]> err.no Git - systemd/log
systemd
12 years agodelta: introduce arg_flags field, to follow our usual scheme
Lennart Poettering [Mon, 21 May 2012 00:01:52 +0000 (02:01 +0200)]
delta: introduce arg_flags field, to follow our usual scheme

12 years agodelta: use same nomenclature for equivalent and redirected everywhere
Lennart Poettering [Wed, 16 May 2012 16:57:57 +0000 (18:57 +0200)]
delta: use same nomenclature for equivalent and redirected everywhere

12 years agodelta: enums are much cooler than defines
Lennart Poettering [Wed, 16 May 2012 16:53:46 +0000 (18:53 +0200)]
delta: enums are much cooler than defines

12 years agodelta: Support filtering what type of deltas to show
Gergely Nagy [Wed, 16 May 2012 16:11:27 +0000 (18:11 +0200)]
delta: Support filtering what type of deltas to show

Not everyone is interested in every kind of deltas (and some might
even be interested knowing which files do not have overrides), so this
here is an implementation of a --type=LIST... option for
systemd-delta, that makes it possible to filter what subset of deltas
we want.

The available modifiers are masked, equivalent, redirected, overriden,
and unchanged - they should be self explanatory, and the man page
explains them in a little more detail anyway.

As a side effect, in case of overriden files, the diff output was made
optional.

By default, everything is shown (with a diff, if appropriate) except
for completely unchanged files.

Signed-off-by: Gergely Nagy <algernon@balabit.hu>
12 years agodelta: add preset dirs
Lennart Poettering [Wed, 16 May 2012 01:39:41 +0000 (03:39 +0200)]
delta: add preset dirs

12 years agodelta: add missing files
Lennart Poettering [Wed, 16 May 2012 01:36:42 +0000 (03:36 +0200)]
delta: add missing files

12 years agoswitch-root: check for absolute paths
Lennart Poettering [Wed, 16 May 2012 01:23:54 +0000 (03:23 +0200)]
switch-root: check for absolute paths

12 years agodelta: add systemd-delta tool to find overriden configuration and unit files
Lennart Poettering [Wed, 16 May 2012 01:23:28 +0000 (03:23 +0200)]
delta: add systemd-delta tool to find overriden configuration and unit files

12 years agoudev: update docs, NEWS
Kay Sievers [Tue, 15 May 2012 21:44:28 +0000 (23:44 +0200)]
udev: update docs, NEWS

12 years agoudev: libudev - bump soname, remove deprecated functions, introduce symbol versions
Kay Sievers [Tue, 15 May 2012 20:57:58 +0000 (22:57 +0200)]
udev: libudev - bump soname, remove deprecated functions, introduce symbol versions

12 years agobuild-sys: support build-from-git without gobject-introspection
Kay Sievers [Tue, 15 May 2012 13:05:36 +0000 (15:05 +0200)]
build-sys: support build-from-git without gobject-introspection

12 years agobuild-sys: fix distcheck
Kay Sievers [Tue, 15 May 2012 12:52:53 +0000 (14:52 +0200)]
build-sys: fix distcheck

12 years agotmpfiles: if we are supposed to write a string to a file, it's OK if we can't write...
Lennart Poettering [Tue, 15 May 2012 12:34:33 +0000 (14:34 +0200)]
tmpfiles: if we are supposed to write a string to a file, it's OK if we can't write the trailing newline

12 years agobuild-sys: fix build
Lennart Poettering [Tue, 15 May 2012 12:34:06 +0000 (14:34 +0200)]
build-sys: fix build

12 years agobuild-sys: fix build for folks without gtk-doc installed
Lennart Poettering [Tue, 15 May 2012 11:55:48 +0000 (13:55 +0200)]
build-sys: fix build for folks without gtk-doc installed

12 years agogit: update gitignore
Lennart Poettering [Tue, 15 May 2012 11:54:05 +0000 (13:54 +0200)]
git: update gitignore

