Paul Gideon Dann’s patchset for Poppler to enable Okular (Qt5) to use Cairo rather than Splash to render PDF files

If you view the same PDF file in Okular (KDE) and Evince (GNOME), you may notice that fonts and lines are rendered better in Evince. Both applications use Poppler to render text and graphics in PDF files, but Poppler uses a different rendering backend in the two applications. For Evince Poppler uses the Cairo library, whereas for Okular Poppler uses Splash, a backend inherited from Poppler’s predecessor Xpdf (still in development). Unfortunately for KDE users, Cairo often does a better job than Splash. However, independent software engineer Paul Gideon Dann came to the rescue by producing the patchset poppler-cairo-backend to modify Poppler in order to make it use the Cairo library instead of Splash when Poppler is used by Okular. To quote the README file for Paul’s patchset:

Purpose of this Patchset

Currently, the default backend for the Qt5 wrapper (used by Okular) is Splash. Unfortunately, Splash does not support subpixel rendering of fonts, so those of us using KDE are stuck with somewhat ugly-looking fonts. This patchset adds support for the Cairo backend to the Qt5 wrapper. It also forces subpixel rendering in the Cairo backend. The upshot of this is that we get beautiful fonts in Okular.

The README focuses on fonts, but in fact the rendering of lines in graphics in PDF files can also be improved by the application of the patchset.

Apparently the Poppler maintainer feels that the introduction of a dependency on Cairo to the Qt5 wrapper (even an optional dependency) in Poppler would be controversial, and he is not willing to merge the patchset. For Okular users who already have Cairo installed (e.g. for Firefox, Inkscape, Scribus and so on), and who are noticing inadequate rendering of some PDF files, Paul’s patchset is worth trying.

In Gentoo Linux, which is a source code-based distribution, it is very easy to apply the patchset. For example, I did the following to apply the patchset for Poppler 0.80.0 in a ~amd64 (Testing Branch) installation:

1. Created a package-specific and version-specific directory to hold the patchset:

root # mkdir -p /etc/portage/patches/app-text/poppler-0.80.0

2. Downloaded the patchset for Poppler 0.80.0 from the following Web page:

https://github.com/giddie/poppler-cairo-backend/tree/76e607bcf010d6d9b8df5cb0f851ef9c91d4caf2

3. Copied the patchset to the directory created in Step 1:

root # cp /home/fitzcarraldo/Downloads/*.patch /etc/portage/patches/app-text/poppler-0.80.0/
root # ls -1 /etc/portage/patches/app-text/poppler-0.80.0
0001-Cairo-backend-added-to-Qt5-wrapper.patch
0002-Setting-default-Qt5-backend-to-Cairo.patch
0003-Apply-subpixel-rendering-in-Cairo-Backend.patch

4. Checked first that the patchset could be applied successfully before actually using it:

root # cd /usr/portage/app-text/poppler
root # ebuild poppler-0.80.0.ebuild clean prepare
 * poppler-0.80.0.tar.xz BLAKE2B SHA512 size ;-) ...                                     [ ok ]
 * checking ebuild checksums ;-) ...                                                     [ ok ]
 * checking auxfile checksums ;-) ...                                                    [ ok ]
 * checking miscfile checksums ;-) ...                                                   [ ok ]
>>> Unpacking source...
>>> Unpacking poppler-0.80.0.tar.xz to /var/tmp/portage/app-text/poppler-0.80.0/work
>>> Source unpacked in /var/tmp/portage/app-text/poppler-0.80.0/work
>>> Preparing source in /var/tmp/portage/app-text/poppler-0.80.0/work/poppler-0.80.0 ...
 * Applying poppler-0.60.1-qt5-dependencies.patch ...                                    [ ok ]
 * Applying poppler-0.28.1-fix-multilib-configuration.patch ...                          [ ok ]
 * Applying poppler-0.78.0-respect-cflags.patch ...                                      [ ok ]
 * Applying poppler-0.61.0-respect-cflags.patch ...                                      [ ok ]
 * Applying poppler-0.57.0-disable-internal-jpx.patch ...                                [ ok ]
 * Applying 0001-Cairo-backend-added-to-Qt5-wrapper.patch ...                            [ ok ]
 * Applying 0002-Setting-default-Qt5-backend-to-Cairo.patch ...                          [ ok ]
 * Applying 0003-Apply-subpixel-rendering-in-Cairo-Backend.patch ...                     [ ok ]
 * User patches applied.
>>> Source prepared.

