Troubleshooting launching xfce/any session on live ISO

EDIT: 2026-07-20T22:00:00Z
Remember, this topic was created 2026-07-09T22:00:00Z and only after a serious amount of local troubleshooting and building ISOs.

The statements made here were true for the local troubleshooting and ISOs.

The problem

It is a general issue with newer xfce builds, I am not sure what it is, but I am fairly certain it has something do with the configuration of lightdm.

  • The Gnome edition uses GDM - no issues
  • The KDE edition uses SDDM - no issues
  • Custom Plasma ISO with plasma login manager - no issues

I have begun pick-axing the installation process in the hope that I find it.

I have a lot of thoughts - but no solutions.

The fact that you can boot the system or not - if not you can leave the system at the error message for a short period of time which makes it settle and you can login - makes me believe there is a race condition possibly with the the systemd logind.

I have found a script which which - according to the script - should fix a panel error.

It is included with the settings profile and lives in the users autostart folder but I can find no documentation on this - why it is there and what it is supposed to fix.

script in ~/.config/autostart resetting xfsettings
#/bin/sh
sleep 5
xfsettingsd --replace &
sleep 15
xfsettingsd --replace &
sleep 25
xfsettingsd --replace &

In the livefs - I have found that the xfce installation lacks certain pam configurations for lightdm whereas there is configurations for plasmalogin and kde.

The configuration in the livefs allows pam to accept a login if the user is member of the wheel group.

But if the manjaro-live script is racing the login daemon it is a hit or miss if the live user has been created or not where the latter will trigger the error message. This speculation is somewhat backed by the above mentioned fact - if you leave the error message long enough you can acknowledge the error - get to the login screen and login with the default manjaro password.

Looking at the Cinnamon edition - it uses lightdm as well - with slick-greeter instead of gtk-greeter and the Cinnamon edition does not exhibit the same behaviour when built against the current stable repo.

I am going to compare the Cinnamon and Xfce iso profiles - see if I can learn something from that comparison.


Perhaps @philm can provide some insight into this peculiar issue?

Please @philm help us understand… and more importantly we need to fix it… possibly in the manjaro-tools-livecd repository

EDIT 2026-07-10T05:13:00Z

I have come a bit closer - if I wait a couple of minutes, then acknowledge the error - the greeter comes up - only have to hit enter - partly autologin - only full autologin is now missing.

In my testing repo I have added pam configuration for lightdm to the livefs overlay.

Searching for the error reveals it is a somewhat common error.

https://sx.nix.dk/search?q=xfce lightdm unable to contact settings server

One topic on Gentoo 6 years ago suggested that the dbus was not build for X, which makes me think - I vaguely recall a package named dbus-x11 - but that does not exist anymore.

When the system fail and error out - the following is written to the file .xsession-errors in the live users home

Failed to connect to user scope bus via local transport: $DBUS_SESSION_BUS_ADDRESS and $XDG_RUNTIME_DIR not defined (consider using --machine=<user>@.host --user to connect to bus of other user)
dbus-update-activation-environment: error: unable to connect to D-Bus: Using X11 for dbus-daemon autolaunch was disabled at compile time, set your DBUS_SESSION_BUS_ADDRESS instead
dbus-update-activation-environment: error: unable to connect to D-Bus: Using X11 for dbus-daemon autolaunch was disabled at compile time, set your DBUS_SESSION_BUS_ADDRESS instead
Failed to connect to user scope bus via local transport: $DBUS_SESSION_BUS_ADDRESS and $XDG_RUNTIME_DIR not defined (consider using --machine=<user>@.host --user to connect to bus of other user)
dbus-update-activation-environment: error: unable to connect to D-Bus: Using X11 for dbus-daemon autolaunch was disabled at compile time, set your DBUS_SESSION_BUS_ADDRESS instead

Looking at the error message I am thinking - the root cause is

Failed to connect to user scope bus via local transport: $DBUS_SESSION_BUS_ADDRESS and $XDG_RUNTIME_DIR not defined (consider using --machine=<user>@.host --user to connect to bus of other user)

The rest is just messages caused by the first error.

This leads me to [SOLVED] Failed to connect to bus: $XDG_RUNTIME_DIR not defined / Newbie Corner / Arch Linux Forums

This does not get me much closer but I do now understand that we need to address the missing variables $DBUS_SESSION_BUS_ADDRESS and $XDG_RUNTIME_DIR somehow.