12 years agobuild-sys: move gtk-doc stuff into its own subdir and reindent autogen.sh
Lennart Poettering [Tue, 15 May 2012 11:53:05 +0000 (13:53 +0200)]
build-sys: move gtk-doc stuff into its own subdir and reindent autogen.sh

We moved most other build noise files into their own subdirs, so let's
do this for gtk-doc too

12 years agoman: overwrite vs. override
Michal Schmidt [Tue, 15 May 2012 06:16:30 +0000 (08:16 +0200)]
man: overwrite vs. override

'overwrite' conveys the idea of the old text being destroyed.
'override' does not. It is more similar to 'overrule'.

12 years agoinstall: fix inverted meaning of '--force' in systemctl enable
Michal Schmidt [Tue, 15 May 2012 06:12:49 +0000 (08:12 +0200)]
install: fix inverted meaning of '--force' in systemctl enable

12 years agomissing: Fix broken syscall(__NR_fanotify_mark... on ppc32
Andreas Schwab [Tue, 8 May 2012 17:18:26 +0000 (19:18 +0200)]
missing: Fix broken syscall(__NR_fanotify_mark... on ppc32

The same situation as on o32 mips.

12 years agosd-login: update header docs a bit
Lennart Poettering [Mon, 14 May 2012 20:41:17 +0000 (22:41 +0200)]
sd-login: update header docs a bit

12 years agonamespace: make PrivateTmp= apply to both /tmp and /var/tmp
Lennart Poettering [Fri, 11 May 2012 15:56:09 +0000 (17:56 +0200)]
namespace: make PrivateTmp= apply to both /tmp and /var/tmp

12 years agoupdate TODO
Lennart Poettering [Fri, 11 May 2012 15:46:58 +0000 (17:46 +0200)]
update TODO

12 years agosystemctl: add switch-root verb
Lennart Poettering [Fri, 11 May 2012 15:35:46 +0000 (17:35 +0200)]
systemctl: add switch-root verb

12 years agocore: add extra safety check before switching root
Lennart Poettering [Fri, 11 May 2012 15:17:57 +0000 (17:17 +0200)]
core: add extra safety check before switching root

12 years agojob: info message if JOB_VERIFY_ACTIVE detects an inactive unit
Michal Schmidt [Sun, 13 May 2012 21:29:19 +0000 (23:29 +0200)]
job: info message if JOB_VERIFY_ACTIVE detects an inactive unit

12 years agojob: report the status of first half of JOB_RESTART the same as JOB_STOP
Michal Schmidt [Sun, 13 May 2012 21:28:46 +0000 (23:28 +0200)]
job: report the status of first half of JOB_RESTART the same as JOB_STOP

12 years agounit: unit type dependent status messages
Michal Schmidt [Sun, 13 May 2012 16:18:54 +0000 (18:18 +0200)]
unit: unit type dependent status messages

Instead of generic "Starting..." and "Started" messages for all unit use
type-dependent messages. For example, mounts will announce "Mounting..."
and "Mounted".

Add status messages to units of types that used to be entirely silent
(automounts, sockets, targets, devices). For unit types whose jobs are
instantaneous, report only the job completion, not the starting event.
Socket units with non-instantaneous jobs are rare (Exec*= is not used
often in socket units), so I chose not to print the starting messages
for them either.

This will hopefully give people better understanding of the boot.

12 years agounit: print the color status marks on the left
Michal Schmidt [Mon, 14 May 2012 10:50:33 +0000 (12:50 +0200)]
unit: print the color status marks on the left

The alignment of the "[  OK  ]" and "[FAILED]" status marks to the right
side of the terminal makes it difficult to link them with the messages
on the left if your console is wide.

