at-spi-bus-launcher segfaults

girzel · 2022-04-22 15:50:35

seth wrote:

Can you manually start the service?
systemctl --user start gvfs-daemon
systemctl --user status gvfs-daemon
If yes, as a workaround you could simply enable the service,
systemctl --user enable gvfs-daemon

No, using systemctl --user to start it just times out. I don't understand this. If I manually run the gvfsd command with --replace, I get:

[eric@clem ~]$ pgrep -l gv
281435 gvfsd
281440 gvfsd-fuse

I kill that, run the systemctl command instead, and get the same "pgrep" output, but then after a while the systemctl call times out. If I manually edit the gvfs-daemon.service file and add a "--replace" in there, same result.

girzel · 2022-04-22 16:15:37

I tried just uninstalling gvfs -- surprisingly, the only package that required it was gimp. I rebooted, and everything works perfectly: geeqie, Libreoffice, the PA volume panel, the GUI file-chooser for all applications -- all start instantly.

It's a bit of a shame to lose gimp, and I guess I'd still like to know what the "real" problem is, but for now it's working. Maybe next time I need gimp I'll re-install it, and see if everything magically starts working again: the system-software-level equivalent of turning it off and turning it back on again.

Edit: Or, if you think we stand a chance of actually figuring this out, and it might be helpful to future bbs-browsers, I'm happy to re-install now and see what happens.

Last edited by girzel (2022-04-22 16:28:45)

seth · 2022-04-22 16:31:56

Can you "systemctl --user list-unit-files" ?

girzel · 2022-04-22 18:31:20

seth wrote:

Can you "systemctl --user list-unit-files" ?

http://0x0.st/oTuz.txt

seth · 2022-04-22 20:04:27

So the user daemon is pricipally ok - might be worth a shot to simply re-install gvfs and if doesn't work, see wheter "systemctl --user edit gvfs-daemon" reveals a local override (and maybe in general whether the problem extends to a different account)

girzel · 2022-04-22 20:46:20

seth wrote:

So the user daemon is pricipally ok - might be worth a shot to simply re-install gvfs and if doesn't work, see wheter "systemctl --user edit gvfs-daemon" reveals a local override (and maybe in general whether the problem extends to a different account)

Will give it a shot tonight, when I'm done with work. By "different account" do you mean setting up another user on the machine and trying to replicate there?

seth · 2022-04-22 23:37:07

By "different account" do you mean setting up another user on the machine and trying to replicate there?

Exactly.

girzel · 2022-04-24 04:40:14

I reinstalled gvfs, and unfortunately the broken behavior came back, same as before.

So I created my alter-ego, Freddy, and logged in as him. Exact same behavior, including running gvfsd manually as a quick fix. Here are Freddy's unit files:

http://0x0.st/oT95.txt

seth · 2022-04-24 06:37:07

The gvfs services are there, but I guess the process doesn't run.

I'l throw a hail-mary: install https://archlinux.org/packages/community/x86_64/dunst/
Don't ask, just try. I'll explain afterwards.

girzel · 2022-04-25 00:15:43

I installed dunst and ran "systemctl --user start dunst", and it timed out in exactly the same way as gvfs-daemon did. I added "--verbosity debug" to the .service file, reloaded, and tried again -- also timed out and provided no more information than before. Mind you, running "dunst --verbosity debug" in the shell directly didn't produce any output either, but I guess nothing tried to use it.

It sort of looks like these commands aren't getting run at all, ie there's something wrong with the "systemctl --user" command. The timeout's coming from there, not the underlying command. Does that seem likely?

seth · 2022-04-25 06:28:39

The initial impression and common cause is a broken user session (bus), yes - but we dealt w/ your xinitrc early on and you can talk to it to list the units.
We also eradicated flatpak/portal as the usual bus-breaker.