I have tried to add a condition to the manjaro-live.service

Before=systemd-user-sessions.service systemd-logind.service

it is a race condition somewhere

Even with the above changes - the pam configs and the extended condition - if I reboot the same stick it may occasionally boot directly to desktop - if I acknowledge the error the greeter will reload and after waiting a short while, hit enter or click login and desktop loads.

I will have to check if the xfsettingsd reload script is causing this.

I have been back and forth on what to keep and I have decided to stay as close to the original xfce ISO.

Which gives me a simple command which stores the build result in /miso folder

build-bobs-iso micro.xfce

I have no idea or maybe, perhaps…

While I went over the list in Packages-Desktop comparing to Packages-Mhwd I noticed a set of doubled packages

libva-intel-driver
libva-mesa-driver

I removed those from Packages-Desktop as they were already present in Packages-Mhwd, the thought was - if mhwd selects and installs the correct driver at install time - they should be included if they are required.

I also tagged nvidia-prime as belonging to >nonfree_x86_64 because it depends on nvidia-settings which is tagged with >nonfree_x86_64.

I will check with my bare-metal test but it is Intel only.

The test system was correctly identified by mhwd to be a candidate for video-linux.

I don’t have other hardware that qualifies for testing.

3 Likes

@linux-aarhus On Clevo optimus:
I put manjaro-xfce-26.0.4-minimal-260327-linux618.iso (from where I installed my current system) and the micro.xfce.iso on a ventoy drive.

manjaro-xfce-26.0.4-minimal-260327-linux618.iso

  • boots straight to desktop every time
  • mhwd -li shows video-linux version 2024.05.06

micro.xfce

  • boots to black screen every time (no popup)
  • editing grub line by replacing nouveau, i915 and radeon-modeset=1 with nomodeset boots to ā€˜unable to connect settings server’ popup, 3 min wait gets me to login and eventually desktop but resolution 1024x758 and mhwd -li shows video-modesetting

manjaro-xfce-26.0.4-minimal-260327-linux618.iso.pkgs

If you have a new iso just upload and I’ll test it.
I’m honored that you think I know how to use buildiso… I once made an Awsome install with architect, though, so there’s hope.

3 Likes

Would it be possible to replace lightdm with ly (console based, doesn’t require systemd) in order to test this?

I have just now built one using sddm - I am about to test it.

3 Likes

Thank you, I guess the xfce bug will be found.

Want to install micro.xfce on my macbook. I currently have manjaro-xfce-26.04-minimal-260327-linux66 on it that booted with the popup but runs and boots fine, no more popups.

I’d add b43-firmware and conky and maybe swap lightdm for ly and follow your ā€˜very advanced’ tutorial to make a fool of myself an iso from the running system. Would be interesting to see if the bug survives.

Here I have a question; It’s boiling hot right now; how hard is the process on the CPU? If it’s anything like compiling a kernel I may have to find a cooler place to do that.

1 Like

You don’t need to follow any tutorials — advanced or otherwise — in order to make a fool of yourself. There are plenty of people here perfectly capable of making fools of themselves without having looked at any tutorial at all. :smiling_face_with_horns:

4 Likes

Trust the voice of experience! :winking_face_with_tongue:

4 Likes

Indeed. I have mistakenly wiped a partition, a live web server (tip: never use drag’n’drop with the mouse for important directions) and a gigabytes of very important documents over the years. First 2 i could recover, the third one was gone forever (that was the backup of already lost data :joy:).

I am back to square one.

With the attempt to create a micro ISO with sddm - I faced the same issue.

So I am back to start.

Does it mean the troublemaker is not lightdm then?

Yes - I think the problem lies within manjaro-live.service.

It is set to run before systemd-user-sessions.service. From what I have been able to gather this is correct. But for some reason the Before=systemd-user-sessions.service directive in manjaro-live.service is not honored.

The service executes the /usr/bin/manjaro-live script, which creates the live user and the properties needed to be able to autologin

At the time the display manager tries to autologin, the user has not yet been created and therefore the autologin process fails.