I considered the options:
 1. Align them to the 80th column regardless of the console width.
    Disadvantage - either:
    - truncating messages needlessly, not using available space; or
    - If the message is long, write the mark over it. => ugly
 2. Write them to the 80th column for short messages,
    and further to the right for longer ones.
    Disadvantage:
    - jagged look
 3. Write the marks on the left, before the message.
    Disadvantage:
    - Breaks tradition from RHL.
    Advantages:
    + slightly simpler code
    + Will annoy holy-traditionalists.

I chose option 3.
BTW, Debian now uses similar marks on the left with its makefile-style
boot.

Special values of the "status" argument to status_vprintf are:
  NULL - no status mark, no message indentation
  ""   - no status mark, message indented as if the mark was there

12 years agojob: change red [ABORT] status to yellow [DEPEND]
Michal Schmidt [Mon, 14 May 2012 10:23:23 +0000 (12:23 +0200)]
job: change red [ABORT] status to yellow [DEPEND]

The red "[ABORT]" for a dependency failure is too scary.
It suggests a crash. And it suggests a problem with the unit itself.
Change it to a yellow "[DEPEND]" message. The color communicates the
level of seriousness better.

12 years agojob: only jobs on the runqueue can be run
Michal Schmidt [Sat, 12 May 2012 19:06:27 +0000 (21:06 +0200)]
job: only jobs on the runqueue can be run

12 years agounits: do not quit plymouth too early
Michal Schmidt [Sat, 12 May 2012 20:01:27 +0000 (22:01 +0200)]
units: do not quit plymouth too early

12 years agokeymap: Add Mio N890
Martin Pitt [Mon, 14 May 2012 11:06:48 +0000 (13:06 +0200)]
keymap: Add Mio N890

https://launchpad.net/bugs/998621

12 years agoupdate mailmap for Martin Pitt
Martin Pitt [Mon, 14 May 2012 11:06:20 +0000 (13:06 +0200)]
update mailmap for Martin Pitt

12 years agoudev: path_id - cciss - set "we have a bus" flag
Kay Sievers [Thu, 10 May 2012 19:35:54 +0000 (21:35 +0200)]
udev: path_id - cciss - set "we have a bus" flag

12 years agoudev: path_id - add cciss support
Hannes Reinecke [Thu, 10 May 2012 11:18:07 +0000 (13:18 +0200)]
udev: path_id - add cciss support

12 years agodbus-manager: fix tainted string
Michal Schmidt [Wed, 9 May 2012 19:42:56 +0000 (21:42 +0200)]
dbus-manager: fix tainted string

The pointer to the end of the string was not advanced after adding
the "cgroups-missing" taint. If "local-hwclock" was detected too,
it would overwrite the previous string.

With 'e' always pointing to the end of the string, removing the last
delimiter is easier.

12 years agomain: fix uninitialized variable
Michal Schmidt [Wed, 9 May 2012 09:20:06 +0000 (11:20 +0200)]
main: fix uninitialized variable

12 years agomain: fix assertion failure due to use of ELEMENTSOF on a non-array
Michal Schmidt [Wed, 9 May 2012 09:13:26 +0000 (11:13 +0200)]
main: fix assertion failure due to use of ELEMENTSOF on a non-array

12 years agobash-completion: use printf instead of echo
Michal Schmidt [Wed, 9 May 2012 07:36:20 +0000 (09:36 +0200)]
bash-completion: use printf instead of echo

If the expanded argument begins with '-', echo would try to interpret
it as options.
printf does not have this problem. Use it everywhere.

12 years agobash-completion: avoid losing backslashes in unit names
Michal Schmidt [Wed, 9 May 2012 07:18:44 +0000 (09:18 +0200)]
bash-completion: avoid losing backslashes in unit names

Use 'read -r' everywhere to consider backslashes as parts of the input line.
Single-quote the arguments to 'compgen -W' to avoid immediate expansion.
compgen itself will expand the argument.