Can you

dbus-send --session --print-reply --dest=org.freedesktop.DBus /org/freedesktop/DBus org.freedesktop.DBus.ListNames | sed '/string/!d; s/^\s*string //g'

Next hail-mary:

echo $SHELL
stat /bin/sh

(The previous idea was btw. that gvfsd tried a sync notification for some reason and failed to talk to the notification daemon - but that would not affect dunst, such notification daemon itself…)

girzel · 2022-04-25 17:31:00

seth wrote:

The initial impression and common cause is a broken user session (bus), yes - but we dealt w/ your xinitrc early on and you can talk to it to list the units.
We also eradicated flatpak/portal as the usual bus-breaker.
Can you
dbus-send --session --print-reply --dest=org.freedesktop.DBus /org/freedesktop/DBus org.freedesktop.DBus.ListNames | sed '/string/!d; s/^\s*string //g'

Without gvfsd running, that gives me:

[eric@clem ~]$ dbus-send --session --print-reply --dest=org.freedesktop.DBus /org/freedesktop/DBus org.freedesktop.DBus.ListNames | sed '/string/!d; s/^\s*string //g'
"org.freedesktop.DBus"
"org.mozilla.firefox.ZGVmYXVsdC0xNDYwMzM5OTU1NTM1"
":1.7"
":1.24"
":1.0"
":1.35"
":1.8"
"org.pulseaudio.Server"
":1.1"
":1.2"
":1.22"

Running "gfvsd --replace" manually and doing it again gives me:

[eric@clem ~]$ dbus-send --session --print-reply --dest=org.freedesktop.DBus /org/freedesktop/DBus org.freedesktop.DBus.ListNames | sed '/string/!d; s/^\s*string //g'
"org.freedesktop.DBus"
":1.7"
":1.8"
":1.22"
"org.mozilla.firefox.ZGVmYXVsdC0xNDYwMzM5OTU1NTM1"
":1.0"
":1.24"
"org.gtk.vfs.Daemon"
":1.1"
"org.pulseaudio.Server"
":1.36"
":1.2"
":1.37"
":1.38"

seth wrote:

Next hail-mary:
echo $SHELL
stat /bin/sh

[eric@clem ~]$ echo $SHELL
/bin/bash
[eric@clem ~]$ stat /bin/sh
  File: /bin/sh -> bash
  Size: 4               Blocks: 0          IO Block: 4096   symbolic link
Device: 8,19    Inode: 661568      Links: 1
Access: (0777/lrwxrwxrwx)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 2022-04-24 16:48:09.003654848 -0700
Modify: 2022-01-08 10:31:11.000000000 -0800
Change: 2022-01-09 18:35:27.950409956 -0800
 Birth: 2022-01-09 18:35:27.950409956 -0800

seth · 2022-04-25 19:55:49

The bus is there, but where's org.freedesktop.systemd1 ?

ps fx

girzel · 2022-04-25 20:43:40

seth wrote:

The bus is there, but where's org.freedesktop.systemd1 ?
ps fx

Without running the manual gvfsd command: http://0x0.st/oTxn.txt

While the manual command is running: http://0x0.st/oTx5.txt

(I just know this is going to turn out to be some stupid thing I did years ago, then forgot...)

seth · 2022-04-25 21:19:32

"/usr/lib/systemd/systemd --user" is there, but it's not registered to "/usr/bin/dbus-daemon --session --address=systemd: --nofork --nopidfile --systemd-activation --syslog-only", though starting it.

I found https://github.com/systemd/systemd/issues/5247 but it's inconclusive (except that lennart defaulted to notmybug…) - are there systemd related errors in the journal?
Did "Freddy" inherit your bashrc/profile and or at least the pattern (iff) that runs startx?

girzel · 2022-04-25 21:39:30

seth wrote:

"/usr/lib/systemd/systemd --user" is there, but it's not registered to "/usr/bin/dbus-daemon --session --address=systemd: --nofork --nopidfile --systemd-activation --syslog-only", though starting it.
I found https://github.com/systemd/systemd/issues/5247 but it's inconclusive (except that lennart defaulted to notmybug…) - are there systemd related errors in the journal?