Both sddm and lightdm generates the same error message at the first attempt of autologin

Failed to connect to user scope bus via local transport: $DBUS_SESSION_BUS_ADDRESS and $XDG_RUNTIME_DIR not defined (consider using --machine=<user>@.host --user to connect to bus of other user)
dbus-update-activation-environment: error: unable to connect to D-Bus: Using X11 for dbus-daemon autolaunch was disabled at compile time, set your DBUS_SESSION_BUS_ADDRESS instead
dbus-update-activation-environment: error: unable to connect to D-Bus: Using X11 for dbus-daemon autolaunch was disabled at compile time, set your DBUS_SESSION_BUS_ADDRESS instead

I have built yet another xfce ISO.

I have booted the ISO straight to desktop on bare-metal and virtual (qemu/kvm)

But I have no idea how it will work on other hardware.

Please test and let me know.

Signed and with checksum https://manjaro.dk/iso/micro

2 Likes

manjaro-micro.xfce-194.05-260713-linux618.iso
Physical machine: full Intel IGP LENOVO ThinkPad T480 … Ventoy
checksum → OK
Unable to contact settings server
Could not connect: No such file or directory

1 Like

Can someone evaluate if this issue from 2023 is related?
Unable to contact Settings Ā· Issue #2646 Ā· neutrinolabs/xrdp Ā· GitHub

The culprit seems to lie somewhere in the systemd service interaction and the oneshot manjaro-live.service which in turn runs the manjaro-live shell script

It appears the system does not recognise the changes made by the script.

I have tried

  • adding Before=display-manager.service to ensure display manager is delayed until
  • setting DBUS_SESSION_BUS_ADDRESS and XDG_RUNTIME_DIR
  • created a custom .xinitrc simplifying the init on the live session by hardcoding the expected variables
    export DBUS_SESSION_BUS_ADDRESS=unix:path=/run/user/1000/bus
    export XDG_RUNTIME_DIR=/run/user/1000
     
    exec dbus-launch --sh-syntax --exit-with-session xfce4-session
    
  • tried manually enable and start user@1000.service

I can make the ISO load on bare-metal and virtual - but after a few successful loading of the desktop - the ISO stops working - yes - that is strange, very, very strange.

This makes me reason that the issue is the timing between the units that is loading the final desktop

  • when the loading fails
    • the user has been created
    • no session bus has been created at socket://run/user/1000/bus
  • when the loading succeeds
    • the user has been created
    • the session bus has been created at socket://run/user/1000/bus

When the desktop loading fails - the system eventually settles, the dbus session is created and it is possible to load the live desktop.

I think you are on the right track on this..

I installed gnome-logs on an ISO and in live mode, I got the following:

Notice 17:50:34 "Failed to resolve user ā€˜1000’ , the user does not yet exist at this point.

17:50:13 kernel: RDSEED32 is broken. Disabling the corresponding CPUID bit.
17:50:13 kernel: virt/tdx: TDX not supported by the host platform
17:50:13 kernel: EXT4-fs (vda): VFS: Can't find ext4 filesystem
17:50:19 systemd-udevd: /usr/lib/udev/rules.d/69-cd-sensors.rules:105 Failed to resolve group 'colord', ignoring: Unknown group
17:50:20 Xorg: (==) Using system config directory "/usr/share/X11/xorg.conf.d"
17:50:27 lightdm: gkr-pam: unable to locate daemon control file
17:50:28 systemd: Failed to start User Runtime Directory /run/user/1000.
17:50:28 lightdm: gkr-pam: unable to locate daemon control file
17:50:28 cinnamon-sessio: CRITICAL: t+0.00286s: Unable to start session: Lost name on bus: org.gnome.SessionManager
17:50:28 Xorg: (==) Using system config directory "/usr/share/X11/xorg.conf.d"
17:50:31 lightdm: gkr-pam: unable to locate daemon control file
17:50:32 cinnamon-sessio: CRITICAL: t+0.00264s: Unable to start session: Lost name on bus: org.gnome.SessionManager
17:50:32 Xorg: (==) Using system config directory "/usr/share/X11/xorg.conf.d"
17:50:32 lightdm: pam_systemd(lightdm-greeter:session): Varlink call io.systemd.Login.CreateSession failed: io.systemd.System
17:50:34 lightdm: gkr-pam: unable to locate daemon control file
17:50:34 lightdm: gkr-pam: unable to locate daemon control file
17:50:34 systemd-user-ru: Failed to resolve user '1000': Unknown user
17:50:34 systemd: Failed to start User Runtime Directory /run/user/1000.
17:50:34 systemd: Failed to start User Runtime Directory /run/user/1000.
17:50:34 lightdm: pam_systemd(lightdm:session): Varlink call io.systemd.Login.CreateSession failed: io.systemd.Login.UnitAllocationFailed
17:50:34 lightdm: gkr-pam: unable to locate daemon control file
17:50:34 lightdm: gkr-pam: unable to locate daemon control file
17:50:35 cinnamon-sessio: CRITICAL: t+0.00254s: Unable to start session: Lost name on bus: org.gnome.SessionManager
17:50:35 Xorg: (==) Using system config directory "/usr/share/X11/xorg.conf.d"
17:50:37 lightdm: gkr-pam: unable to locate daemon control file
17:50:38 cinnamon-sessio: CRITICAL: t+0.00294s: Unable to start session: Lost name on bus: org.gnome.SessionManager
17:50:38 Xorg: (==) Using system config directory "/usr/share/X11/xorg.conf.d"
17:50:41 lightdm: gkr-pam: unable to locate daemon control file
17:50:43 spice-vdagentd: An agent is already connected for this session