Fixes a possible reason for "Failed to issue method call: Unknown unit"
after requesting completion.

https://bugzilla.redhat.com/show_bug.cgi?id=814966

12 years agoupdate TODO
Lennart Poettering [Tue, 8 May 2012 23:26:30 +0000 (01:26 +0200)]
update TODO

12 years agoutil: a few updates for rm_rf()
Lennart Poettering [Tue, 8 May 2012 23:25:52 +0000 (01:25 +0200)]
util: a few updates for rm_rf()

12 years agomanager: introduce SwitchRoot bus call for initrd/main transition
Lennart Poettering [Tue, 8 May 2012 23:24:50 +0000 (01:24 +0200)]
manager: introduce SwitchRoot bus call for initrd/main transition

12 years agopath-util: there is no function path_parent()
Lennart Poettering [Tue, 8 May 2012 21:37:25 +0000 (23:37 +0200)]
path-util: there is no function path_parent()

12 years agomain: simplify+unify logic for parsing runtime booleans
Lennart Poettering [Tue, 8 May 2012 21:36:55 +0000 (23:36 +0200)]
main: simplify+unify logic for parsing runtime booleans

12 years agologind: implement suspend/hibernate calls with inhibition logic
Lennart Poettering [Tue, 8 May 2012 17:02:25 +0000 (19:02 +0200)]
logind: implement suspend/hibernate calls with inhibition logic

12 years agosleep: implement suspend/hibernate as first class targets
Lennart Poettering [Sat, 5 May 2012 00:06:58 +0000 (02:06 +0200)]
sleep: implement suspend/hibernate as first class targets

12 years agologind: use "sleep" as generic term for "suspend", "hibernate", and later on "hybrid...
Lennart Poettering [Fri, 4 May 2012 22:46:31 +0000 (00:46 +0200)]
logind: use "sleep" as generic term for "suspend", "hibernate", and later on "hybrid suspend"

12 years agologind: fix test-inhibit
Lennart Poettering [Fri, 4 May 2012 22:39:24 +0000 (00:39 +0200)]
logind: fix test-inhibit

12 years agoautogen.sh: undef _FORTIFY_SOURCE, which now logs warnings when used with -O0
Kay Sievers [Tue, 8 May 2012 00:51:26 +0000 (02:51 +0200)]
autogen.sh: undef _FORTIFY_SOURCE, which now logs warnings when used with -O0

12 years agoutil: split-out path-util.[ch]
Kay Sievers [Mon, 7 May 2012 19:36:12 +0000 (21:36 +0200)]
util: split-out path-util.[ch]

12 years agoutil: split-out hwclock.[ch]
Kay Sievers [Mon, 7 May 2012 19:06:55 +0000 (21:06 +0200)]
util: split-out hwclock.[ch]

12 years agologind: fix memory leak
Frederic Crozat [Fri, 4 May 2012 14:14:19 +0000 (16:14 +0200)]
logind: fix memory leak

12 years agoupdate TODO
Lennart Poettering [Mon, 7 May 2012 20:20:08 +0000 (22:20 +0200)]
update TODO

12 years agoUpdate TODO
Lennart Poettering [Mon, 7 May 2012 15:16:50 +0000 (17:16 +0200)]
Update TODO

12 years agoutil: split-out conf-file.[ch]
Kay Sievers [Mon, 7 May 2012 16:55:45 +0000 (18:55 +0200)]
util: split-out conf-file.[ch]

12 years agoudev: path_id - skip PCI-only-parents for block devices
Kay Sievers [Mon, 7 May 2012 16:47:58 +0000 (18:47 +0200)]
udev: path_id - skip PCI-only-parents for block devices

12 years agoudev: use conf_files_list() to search rules files
Kay Sievers [Mon, 7 May 2012 11:21:05 +0000 (13:21 +0200)]
udev: use conf_files_list() to search rules files

12 years agoconf_files_list(): split out conf_files_list_strv()
Kay Sievers [Mon, 7 May 2012 11:20:29 +0000 (13:20 +0200)]
conf_files_list(): split out conf_files_list_strv()

12 years agoconf_files_list(): files-add() - do not canonicalize file names
Kay Sievers [Mon, 7 May 2012 11:15:25 +0000 (13:15 +0200)]
conf_files_list(): files-add() - do not canonicalize file names

File names in /etc, /run, /usr/lib are sorted/overridden by basename.
Sorting things like "/dev/null" with the basename "null" in the hash
of config files breaks the ordering and the overriding logic.