Yes there are!

[root@clem home]# journalctl -b --grep systemd1
Apr 25 10:48:27 clem dbus-daemon[430]: [system] Successfully activated service 'org.freedesktop.systemd1'
Apr 25 10:51:27 clem dbus-daemon[1018]: [session uid=1000 pid=1018] Failed to activate service 'org.freedesktop.systemd1': timed out (service_start_timeout=120000ms)
Apr 25 11:28:08 clem dbus-daemon[1018]: [session uid=1000 pid=1018] Failed to activate service 'org.freedesktop.systemd1': timed out (service_start_timeout=120000ms)
Apr 25 11:32:32 clem dbus-daemon[1018]: [session uid=1000 pid=1018] Failed to activate service 'org.freedesktop.systemd1': timed out (service_start_timeout=120000ms)
Apr 25 13:43:40 clem dbus-daemon[1018]: [session uid=1000 pid=1018] Failed to activate service 'org.freedesktop.systemd1': timed out (service_start_timeout=120000ms)

seth wrote:

Did "Freddy" inherit your bashrc/profile and or at least the pattern (iff) that runs startx?

Yes, I did the same routine of copying in the template, commenting out the twm/xclock stuff at the bottom, and adding "exec /usr/bin/i3"

seth · 2022-04-25 21:42:31

Maybe we should take a look at the bashrc and bash_profile then…

girzel · 2022-04-25 21:51:42

seth wrote:

Maybe we should take a look at the bashrc and bash_profile then…

Here's all of ~/.bash_profile, I have no bashrc.

export PATH=/home/eric/.npm-global/bin:/home/eric/.local/bin:/home/eric/.cabal/bin:$PATH
export HISTCONTROL=ignoreboth:erasedups

alias ll='ls -l'
alias la='ls -A'
alias ls='ls --color=auto'

#export DBUS_SESSION_BUS_ADDRESS=unix:path=/run/user/$UID/dbus/user_bus_socket

export GPGKEY=9903E6EE

alias loct='/usr/bin/locate -i -d /home/eric/.mlocate.db'

alias ec="emacsclient -c -n"

alias updatedb_local='/usr/bin/updatedb -U /home/eric -o /home/eric/.mlocate.db'

source /usr/share/bash-completion/completions/pass
source /usr/share/bash-completion/completions/git
source /usr/share/bash-completion/completions/python
source /usr/share/bash-completion/completions/ssh
source /usr/share/bash-completion/completions/kill
source /usr/share/bash-completion/completions/pkill
source /usr/share/bash-completion/completions/killall
source /usr/share/bash-completion/completions/xdg-mime
source /usr/share/bash-completion/completions/systemctl
source /usr/share/bash-completion/completions/networkctl
#source /usr/share/bash-completion/completions/pmount
source /usr/share/bash-completion/completions/netctl
source /usr/share/bash-completion/completions/pacman
source /usr/share/bash-completion/completions/journalctl
source /usr/share/bash-completion/completions/pikaur

# OPAM configuration

export VISUAL="emacsclient -c"
export EDITOR="emacsclient -c"
export ALTERNATE_EDITOR=""

source /usr/share/nvm/init-nvm.sh

seth · 2022-04-26 06:18:22

#export DBUS_SESSION_BUS_ADDRESS=unix:path=/run/user/$UID/dbus/user_bus_socket

I was so happy before I saw the hash… do you maybe export that somewhere else (in /etc/profile* or so)?

girzel · 2022-04-27 00:39:23

seth wrote:

#export DBUS_SESSION_BUS_ADDRESS=unix:path=/run/user/$UID/dbus/user_bus_socket
I was so happy before I saw the hash… do you maybe export that somewhere else (in /etc/profile* or so)?

