This adds support for executing systemctl operations remotely or as
privileged user while still running systemctl itself unprivileged and
locally.
This currently requires a D-Bus patch to work properly.
https://bugs.freedesktop.org/show_bug.cgi?id=35230
This merges several separate patches that I carry as part of
Mandriva systemd RPM. They touch those parts that are very
unlikely to be changed in near future and do not impose any
functionality change for systemd core. I also think it is
useful for troubleshooting to have real distribution name in
system logs, espicially when someone reports problem upstream.
The patch looks bigger than sum of replaced patches because
- previous patches were applied on top of distro=fedora, now
I need to add all those bits for distro=mandriva as well
- part of patch was done as spec file magic, but it seems more
logical to ship all these bits together
Since we want to replace the bridge dynamically by the real syslog
implementation such as rsyslog we need to make sure that the the bridge
stays running right to the moment rsyslog is up so that we process
messages enqueued by other processes started before rsyslog, so that
those clients don't stay stuck.
This is supposed to play the same roles /var/lib/dbus/machine-id,
however fixes a couple of problems:
- It is available during early boot since it is stored in /etc
- Removes the ID from the D-Bus context and moves it into a system
context, thus hopefully lowering hesitation by people to use it.
- It is generated at installation time. If the file is empty at boot
time it will be mounted over with a randomly generated ID, which is
not saved to disk. This is useful to support state-less machines with
no transient or writable /etc configuration.
That unity pulls in OpenRC which in turn pulls in most of legacy
system that causes lots of troubles as it is too smart, thus not
recommended.
Moreover, SystemD developers seems to agree that a service file per DM
is the best approach, so having gdm.service, kdm.service, slim.service
is better than a single wrapper for them.
When cross-compiling systemd, the introspection XML files fail to be
generated because the systemd host binary is not executable. This patch
works around this by putting the introspection XML data into separate
ELF sections and extracting them from the binary when generating the XML
files.
The extracted XML data is passed through the strings utility in order to
strip the trailing NUL character. A small AWK script is used to prepend
the doctype and add the opening and closing node tags respectively.
Finally, the C preprocessor is used to substitute the correct doctype
information from the D-Bus header files.
With the introduction of native shutdown/reboot, the killall.service was
removed (as this functionality was moved into systemd-shutdown).
Without killall.service though, the umount*.service files no longer work
correctly.
Wit native mount support those files are also no longer necessary, so
remove them.
I've been playing recently with systemd on Arch, and had much fun. But
soon, alas, my fingers started to ache from repeatedly writing
systemctl restart some-long-service.service. So, I wrote a completion
script. I figured other people may want to use it, so I prepared a
patch against systemd-git (attached).
There are some notes/disclaimers, however:
- It requires bash>=4.0, sed, grep and awk. A bash-completion package
is not strictly needed; sourcing the file is enough.
- It wouldn't work properly with --session, as I had no way to test it.
- It uses the output of systemctl list-units directly when that's
enough, but also runs systemctl show when completing on some verbs
(for example, to check for AllowIsolate=yes). This /may/ be somewhat
slow once there are many units, since it calls a dbus method on each
one. Is there a faster way to have that information?
- The code is perhaps a bit long and messy; honestly, I blame the tool ;)
One way to improve on the situation is to integrate some completion
code in systemctl itself, the way e.g. gdbus, gsettings and django do
it. This will allow for finer grained and faster completions, and it
won't be necessary to keep the verb/option tables in sync with some
other file. But it does mean adding all of this code in C. If this is
acceptable, I'll try to have a go at it.
Finally, a couple of completion tips I run into:
- If you alias systemctl to, say, sctl, you get completions on that
too by running to following command:
complete -F _systemctl sctl
- Add the following line to your .inputrc, to have the completion show
after only a single tab press:
set show-all-if-ambiguous on
It makes the shell quite more pleasant.
Hope it's good enough!
Ran