12 years agologind: implement delay inhibitor locks in addition to block inhibitor locks
Lennart Poettering [Fri, 4 May 2012 22:34:48 +0000 (00:34 +0200)]
logind: implement delay inhibitor locks in addition to block inhibitor locks

This is useful to allow applications to synchronously save data before
the system is suspended or shut down.

12 years agoupdate TODO
Kay Sievers [Fri, 4 May 2012 17:06:07 +0000 (19:06 +0200)]
update TODO

12 years agoudev: fix the ENV{} blacklist logic
Kay Sievers [Fri, 4 May 2012 00:26:08 +0000 (02:26 +0200)]
udev: fix the ENV{} blacklist logic

12 years agosystemctl: fix typo
Lennart Poettering [Thu, 3 May 2012 23:02:32 +0000 (01:02 +0200)]
systemctl: fix typo

as Sergey Ptashnick pointed out we should write "poweroff" once instead
of "reboot" twice.

12 years agosystemd-analyze: add a --user option to support user instances
Patrick McCarty [Thu, 3 May 2012 20:42:45 +0000 (13:42 -0700)]
systemd-analyze: add a --user option to support user instances

12 years agoman: clarify _TRANSPORT
Shawn Landden [Thu, 3 May 2012 22:19:01 +0000 (15:19 -0700)]
man: clarify _TRANSPORT

"resp." where it was confuses reader that stdout and stderr might go to differn't places

12 years agoupdate TODO
Lennart Poettering [Thu, 3 May 2012 22:43:36 +0000 (00:43 +0200)]
update TODO

12 years agoreadahead: store inode numbers in pack file
Lennart Poettering [Thu, 3 May 2012 22:34:12 +0000 (00:34 +0200)]
readahead: store inode numbers in pack file

If the inode nr for each file is available in the pack file we can
easily detect replaced files (like they result from package upgrades)
which we can then skip to readahead.

12 years agounits: use OOMScoreAdjust= in the unit files to set OOM score adjust
Lennart Poettering [Thu, 3 May 2012 22:15:21 +0000 (00:15 +0200)]
units: use OOMScoreAdjust= in the unit files to set OOM score adjust

12 years agoreadhead: temporarily lower the kernel's read_ahead_kb setting while collecting
Lennart Poettering [Thu, 3 May 2012 22:13:20 +0000 (00:13 +0200)]
readhead: temporarily lower the kernel's read_ahead_kb setting while collecting

While collecting readahead data we want to know exactly what userspace
accesses unblurred by the kernel's read_ahead_kb. Hence lower this
during collection, and raise it afterwards.

This is mostly based on ideas and code by Auke Kok.

12 years agocgroup: fix alloca() misuse in cg_shorten_controllers()
Lennart Poettering [Thu, 3 May 2012 21:23:38 +0000 (23:23 +0200)]
cgroup: fix alloca() misuse in cg_shorten_controllers()

12 years agodbus: handle invalid enum values better
Lennart Poettering [Thu, 3 May 2012 21:06:19 +0000 (23:06 +0200)]
dbus: handle invalid enum values better

12 years agodbus: include unit name in JobNew/JobRemoved signals
Lennart Poettering [Thu, 3 May 2012 20:53:25 +0000 (22:53 +0200)]
dbus: include unit name in JobNew/JobRemoved signals

This breaks D-Bus interface slightly, but since the D-Bus API isn't
covered by the interface stability promise this should be OK.