5. Re-merged Poppler to apply the patchset to the Poppler source code and rebuild the patched package:

root # emerge -1v poppler

These are the packages that would be merged, in order:

Calculating dependencies... done!
[ebuild   R    ] app-text/poppler-0.80.0:0/90::gentoo  USE="cairo cjk cxx introspection jpeg jpeg2k lcms png qt5 tiff utils -curl -debug -doc -nss" 0 KiB

Total: 1 package (1 reinstall), Size of downloads: 0 KiB

>>> Verifying ebuild manifests
>>> Emerging (1 of 1) app-text/poppler-0.80.0::gentoo
>>> Installing (1 of 1) app-text/poppler-0.80.0::gentoo
>>> Jobs: 1 of 1 complete                           Load avg: 1.06, 1.11, 0.95
>>> Auto-cleaning packages...

>>> No outdated packages were found on your system.

 * GNU info directory index is up-to-date.

6. Re-merged Okular so that it uses the patched Poppler dependency:

root # emerge -1v okular

These are the packages that would be merged, in order:

Calculating dependencies... done!
[ebuild   R    ] kde-apps/okular-19.08.1:5::gentoo  USE="chm crypt djvu image-backend pdf postscript tiff -debug -epub -handbook -markdown -mobi -mobile -plucker -share -speech -test" 0 KiB

Total: 1 package (1 reinstall), Size of downloads: 0 KiB

>>> Verifying ebuild manifests
>>> Emerging (1 of 1) kde-apps/okular-19.08.1::gentoo
>>> Installing (1 of 1) kde-apps/okular-19.08.1::gentoo
>>> Jobs: 1 of 1 complete                           Load avg: 1.17, 1.13, 1.04
>>> Auto-cleaning packages...

>>> No outdated packages were found on your system.

 * GNU info directory index is up-to-date.

My thanks go to Paul for taking the time to produce the patchset.

Preventing Lubuntu 18.04 from leaving a user process running after the user logs out

My family’s desktop machine has Lubuntu 18.04 installed, which uses systemd and the LXDE desktop environment. Each family member has their own user account, thus the installation is a single-seat, multi-user installation. For each user’s account I set up the virus-checking scheme described in an earlier post, suitably modified to take into account the differences between Lubuntu 18.04 and Gentoo Linux running KDE. For example, the monitorDownloadsGUI script in Lubuntu 18.04 uses zenity rather than kdialog, and, as Lubuntu 18.04 uses systemd, the ClamAV daemon’s service file in Lubuntu 18.04 is /lib/systemd/system/clamav-daemon.service rather than the OpenRC init file /etc/init.d/clamd used in my Gentoo Linux installations.

The virus-checking script ~/.monitorDownloadGUI in each user’s home directory is launched automatically by LXDE at login because I created a Desktop Configuration File ~/.config/autostart/monitorDownloadsGUI.desktop in each user’s account. For example, the contents of the file in my account are as follows:

[Desktop Entry]
Type=Application
Exec=/home/fitzcarraldo/.monitorDownloadsGUI

However, I recently noticed that Lubuntu 18.04 does not terminate the monitorDownloadsGUI process when the user logs out. I do not see this behaviour on my laptops running Gentoo Linux with OpenRC and KDE, so I am not sure why this is happening in Lubuntu 18.04 with systemd and LXDE. The output of the ‘ps -ef‘ command after each of the three example steps shown below illustrates the behaviour.

Step 1. george is the only user who is logged-in.

$ ps -ef | grep bash | grep -v grep
george    1410     1  0 02:05 ?        00:00:00 /bin/bash /home/george/.monitorDownloadsGUI
george    1597  1358  0 02:05 pts/0    00:00:00 /bin/bash

Step 2. ringo uses ‘Logout’ > ‘Switch User’ to login to his account.

$ ps -ef | grep bash | grep -v grep
george    1410     1  0 02:05 ?        00:00:00 /bin/bash /home/george/.monitorDownloadsGUI
george    1597  1358  0 02:05 pts/0    00:00:00 /bin/bash
ringo     2382     1  0 02:06 ?        00:00:00 /bin/bash /home/ringo/.monitorDownloadsGUI

Step 3. ringo logs out of his account.

$ ps -ef | grep bash | grep -v grep
george    1410     1  0 02:05 ?        00:00:00 /bin/bash /home/george/.monitorDownloadsGUI
george    1597  1358  0 02:05 pts/0    00:00:00 /bin/bash
ringo     2382     1  0 02:06 ?        00:00:00 /bin/bash /home/ringo/.monitorDownloadsGUI

Notice that the process with PID 2382 is still running, even though user ringo is no longer logged in.

