in once place would be a good start.
For example, this works:
qemu-system-x86_64 -enable-kvm -m 2048 devuan_excalibur_6.1.1_amd64_desktop-live.isoBut trying to boot that iso in qemu uefi fails.
Meanwhile, this one works in uefi even though it was made the same way as the excalibur iso.
qemu-system-x86_64 -enable-kvm -m 2048 -bios /usr/share/ovmf/OVMF.fd devuan_freia_7-preview-2026-01-12_amd64_desktop-live.isoAnd this one works (excalibur):
qemu-system-x86_64 -enable-kvm -m 2048 -bios /usr/share/ovmf/OVMF.fd refracta-isos/13.2/refracta_13.2_xfce_amd64-20251205_1133.isoEdit: This one that I made for a wifi test yesterday works in uefi, too. It's just the default devuan excalibur desktop install that was created with refractasnapshot instead of live-sdk. The uefi part of the build is the same for both methods, so I'm at a loss for an explanation.
https://get.refracta.org/files/experimental/excalibur_nonfree-20260224_2014.isoInstall ntpsec-ntpdate and the command ntpdate-debian will work again.
Please try this and let me know if it works. There's a gpg sig file in the same directory.
https://get.refracta.org/files/experime … 4_2014.iso
sha256sum:
31c2849c144cf99b742cce27e1781baa2d6df935f0b3bd86bee14e7545008a59 excalibur_nonfree-20260224_2014.isoSome of the broadcoms are left out because they require the user to agree to certain conditions. I thought this was one of those, but now I'm not so sure. I might try making a live iso with broadcom-sta-dkms, and if it works, I can upload it for you to use. Instructions for installing all the broadcom drivers can be found at wiki.debian.org, and they might require that you have an internet connection to get what you need to have an internet connection.
Easiest way to get what you need is if you can use an ethernet cable. You need to add non-free to the active lines in sources.list.
Alternatively, you could download the packages in windows and then install them when you boot into devuan. Here's the list of what you need:
# apt install broadcom-sta-dkms linux-headers-$(uname -r)
Installing:
broadcom-sta-dkms linux-headers-6.12.57+deb13-amd64
Installing dependencies:
binutils libalgorithm-diff-perl libitm1
binutils-common libalgorithm-diff-xs-perl liblsan0
binutils-x86-64-linux-gnu libalgorithm-merge-perl libquadmath0
build-essential libasan8 libsframe1
dkms libbinutils libstdc++-14-dev
dpkg-dev libc-dev-bin libtsan2
fakeroot libc6-dev libubsan1
g++ libcc1-0 linux-headers-6.12.57+deb13-common
g++-14 libcrypt-dev linux-kbuild-6.12.57+deb13
g++-14-x86-64-linux-gnu libctf-nobfd0 linux-libc-dev
g++-x86-64-linux-gnu libctf0 make
gcc libfakeroot manpages
gcc-14 libgcc-14-dev manpages-dev
gcc-14-x86-64-linux-gnu libgprofng0 pahole
gcc-x86-64-linux-gnu libhwasan0 rpcsvc-protoEdit:
I installed into a VM and was able to install broadcom-sta-dkms without having to agree to a license. I'm surprised. I'm making a new iso now that you can test. I don't have hardware for that.
sources.list will still be used if you don't "modernize" the sources when apt asks you. It will also be used if you change your mind and move the modernized files out of the way and replace your sources.list file.
I fixed the broken link. Here's the right one.
Devarch, you are mistaken. I did not remove support for FDE. You have to check a box (in the gui version) or say 'yes' (in the cli version) to encrypt the root partition. If /boot is in that partition, the installer will add the appropriate line to /etc/default/grub.
You might look at the Help in the installer or the readme in /usr/share/doc/refractainstaller-base. You only need to read four short lines to get to the answer.
STARTING REFRACTA INSTALLER (9.5.x)
*** NEW in 9.4 ***
- UEFI and BIOS installers have been merged (both gui and cli scripts)
- Installer supports gpt disk with bios boot. (special partition needed)
- Installer supports full-disk encryption (no separate /boot partition)Note: If you want to get into playing with chroot and doing some manual tasks, you can use the cli installer to do lvm or raid installs, which aren't overtly supported in refractainstaller.
Note2: If you want to tell me that it's not proper to have /boot inside the root partition, you'll need to explain why. The reason for having it outside the encrypted partition all these years was so the bootloader could find the kernel. Well, now grub can find it in an encrypted partition.
ps_mem.py shows:
# old thinkpad with 8G ram running daedalus, xfce
7.1 MiB + 1.1 MiB = 8.2 MiB NetworkManager
15.4 MiB + 3.2 MiB = 18.6 MiB nm-applet
# satellite with 8G ram running excalibur, xfce
17.5 MiB + 1.4 MiB = 18.9 MiB wicd
24.9 MiB + 3.7 MiB = 28.6 MiB wicd-client
# newer old thinkpad with 8G ram running excalibur, lxqt
8.0 MiB + 814.5 KiB = 8.8 MiB NetworkManager
11.1 MiB + 4.3 MiB = 15.5 MiB nm-traygrub-pc-bin was there. I normally include that and even the ia32-bin package, just in case someone needs it. (old macbook pro)
I thought you need to have the right package, grub-pc or grub-efi-amd64 to install the bootloader correctly, so I always make sure I have the right one before any grub-install happens. Two days ago, I did a bios install with grub-efi-amd64 installed and I rebooted to a grub rescue prompt. But that was with a misconfigured calamares, so maybe it doesn't count.
Looking at the files that grub-efi-amd64 provides, I have no idea what it does. Looks like nothing. The script gathers information for a bug report.
$ apt-file list grub-efi-amd64
grub-efi-amd64: /usr/share/bug/grub-efi-amd64/presubj
grub-efi-amd64: /usr/share/bug/grub-efi-amd64/script
grub-efi-amd64: /usr/share/doc/grub-efi-amd64FYI, refractainstaller had FDE before debian-installer did.
In my last build, I copied all of the user's home to /etc/skel and uncommented the appropriate line in the modules. It worked - it kept all the user configs and changed the user name.
And no, I didn't run the scripts. I'll have to try a no-network install to see what happens.
You're right. I just checked a beowulf live-iso and wicd disconnects the active connection before it completes the new connection. That's consistent when going from wired to wireless and from wireless to wired.
I did the test with two routers in the house to see if I could be connected to two different networks. I can't do that with wicd. I can do it with network-manager - I could ping both routers. Maybe I'll make a wishlist bug report for that some time. Meanwhile, I'm looking forward to seeing more test reports from people.
"Try it, you'll like it."
Installs and works on excalibur. That's a sight for sore eyes. It's like seeing an old friend I'd forgotten about.
As mentioned in IRC, there's an extra entry in the apps menu. Other annoyance is that it disconnects the wired interface when you connect to wireless. I don't remember if the old wicd did that. N-M lets me have both interfaces up at the same time.
@rations:
Yesterday I tried again using your modules and configs instead of my minimally edited debian files. It works. (replaced the user with what was entered in the installer)
One point (not the only) that I'm confused about is grub. I changed one line in the bootloader module so that it would boot efi or bios, and I'm wondering how it does that. Is a network connection required for that to succeed? My iso has grub-efi-amd64 installed, but I did a bios install and it got grub-pc from somewhere. (either network or it found the deb package that I put in the root of the squashed filesystem)
I'm going to keep playing with this, but it's probably going to go slowly. I have too much other stuff to do. Thanks for this contribution. I now have less hate for calamares. Maybe there can eventually be a calamares-settings-devuan package.