12 years agoservice: explicitly remove control/ subcgroup after each control command
Lennart Poettering [Thu, 3 May 2012 19:54:44 +0000 (21:54 +0200)]
service: explicitly remove control/ subcgroup after each control command

The kernel will only notify us of cgroups running empty if no subcgroups
exist anymore. Hence make sure we don't leave our own control/ subcgroup
around longer than necessary.

https://bugzilla.redhat.com/show_bug.cgi?id=818381

12 years agosystemctl: allow 'systemctl reboot -ff' to succeed even without D-Bus
Lennart Poettering [Thu, 3 May 2012 14:29:15 +0000 (16:29 +0200)]
systemctl: allow 'systemctl reboot -ff' to succeed even without D-Bus

12 years agosystemctl: print a nice error message if an unprivileged user calls systemctl reboot -ff
Lennart Poettering [Thu, 3 May 2012 14:17:58 +0000 (16:17 +0200)]
systemctl: print a nice error message if an unprivileged user calls systemctl reboot -ff

12 years agosystemctl: get rid of arg_immediate and fold it into arg_force
Lennart Poettering [Thu, 3 May 2012 14:17:39 +0000 (16:17 +0200)]
systemctl: get rid of arg_immediate and fold it into arg_force

12 years agorc-local-generator: hook halt-local in based on generator too
Lennart Poettering [Thu, 3 May 2012 13:55:38 +0000 (15:55 +0200)]
rc-local-generator: hook halt-local in based on generator too

12 years agohwclock: add taint flag for non-local hwclock
Lennart Poettering [Thu, 3 May 2012 13:46:04 +0000 (15:46 +0200)]
hwclock: add taint flag for non-local hwclock

12 years agovconsole: fix some error messages
Lennart Poettering [Thu, 3 May 2012 13:10:09 +0000 (15:10 +0200)]
vconsole: fix some error messages

12 years agounits: explicit Type=dbus is now redundant
Lennart Poettering [Thu, 3 May 2012 12:48:35 +0000 (14:48 +0200)]
units: explicit Type=dbus is now redundant

12 years agoservice: default to Type=dbus if BusName= is specified
Lennart Poettering [Thu, 3 May 2012 12:46:29 +0000 (14:46 +0200)]
service: default to Type=dbus if BusName= is specified

12 years agoservice: warn if a dbus name is specified but the service type is not dbus
Lennart Poettering [Thu, 3 May 2012 12:42:49 +0000 (14:42 +0200)]
service: warn if a dbus name is specified but the service type is not dbus

12 years agovconsole: fix error messages
Lennart Poettering [Thu, 3 May 2012 00:28:26 +0000 (02:28 +0200)]
vconsole: fix error messages

https://bugs.freedesktop.org/show_bug.cgi?id=49192

12 years agoupdate TODO
Lennart Poettering [Thu, 3 May 2012 00:25:49 +0000 (02:25 +0200)]
update TODO

12 years agoman: fix typo
Kay Sievers [Mon, 30 Apr 2012 13:30:57 +0000 (15:30 +0200)]
man: fix typo

12 years agounits: make sure /var is writable before initializing random seed
Lennart Poettering [Mon, 30 Apr 2012 08:56:46 +0000 (10:56 +0200)]
units: make sure /var is writable before initializing random seed

12 years agoTODO
Lennart Poettering [Mon, 30 Apr 2012 08:49:39 +0000 (10:49 +0200)]
TODO

12 years agounit: add new dependency type RequiresMountsFor=
Lennart Poettering [Sun, 29 Apr 2012 12:26:07 +0000 (14:26 +0200)]
unit: add new dependency type RequiresMountsFor=

RequiresMountsFor= is a shortcut for adding requires and after
dependencies to all mount units neeed for the specified paths.

This solves a couple of issues regarding dep loop cycles for encrypted
swap.