If a user logs out and logs in again, or if users switch between sessions using ‘Logout’ > ‘Switch User’, it is also possible for multiple instances of the script per user to be running. For example:

$ ps -ef | grep bash | grep -v grep
george    1564     1  0 11:14 ?        00:00:00 /bin/bash /home/george/.monitorDownloadsGUI
ringo     2522     1  0 11:16 ?        00:00:00 /bin/bash /home/ringo/.monitorDownloadsGUI
george    3803     1  0 11:17 ?        00:00:00 /bin/bash /home/george/.monitorDownloadsGUI
george    5997     1  0 11:19 ?        00:00:00 /bin/bash /home/george/.monitorDownloadsGUI
george    6054  5881  0 11:19 pts/0    00:00:00 /bin/bash

Notice that several instances of the script are running for user george. There should only be one instance.

In order to prevent these multiple instances, I added the shell script lines below to the existing LightDM session-cleanup-script that I had created previously to solve a different problem in the Lubuntu 18.04 installation (see an earlier blog post).

# Get rid of duplicate instances (if any) per user of the virus-checker script's process
who -u | grep -v "\." > /tmp/logged-in_users
while IFS=: read -r f1 f2 f3 f4 f5 f6 f7
# $f1 is username
# $f2 is password ('x')
# $f3 is UID
# $f4 is GID
# $f5 is UID info
# $f6 is home directory
# $f7 is command/shell
do
    match=0
    while read a b c d e f g h # Use this if this script is launched by LightDM in Lubuntu 18.04
#    while read a b c d e f g # Use this if you launch this script from a terminal in Lubuntu 18.04
    #
    # If this script is launched by a user, 'who -u' returns the following fields:
    # "john     tty7         2019-08-31 17:08 00:01        1624 (:0)"
    # If this script is launched by LightDM, 'who -u' returns the following fields:
    # "john     tty7        Aug 31 17:08 00:01        1624 (:0)"
    #
    do
        if [[ $f6 == *"/home/"* ]] && [[ $f7 == "/bin/bash" ]] && [[ $a == $f1 ]]; then
            match=1
            user=$f1
            tty=$b
        fi
    done < /tmp/logged-in_users
    if [[ $match -eq 1 ]] && [[ $(echo $tty | sed 's/[^0-9]*//g') -gt 6 ]]; then
        if [[ `ps -ef | grep bash | grep "$user" | grep monitorDownloadsGUI | awk -F' ' '{print $2}' | wc -l` -gt 1 ]]; then
            kill `ps -ef | grep bash | grep "$user" | grep monitorDownloadsGUI | awk -F' ' '{print $2}' | tail -n +2`
        fi
    elif [[ $match -ne 1 ]]; then
        if [[ $f6 == *"/home/"* ]] && [[ $f7 == "/bin/bash" ]] && [[ `ps -ef | grep bash | grep "$f1" | grep monitorDownloadsGUI | awk -F' ' '{print $2}' | wc -l` -gt 1 ]]; then
            kill `ps -ef | grep bash | grep "$f1" | grep monitorDownloadsGUI | awk -F' ' '{print $2}' | tail -n +2`
        elif [[ $f6 == *"/home/"* ]] && [[ $f7 == "/bin/bash" ]] && [[ `ps -ef | grep bash | grep "$f1" | grep monitorDownloadsGUI | awk -F' ' '{print $2}' | wc -l` -eq 1 ]]; then
            kill `ps -ef | grep bash | grep "$f1" | grep monitorDownloadsGUI | awk -F' ' '{print $2}'`
        fi
    fi
done < /etc/passwd
rm /tmp/logged-in_users

The above lines of Bash script kill additional instances of monitorDownloadGUI on a per-user basis when a user session ends. If LightDM’s session-cleanup-script does this, there will be no more than one instance of a monitorDownloadsGUI process per logged-in user, and no instances of a monitorDownloadGUI process for users who have logged out:

$ ps -ef | grep bash | grep -v grep
george    1473     1  0 12:32 ?        00:00:00 /bin/bash /home/george/.monitorDownloadsGUI
george    1693  1412  0 12:32 pts/0    00:00:00 /bin/bash

Problem solved. Well, worked around. I would like to know what causes the problem to happen in the first place. I assume it is either systemd or LXDE.