No sorry (this was definitely from a much earlier ill-advised copy-n-paste adventure), and unfortunately, wherever it does get set, it seems to be set correctly:

[eric@clem ~]$ env | grep -i bus
DBUS_SESSION_BUS_ADDRESS=unix:path=/run/user/1000/bus

seth · 2022-05-07 13:32:54

So let's revisit this mess (if you want)

    851 ?        Ss     0:00 /usr/lib/systemd/systemd --user
    852 ?        S      0:00  \_ (sd-pam)
    859 ?        Ssl    0:00  \_ /usr/bin/mpd --no-daemon /home/eric/.mpd/mpd.conf
    861 ?        Ssl    0:00  \_ /usr/bin/syncthing serve --no-browser --no-restart --logflags=0
    875 ?        SNl    0:21  |   \_ /usr/bin/syncthing serve --no-browser --no-restart --logflags=0
   1013 ?        Ssl    0:00  \_ /usr/bin/pipewire
   1014 ?        Ssl    0:00  \_ /usr/bin/pipewire-media-session
   1015 ?        Ssl    0:00  \_ /usr/bin/pipewire-pulse
   1018 ?        Ss     0:00  \_ /usr/bin/dbus-daemon --session --address=systemd: --nofork --nopidfile --systemd-activation --syslog-only
   1476 ?        SLs    0:04  \_ /usr/bin/gpg-agent --supervised
   1560 ?        SLl    0:00      \_ scdaemon --multi-server

mpd, I guess by inference pipewire and syncthing are all started before the dbus-daemon - I assume somehow w/ the session.
I feel the most uneasy about pipewire and if we assume that it's drawn in by mpd: what if you skip/delay the mpd launch?

You don't have to install gvfs for testing this, "org.freedesktop.systemd1" showing up in "dbus-send --session --print-reply --dest=org.freedesktop.DBus /org/freedesktop/DBus org.freedesktop.DBus.ListNames | sed '/string/!d; s/^\s*string //g'" would likely be significant.

Arch Linux

#26 2022-04-22 15:50:35

Re: at-spi-bus-launcher segfaults

#27 2022-04-22 16:15:37

Re: at-spi-bus-launcher segfaults

#28 2022-04-22 16:31:56

Re: at-spi-bus-launcher segfaults

#29 2022-04-22 18:31:20

Re: at-spi-bus-launcher segfaults

#30 2022-04-22 20:04:27

Re: at-spi-bus-launcher segfaults

#31 2022-04-22 20:46:20

Re: at-spi-bus-launcher segfaults

#32 2022-04-22 23:37:07

Re: at-spi-bus-launcher segfaults

#33 2022-04-24 04:40:14

Re: at-spi-bus-launcher segfaults

#34 2022-04-24 06:37:07

Re: at-spi-bus-launcher segfaults

#35 2022-04-25 00:15:43

Re: at-spi-bus-launcher segfaults

#36 2022-04-25 06:28:39

Re: at-spi-bus-launcher segfaults

#37 2022-04-25 17:31:00

Re: at-spi-bus-launcher segfaults

#38 2022-04-25 19:55:49

Re: at-spi-bus-launcher segfaults

#39 2022-04-25 20:43:40

Re: at-spi-bus-launcher segfaults

#40 2022-04-25 21:19:32

Re: at-spi-bus-launcher segfaults

#41 2022-04-25 21:39:30

Re: at-spi-bus-launcher segfaults

#42 2022-04-25 21:42:31

Re: at-spi-bus-launcher segfaults

#43 2022-04-25 21:51:42

Re: at-spi-bus-launcher segfaults

#44 2022-04-26 06:18:22

Re: at-spi-bus-launcher segfaults

#45 2022-04-27 00:39:23

Re: at-spi-bus-launcher segfaults

#46 2022-05-07 13:32:54

Re: at-spi-bus-launcher segfaults

Board footer