12 years agoservice: document new Type=idle
Lennart Poettering [Thu, 26 Apr 2012 18:32:26 +0000 (20:32 +0200)]
service: document new Type=idle

12 years agoupdate TODO
Lennart Poettering [Thu, 26 Apr 2012 18:32:07 +0000 (20:32 +0200)]
update TODO

12 years agoupdate TODO
Kay Sievers [Sun, 29 Apr 2012 22:46:38 +0000 (00:46 +0200)]
update TODO

12 years agoudev: use container_of() macro
Lucas De Marchi [Thu, 26 Apr 2012 16:36:02 +0000 (13:36 -0300)]
udev: use container_of() macro

12 years agoutil: introduce container_of() macro
Lucas De Marchi [Thu, 26 Apr 2012 16:36:01 +0000 (13:36 -0300)]
util: introduce container_of() macro

This macro comes from kernel and it's useful for unwrapping structs
inside another one. The generated code is actually the same to the one
where this logic is used in udev, but using this macro is much cleaner
and less error prone.

12 years agolibudev: prefix log macros with 'udev_'
Kay Sievers [Thu, 26 Apr 2012 15:43:48 +0000 (17:43 +0200)]
libudev: prefix log macros with 'udev_'

12 years agoupdate TODO
Kay Sievers [Thu, 26 Apr 2012 00:03:31 +0000 (02:03 +0200)]
update TODO

12 years agoudev: use only oom_score_adj interface
Kay Sievers [Wed, 25 Apr 2012 23:49:15 +0000 (01:49 +0200)]
udev: use only oom_score_adj interface

12 years agoudev: docs - it's 'GUdev'
Kay Sievers [Wed, 25 Apr 2012 23:48:24 +0000 (01:48 +0200)]
udev: docs - it's 'GUdev'

12 years agocore: add NOP jobs, job type collapsing
Michal Schmidt [Wed, 25 Apr 2012 09:58:27 +0000 (11:58 +0200)]
core: add NOP jobs, job type collapsing

Two of our current job types are special:
JOB_TRY_RESTART, JOB_RELOAD_OR_START.

They differ from other job types by being sensitive to the unit active state.
They perform some action when the unit is active and some other action
otherwise. This raises a question: when exactly should the unit state be
checked to make the decision?

Currently the unit state is checked when the job becomes runnable. It's more
sensible to check the state immediately when the job is added by the user.
When the user types "systemctl try-restart foo.service", he really intends
to restart the service if it's running right now. If it isn't running right
now, the restart is pointless.

Consider the example (from Bugzilla[1]):

sleep.service takes some time to start.
hello.service has After=sleep.service.
Both services get started. Two jobs will appear:
 hello.service/start  waiting
 sleep.service/start  running
Then someone runs "systemctl try-restart hello.service".

Currently the try-restart operation will block and wait for
sleep.service/start to complete.

The correct result is to complete the try-restart operation immediately
with success, because hello.service is not running. The two original
jobs must not be disturbed by this.

To fix this we introduce two new concepts:
- a new job type: JOB_NOP
  A JOB_NOP job does not do anything to the unit. It does not pull in any
  dependencies. It is always immediately runnable. When installed to a unit,
  it sits in a special slot (u->nop_job) where it never conflicts with
  the installed job (u->job) of a different type. It never merges with jobs
  of other types, but it can merge into an already installed JOB_NOP job.

- "collapsing" of job types
  When a job of one of the two special types is added, the state of the unit
  is checked immediately and the job type changes:
  JOB_TRY_RESTART     -> JOB_RESTART or JOB_NOP
  JOB_RELOAD_OR_START -> JOB_RELOAD or JOB_START
  Should a job type JOB_RELOAD_OR_START appear later during job merging, it
  collapses immediately afterwards.
  Collapsing actually makes some things simpler, because there are now fewer
  job types that are allowed in the transaction.

[1] Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=753586