How to run KDE Dolphin, Kate and KWrite as root user

When using KDE I occasionally wish to launch KWrite or Kate as root user in order to edit system files more easily than using a TUI editor in a terminal window (either launched as root user or by using the sudoedit command). Being able to browse using Dolphin as the root user occasionally is also useful. These all used to be possible by launching the application with the kdesu command, but in 2017 KDE developer Martin Gräßlin removed this option on security grounds (see his blog post ‘Editing files as root‘). Attempting to launch e.g. Kate using the sudo command results in the following message:

$ sudo kate
Executing Kate with sudo is not possible due to unfixable security vulnerabilities.

Attempting to launch e.g. Kate using the kdesu command results in a pop-up window prompting me to enter the root user’s password, but then does not launch Kate:

$ kdesu kate
$

I am willing to accept a small risk despite the ‘unfixable security vulnerabilities’ , and a 2018 Kubuntu Forums post by KDE user Rog131 provided me with a solution. It is possible to launch Dolphin, Kate and KWrite as root from your user account by using the pkexec command. For example, to launch Dolphin you can enter:

$ pkexec env DISPLAY=$DISPLAY XAUTHORITY=$XAUTHORITY KDE_SESSION_VERSION=5 KDE_FULL_SESSION=true dolphin

Dolphin first displays an orange-coloured box with the warning message ‘Running Dolphin as root can be dangerous. Please be careful.’ and you can then browse and open root-owned directories and files.

You can also launch Kate and KWrite as root from your user account in the same way:

$ pkexec env DISPLAY=$DISPLAY XAUTHORITY=$XAUTHORITY KDE_SESSION_VERSION=5 KDE_FULL_SESSION=true kate

$ pkexec env DISPLAY=$DISPLAY XAUTHORITY=$XAUTHORITY KDE_SESSION_VERSION=5 KDE_FULL_SESSION=true kwrite

To make it easy to launch them as root user from e.g. Konsole or Yakuake you could set aliases for the three commands in your ~/.bashrc file:

$ tail -n 3 ~/.bashrc
alias dolroot="pkexec env DISPLAY=$DISPLAY XAUTHORITY=$XAUTHORITY KDE_SESSION_VERSION=5 KDE_FULL_SESSION=true dolphin"
alias kateroot="pkexec env DISPLAY=$DISPLAY XAUTHORITY=$XAUTHORITY KDE_SESSION_VERSION=5 KDE_FULL_SESSION=true kate"
alias kwriteroot="pkexec env DISPLAY=$DISPLAY XAUTHORITY=$XAUTHORITY KDE_SESSION_VERSION=5 KDE_FULL_SESSION=true kwrite"

Then all you would need to type in a terminal window would be:

$ dolroot
$ kateroot
$ kwriteroot

which are no more difficult than having to type:

$ kdesu dolphin
$ kdesu kate
$ kdesu kwrite

If an alias is used, rooted-Dolphin/Kate/KWrite can be launched from the command line but cannot be launched via KDE’s Application Launcher menu or KRunner. On the other hand, if a wrapper script is used, rooted-Dolphin/Kate/KWrite can be launched from the user’s command line and via KDE’s Application Launcher menu (and therefore via KRunner too). For example, I created three tiny Bash scripts dolroot, kateroot and kwriteroot. The scripts simply contain the aforementioned pkexec command. For example, dolroot contains:

#!/bin/bash
pkexec env DISPLAY=$DISPLAY XAUTHORITY=$XAUTHORITY KDE_SESSION_VERSION=5 KDE_FULL_SESSION=true dolphin

Don’t forget to make them executable:

$ chmod 700 dolroot
$ chmod 700 kateroot
$ chmod 700 kwriteroot
$ ls -la *root
-rwx------ 1 fitzcarraldo fitzcarraldo 115 Jul 30 15:33 dolroot
-rwx------ 1 fitzcarraldo fitzcarraldo 112 Jul 30 15:34 kateroot
-rwx------ 1 fitzcarraldo fitzcarraldo 114 Jul 30 15:34 kwriteroot	

After adding entries for dolroot, kateroot and kwriteroot to the KDE Application Launcher’s menu, you can press Alt+F2 as usual to display the KRunner launcher then enter ‘dolroot’, ‘kateroot’ or ‘kwriteroot’ (without the quotes, obviously) in the KRunner window to launch Dolphin/Kate/KWrite as root user. A window will pop-up for you to enter the root user’s password. Once you have entered the root user’s password, the application will be launched.