I this thread I posted that manjaro-live.log was showing the root user being configured late int he events, maybe there is an issue there after all?

There has been no changes to the tools for a long time.

I have dissected the initialisation logic and there is nothing indicating an issue with the the initialisation logic.

The live user - which you can change in the manjaro-tools.conf - is generated when the ISO starts.

The script is run by a service manjaro-live.service and it is the output from that script you see as /var/log/manjaro-live.log.

The configuration of the root user is irrelevant.

The issue - at least as far as I could track it - is that the dbus socket://run/user/1000/bus sometimes fail to get created.

It is a hit or miss scenario - but why? - what is the cause? - as the issue is not fully reproducible - what is the trigger or missing trigger?

The evidence is clear when you switch a second tty - login using the default manjaro credentials - then list the content of /run/user/ - when it fails there is no folder for UID 1000.

The fact that you can login using the default manjaro credentials proves that the manjaro-live script work as intended.

When it fails I have also checked with

systemd-analyse critical-chain

and

systemd-analyze blame

And the dbus service is nowhere to be found

So I am thinking it has something to do with the dbus as it fails to activate and therefore does not create a socket for the UID 1000.

4 Likes

Hello @linux-aarhus,

I’ve been following this thread with great interest, as I’m experiencing the exact same autologin failure on a GNOME live ISO – but interestingly, not on my XFCE build from the very same day.

On GNOME, after the boot finishes I’m presented with the login screen and have to type the manjaro password manually. Additionally, even if I select the Hungarian keyboard layout in GRUB, the live session completely ignores it and falls back to US English.

However, my XFCE ISO built on 10 July 2026 (the day you opened this topic) does not exhibit this behaviour – autologin works as expected.
Build details of that XFCE ISO:

  • :package: Date: 2026-07-10 14:26 UTC
  • Version: 2026-07-Respin
  • Branch: Manjaro TESTING
  • Kernel: linux71

The ISOs can be found here:

The corresponding GitHub Actions workflows are here ( Mind you! All of my actions located in master branch, but both building from different branches.):

This makes me suspect the issue may not be a date-related repository change, but rather something in the live session startup that affects GNOME (GDM) and XFCE (LightDM) differently – perhaps the race condition you’ve already identified.

A question for you and others affected: when you select a non-US keyboard layout in the GRUB menu (e.g. French, German, Hungarian), does it actually get picked up in the live desktop, or does it silently revert to US?
In my GNOME case the layout is lost entirely, which might be a related symptom of the D‑Bus/session bus problem.

If it helps, I can trigger a fresh build of the XFCE ISO right now, so we can see if the situation has changed in the last few days…

Thank you for your detailed investigation – the $XDG_RUNTIME_DIR / $DBUS_SESSION_BUS_ADDRESS findings are very illuminating.

Best regards

1 Like

Did we try:

[Unit]
Description=LiveMedia MHWD Script
Before=basic.target
After=local-fs.target sysinit.target
DefaultDependencies=no

[Service]
Type=oneshot
ExecStart=/usr/bin/mhwd-live

[Install]
WantedBy=basic.target