Thankfully KDE’s Nathaniel Graham is pragmatic:

D12795 – Re-allow running Dolphin as the root user (but still not using sudo)
D12732 – Show a warning when running as the root user

How to change the height of the Kickoff Application Launcher menu in KDE Plasma

The height of the KDE Plasma Kickoff Application Launcher menu is not user-configurable, which is odd in a Desktop Environment with a reputation for being highly user-configurable.

It turns out that the height and width of the pop-up menu are hard-coded in the ASCII file /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml:

root # grep -E "Layout.minimumHeight.*units.gridUnit" /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml
    Layout.minimumHeight: units.gridUnit * 34

root # grep -E "Layout.minimumWidth.*units.gridUnit" /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml
    Layout.minimumWidth: units.gridUnit * 26

Now, I was a bit fed up having to scroll up and down the launcher menu to see all fourteen entries in my Favourites list, so I decided to increase the height of the menu, which I did by editing /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml as root user:

root # nano /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml
root # grep -E "Layout.minimumHeight.*units.gridUnit" /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml
    Layout.minimumHeight: units.gridUnit * 44

The only downside to this is that the file will be overwritten when the package kde-plasma/plasma-desktop is upgraded.

The following command would allow me to make sure the file contains the height value of ’44’ that I want:

root # sed -i '/Layout.minimumHeight: units.gridUnit/ c\    Layout.minimumHeight: units.gridUnit * 44' /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml

Therefore, to automate the editing of the file in my Gentoo installations that use OpenRC I created a shell script /etc/local.d/50-set_Kickoff_height.start with the following contents:

#!/bin/bash
if [ -e /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml ]; then
    sed -i '/Layout.minimumHeight: units.gridUnit/ c\    Layout.minimumHeight: units.gridUnit * 44' /usr/share/plasma/plasmoids/org.kde.plasma.kickoff/contents/ui/FullRepresentation.qml
fi

The FullRepresentation.qml file will then be edited every time the machine boots, which is a tad inefficient but not a big overhead.

This is not a perfect solution because the menu will revert to its default height following an upgrade to the package kde-plasma/plasma-desktop until I reboot the machine, but it is good enough for me.

How to stop inactive user sessions triggering Suspend to RAM in a single-seat, multi-user installation of Lubuntu 18.04

In my previous post I mentioned a problem that I had still not been able to fix in a single-seat, multi-user installation of Lubuntu 18.04: Xfce Power Manager in each user’s account can cause the installation to suspend to RAM if a user has not logged out of his/her session and another user is using a different session. Each user account in Lubuntu 18.04 has its own XfcePower Manager settings, stored in the file ~/.config/xfce4/xfconf/xfce-perchannel-xml/xfce4-power-manager.xml. If the property /xfce4-power-manager/inactivity-on-ac has a value of 15 (minutes) or higher, that session can cause the machine to suspend to RAM even if the session is not active while someone else’s session is active. The example below illustrates the effect.

Consider five users mick, christine, john, stevie and lindsey with the following settings for the number of minutes of inactivity that will trigger suspension to RAM:

mick@aspirexc600:~$ xfconf-query -c xfce4-power-manager -p /xfce4-power-manager/inactivity-on-ac
30

christine@aspirexc600:~$ xfconf-query -c xfce4-power-manager -p /xfce4-power-manager/inactivity-on-ac
25

john@aspirexc600:~$ xfconf-query -c xfce4-power-manager -p /xfce4-power-manager/inactivity-on-ac
45

stevie@aspirexc600:~$ xfconf-query -c xfce4-power-manager -p /xfce4-power-manager/inactivity-on-ac
15

lindsey@aspirexc600:~$ xfconf-query -c xfce4-power-manager -p /xfce4-power-manager/inactivity-on-ac
30

Now, suppose that john boots the machine, logs in to his account to check his e-mail, leaves the e-mail client open and goes off to grab lunch without logging out. Then stevie comes along and clicks on ‘Logout’ > ‘Switch User’ to display the LightDM greeter screen (or the greeter screen is already displayed because john‘s session has already been locked), logs in to her account and begins to use, say, LibreOffice Writer. Even though stevie is busy typing, the machine will suspend to RAM after 45 minutes of inactivity by john. This can be very annoying.

In addition to the individual users’ Xfce Power Manager configuration files in Lubuntu 18.04, I found the following Xfce Power Manager configuration files which appear to be system-wide:

/etc/xdg/xdg-Lubuntu/xfce4/xfconf/xfce-perchannel-xml/xfce4-power-manager.xml
/etc/xdg/xfce4/xfconf/xfce-perchannel-xml/xfce4-power-manager.xml

First attempt at fixing the problem

I asked all the users to configure their accounts to never cause the machine to suspend, by using the Xfce Power Manager settings GUI in their session and selecting ‘Never’. I noticed this caused each user’s /xfce4-power-manager/inactivity-on-ac property to become ‘14‘:

user $ xfconf-query -c xfce4-power-manager -p /xfce4-power-manager/inactivity-on-ac
14

Note that users must not edit their file ~/.config/xfce4/xfconf/xfce-perchannel-xml/xfce4-power-manager.xml; if they do, the settings shown in the Xfce Power Manager settings GUI will not be updated. Users must either use the Xfce Power Manager settings GUI or xfconf-query commands as explained on the askubuntu Web page ‘Change xfce4-power-manager option from terminal‘.

Then I edited the file /etc/xdg/xdg-Lubuntu/xfce4/xfconf/xfce-perchannel-xml/xfce4-power-manager.xml to make its contents the same as the previous contents of the individual users’ settings when Suspend to RAM was enabled individually):

<?xml version="1.0" encoding="UTF-8"?>

<channel name="xfce4-power-manager" version="1.0">
  <property name="xfce4-power-manager" type="empty">
    <property name="power-button-action" type="uint" value="3"/>
    <property name="show-tray-icon" type="bool" value="true"/>
    <property name="brightness-switch-restore-on-exit" type="int" value="1"/>
    <property name="brightness-switch" type="int" value="0"/>
    <property name="presentation-mode" type="bool" value="false"/>
    <property name="inactivity-on-ac" type="uint" value="30"/>
    <property name="blank-on-ac" type="int" value="10"/>
    <property name="dpms-on-ac-sleep" type="uint" value="0"/>
    <property name="dpms-on-ac-off" type="uint" value="0"/>
    <property name="brightness-on-ac" type="uint" value="9"/>
    <property name="lock-screen-suspend-hibernate" type="bool" value="true"/>
    <property name="logind-handle-lid-switch" type="bool" value="false"/>
    <property name="dpms-enabled" type="bool" value="false"/>
    <property name="general-notification" type="bool" value="true"/>
    <property name="sleep-button-action" type="uint" value="0"/>
    <property name="hibernate-button-action" type="uint" value="0"/>
  </property>
</channel>

After rebooting, leaving one or more users logged in without any activity did not cause the installation to suspend to RAM after 30 minutes of no activity in any session.

So I then edited the file /etc/xdg/xfce4/xfconf/xfce-perchannel-xml/xfce4-power-manager.xml to make its contents the same as the above. After rebooting, leaving one or more users logged in without any activity did not cause the installation to suspend to RAM after 30 minutes of no activity in any session.

I therefore assume that the above two files are ignored by Xfce Power Manager during normal operation.

Second attempt at fixing the problem

I used the procedure given in the Xfce4-power-manager FAQs to check if Xfce Power Manager in Lubuntu 18.04 uses systemd-logind to suspend the installation, and indeed it does:

TRACE[xfpm-polkit.c:366] xfpm_polkit_check_auth_intern(): Action=org.freedesktop.login1.suspend is authorized=TRUE

Therefore I edited /etc/systemd/logind.conf to add IdleAction=suspend and IdleActionSec=30min, and rebooted. However, this had no discernable effect either. Leaving one or more users logged in without any activity did not cause the installation to suspend to RAM after 30 minutes of no activity in any session.

Third attempt (successful) at fixing the problem

So, what to do?! In my previous post I explained how I had fixed the problem of not being able to suspend to RAM automatically from the LightDM greeter screen. I decided to keep the scripts from that post and add a new script sessions_sleep.sh to the root crontab. The contents of all the files and the crontab are shown below.

user $ cd /etc/lightdm/lightdm.conf.d/
user $ cat 10_lubuntu.conf 
[Seat:*]
greeter-setup-script=/etc/lightdm/lightdm.conf.d/lightdm_sleep.sh
session-setup-script=/etc/lightdm/lightdm.conf.d/lightdm_kill_sleep.sh
session-cleanup-script=/etc/lightdm/lightdm.conf.d/unmount_FREECOM_HDD.sh

user $ ls
05_lubuntu.conf  lightdm_kill_sleep.sh       sessions_sleep.sh
10_lubuntu.conf  lightdm_sleep.sh            unmount_FREECOM_HDD.sh

user $ cat lightdm_sleep.sh 
#!/bin/bash
# This forms part of the scheme to provide automatic suspension while the greeter screen is displayed
file="/tmp/unique_identifier"
(while true; do sleep 30m; systemctl suspend; done) &
echo $! > $file

user $ cat lightdm_kill_sleep.sh
#!/bin/bash
# This forms part of the scheme to provide automatic suspension while the greeter screen is displayed
file="/tmp/unique_identifier"
if [ -f "$file" ]; then
    kill `cat $file`
    rm $file
fi

user $ tail -n 11 unmount_FREECOM_HDD.sh
#----------------------------------------------------------------------------------------------------
#
# This forms part of the scheme to provide automatic suspension while the greeter screen is displayed
file="/tmp/unique_identifier"
if [ -f "$file" ]; then
    kill `cat $file`
    rm $file
fi
#
#----------------------------------------------------------------------------------------------------
exit 0

If the machine did not already have a permanently-connected external USB HDD (LABEL=”FREECOM HDD”) then it would have sufficed to specify a script named, for example, lightdm_kill_sleep2.sh instead of unmount_FREECOM_HDD.sh:

user $ cat 10_lubuntu.conf 
[Seat:*]
greeter-setup-script=/etc/lightdm/lightdm.conf.d/lightdm_sleep.sh
session-setup-script=/etc/lightdm/lightdm.conf.d/lightdm_kill_sleep.sh
session-cleanup-script=/etc/lightdm/lightdm.conf.d/lightdm_kill_sleep2.sh
user $ cat lightdm_kill_sleep2.sh
#!/bin/bash
# This forms part of the scheme to provide automatic suspension while the greeter screen is displayed
file="/tmp/unique_identifier"
if [ -f "$file" ]; then
    kill `cat $file`
    rm $file
fi

user $ sudo nano sessions_sleep.sh
user $ sudo chmod +x sessions_sleep.sh
user $ cat sessions_sleep.sh 
#!/bin/bash
date +%s > /tmp/datetime_suspended # Initialise variable
while true
do
    # Only monitor idle time and suspend after specified inactivity if lightdm_sleep.sh is not taking care of those
    if [[ `ps -ef | grep bash | grep lightdm_sleep.sh | wc -l` -eq 0 ]]; then
        #-------------------------------STAGE 1: FIND OUT WHO IS THE ACTIVE USER--------------------------------------
        #
        while IFS=: read -r f1 f2 f3 f4 f5 f6 f7
        # $f1 is username
        # $f2 is password ('x')
        # $f3 is UID
        # $f4 is GID
        # $f5 is UID info
        # $f6 is home directory
        # $f7 is command/shell
        do
            if [[ $f6 == *"/home/"* ]] && [[ $f7 == "/bin/bash" ]]; then
                if `loginctl list-users | grep -ve '^$\|USER\|listed' | awk -F' ' '{print $2}' | grep -q $f1`; then
                    state=`loginctl show-user $f3 | grep State | awk -F'=' '{print $2}'`
                    if [[ $state != "active" ]]; then
                        inactive_user=$f1
                    elif [[ $state == "active" ]]; then
                        active_user=$f1
                    fi
                fi
            fi
        done < /etc/passwd
        #
        #-------------------------------STAGE 2: ASCERTAIN USER SESSIONS---------------------------------------------
        #
        # Find idle time for each X Windows session and suspend to RAM if the active user has been idle for >=30min.
        #
        who -u | grep -v "\." > /tmp/logged-in_users
        #
        while read a b c d e f g
        # $a is username
        # $b is the tty (tty1 to tty12)
        # $c is the date (yyyy-mm-dd)
        # $d is the time (hh:mm)
        # $e is the idle time (hh:mm) which does not reflect reality in this installation, for some reason
        # $f is the PID
        # $g is the display e.g. "(:1)"
        # Example: "john     tty7         2019-08-31 17:08 00:01        1624 (:0)"
        do
            if [[ $(echo $b | sed 's/[^0-9]*//g') -gt 6 ]]; then
                display=$(echo $g | sed 's/[^0-9]*//g')
                idle_millisecs=$(env DISPLAY=:$display sudo -u $a xprintidle)
                let idle_minutes=$idle_millisecs/60000
                if [[ $idle_minutes -ge 30 ]] && [[ $a == "$active_user" ]]; then
                    datetime_now=$(date +%s)
                    diffsecs=$(expr $datetime_now - $(cat /tmp/datetime_suspended))
                    # Prevent suspending immediately after resuming
                    if [ $diffsecs -gt 180 ]; then
                        date +%s > /tmp/datetime_suspended
                        systemctl suspend
                    fi
                fi
            fi
        done < /tmp/logged-in_users
        rm /tmp/logged-in_users
        #
        #------------------------------------------------------------------------------------------------------------
        sleep 10 # Frequency to repeat check
    fi
done

I installed the utility xprintidle via the Linux distribution’s package manager. As the name of the utility suggests, it returns the time (in milliseconds) that an X Windows session has been idle. Nice utility, by the way.

user $ sudo crontab -e
user $ sudo crontab -l | grep -v ^#
@reboot sudo /etc/lightdm/lightdm.conf.d/sessions_sleep.sh

Note that, despite its name, ‘@reboot‘ in the cron job will run the script after a cold boot as well as after a warm boot (reboot). Also note that the use of ‘sudo‘ in the root cron job is not an error; it makes the root cron job use the root user’s environment variables.

Remember that the property /xfce4-power-manager/inactivity-on-ac has to be configured to have a value of 14 (which corresponds to ‘Never’ in the Xfce Power Manager settings GUI) for every user. This should be done by each user using the Xfce Power Manager settings GUI in their own session.

Basically, the scheme works as follows: At boot, Lubuntu 18.04 launches the looping Bash script sessions_sleep.sh, which remains running but does nothing because no X Windows users are logged in. When LightDM runs the greeter-setup-script (lightdm_sleep.sh) and displays the greeter screen, sessions_sleep.sh still does nothing while lightdm_sleep.sh is running and taking care of managing suspension. When an X Windows user logs in and LightDM’s session-setup-script (lightdm_kill_sleep.sh) kills the running script lightdm_sleep.sh, the script sessions_sleep.sh then takes over monitoring users’ activity in X Windows and triggers suspension if the active user has not used his/her session for 30 minutes. If an X Windows user logs out, LightDM’s session-cleanup-script (unmount_FREECOM_HDD.sh) also kills lightdm_sleep.sh if it is running. When LightDM again runs its greeter-setup-script (lightdm_sleep.sh) and displays the greeter screen, that again inhibits sessions_sleep.sh from taking any action if no X Windows user is logged in. This all sounds convoluted, but it seems to work fine so far.

Because Xfce Power Manager is no longer used to monitor idle time and trigger suspension, ‘Presentation mode’ in Xfce Power Manager can no longer prevent the system from suspending after 30 minutes of inactivity while someone is watching a long video or playing music, for example. However this is not a problem; to temporarily inhibit suspension the user can use the method given in my earlier post ‘How to move a mouse pointer automatically in Linux to simulate user activity‘.

The Lubuntu 18.04 architecture

I suspect most Lubuntu 18.04 installations are on laptops or desktop machines with a single user, i.e. single-seat, single-user installations. In such a case, unless the user has created multiple user accounts that he/she logs into concurrently (by using ‘Switch User’, for example), the machine will never suspend unexpectedly while the user is logged in and using the session. I think the way LightDM, light-locker, systemd-logind and Xfce Power Manager have been bundled in Lubuntu 18.04 to manage suspending to RAM is a dog’s breakfast. The design apparently does not take into consideration that different people could be logged in concurrently in a single-seat installation. Try forcing people to log off so that only one person is ever logged in — it won’t happen! To be interrupted by Suspend to RAM triggered by Xfce Power Manager due to inactivity in a different session is illogical; the system should not suspend when someone is actively using the system. Therefore, in my opinion, management of suspension (and hibernation) ought to be configured and managed system-wide, not on a per-user basis, and a design should not require users to hack the installation to the extent I have described above. I was ‘scratching an itch’, but users should not have to jump through hoops to get an installation to function in a sensible manner. For all I know there may be a simpler way of achieving the functionality in Lubuntu 18.04 that I have described in this post and my previous post, but, if there is, it is not obvious. LightDM, light-locker, systemd and Xfce Power Manager are developed by different people, and functionality such as suspension and hibernation does not seem to have been considered using ‘helicopter vision’. Designing disparate applications developed separately to work together holistically is not a trivial task.

Anyway, hopefully I have fixed the problem and also ‘scratched my itch’. No more unexpected suspensions while I am using the family desktop machine!