chiark / gitweb /
elogind.git
11 years agoman/os-release: Add BUILD_ID field
William Douglas [Mon, 25 Mar 2013 21:07:49 +0000 (14:07 -0700)]
man/os-release: Add BUILD_ID field

BUILD_ID is a fairly generic field used to identify the system image
that was used to install the distribution.

11 years agocatalog: fix error codes handling in catalog_list_items
Lukas Nykryn [Wed, 27 Mar 2013 16:35:55 +0000 (17:35 +0100)]
catalog: fix error codes handling in catalog_list_items

Previously r was set to zero and so if(r<0) was never true.

11 years agolog: fix error codes handling in catalog_list_items
Lukas Nykryn [Wed, 27 Mar 2013 09:44:21 +0000 (10:44 +0100)]
log: fix error codes handling in catalog_list_items

It does not make sense to print error code from previous loop.

11 years agosd-bus: check c->path for null instead of path
Lukas Nykryn [Wed, 27 Mar 2013 09:44:20 +0000 (10:44 +0100)]
sd-bus: check c->path for null instead of path

11 years agosystemd-sysctl: Handle missing /etc/sysctl.conf properly
Eelco Dolstra [Wed, 27 Mar 2013 12:41:59 +0000 (13:41 +0100)]
systemd-sysctl: Handle missing /etc/sysctl.conf properly

Since fabe5c0e5fce730aa66e10a9c4f9fdd443d7aeda, systemd-sysctl returns
a non-zero exit code if /etc/sysctl.conf does not exist, due to a
broken ENOENT check.

11 years agoshared: free dt (temporary dir name) on fail
Václav Pavlín [Wed, 27 Mar 2013 14:16:37 +0000 (15:16 +0100)]
shared: free dt (temporary dir name) on fail

[zj: modified to not to try to rmdir() dir we haven't created.]

11 years agocore: check return value of rm_rf_dangerous and warn if it fails
Václav Pavlín [Wed, 27 Mar 2013 14:16:35 +0000 (15:16 +0100)]
core: check return value of rm_rf_dangerous and warn if it fails

11 years agobootctl: rename status output header
Kay Sievers [Wed, 27 Mar 2013 03:26:12 +0000 (04:26 +0100)]
bootctl: rename status output header

11 years agokeymap: Fix Touchpad Toggle on MSI Wind U90/U100
Martin Pitt [Wed, 27 Mar 2013 07:15:12 +0000 (08:15 +0100)]
keymap: Fix Touchpad Toggle on MSI Wind U90/U100

This key is handled by the hardware already, so handling it again in software
nullifies the effect. Newer kernels read the real state and send out a separate
KEY_TOUCHPAD_ON or KEY_TOUCHPAD_OFF event, so in both cases we need to ignore
that key.

https://bugs.freedesktop.org/show_bug.cgi?id=62404

11 years agotests: skip bus test if bus cannot be opened
Zbigniew Jędrzejewski-Szmek [Wed, 27 Mar 2013 01:07:46 +0000 (21:07 -0400)]
tests: skip bus test if bus cannot be opened

To make the result more visible, special return value
is used to tell automake that the test was skipped. While
at it, use the same return value in other skipped tests.

11 years agotests: fix size_t in format string
Zbigniew Jędrzejewski-Szmek [Wed, 27 Mar 2013 03:47:25 +0000 (23:47 -0400)]
tests: fix size_t in format string

11 years agoSimplify the meaning of %s
Zbigniew Jędrzejewski-Szmek [Tue, 26 Mar 2013 23:37:14 +0000 (19:37 -0400)]
Simplify the meaning of %s

The rules governing %s where just too complicated. First of
all, looking at $SHELL is dangerous. For systemd --system,
it usually wouldn't be set. But it could be set if the admin
first started a debug shell, let's say /sbin/sash, and then
launched systemd from it. This shouldn't influence how daemons
are started later on, so is better ignored. Similar reasoning
holds for session mode. Some shells set $SHELL, while other
set it only when it wasn't set previously (e.g. zsh). This
results in fragility that is better avoided by ignoring $SHELL
totally.

With $SHELL out of the way, simplify things by saying that
%s==/bin/sh for root, and the configured shell otherwise.
get_shell() is the only caller, so it can be inlined.

Fixes one issue seen with 'make check'.

11 years agounits: automatically order all mount units after network.target
Lennart Poettering [Wed, 27 Mar 2013 01:51:33 +0000 (02:51 +0100)]
units: automatically order all mount units after network.target

Previously it was necessary to pull in remote-fs-pre.target to order the
mount units against network.target since the ordering was done
transitively via remote-fs-pre.target.

As network implementations shouldn't need to know about the specific
use-case of network mounts we instead now simply order network.target
against all mounts too. This should make it unnecessary for network
managing services to import remote-fs-pre.target explicitly, as
network.target will now suffice.

11 years agoman: network.target is also a passive target
Lennart Poettering [Wed, 27 Mar 2013 01:49:50 +0000 (02:49 +0100)]
man: network.target is also a passive target

11 years agoreadahead: cleanups
Auke Kok [Tue, 26 Mar 2013 18:13:47 +0000 (11:13 -0700)]
readahead: cleanups

- check for OOM
- no need to use floats and round()

11 years agoreadahead: chunk on spinning media
Auke Kok [Fri, 22 Mar 2013 22:09:45 +0000 (15:09 -0700)]
readahead: chunk on spinning media

Readahead has all sorts of bad side effects depending on your
storage media. On rotating disks, it may be degrading startup
performance if enough requests are queued spanning linearly
over all blocks early at boot, and mount, blkid and friends
want to insert reads to the start of these block devices after.

The end result is that on spinning disks with ext3/4 that udev
and mounts take a very long time, and nothing really happens until
readahead is completely finished.

This has the net effect that the CPU is almost entirely idle
for the entire period that readahead is working. We could have
finished starting up quite a lot of services in this time if
we were smarter at how we do readahead.

This patch sorts all requests into 2 second "chunks" and sub-sorts
each chunk by block. This adds a single cross-drive seek per "chunk"
but has the benefit that we will have a lot of the blocks we need
early on in the boot sequence loaded into memory faster.

For a comparison of how before/after bootcharts look (ext4 on a
mobile 5400rpm 250GB drive) please look at:

    http://foo-projects.org/~sofar/blocked-tests/

There are bootcharts in the "before" and "after" folders where you
should be able to see that many low-level services finish 5-7
seconds earlier with the patch applied (after).

11 years agoNEWS: tweak grammar
Zbigniew Jędrzejewski-Szmek [Tue, 26 Mar 2013 15:56:01 +0000 (11:56 -0400)]
NEWS: tweak grammar

11 years agoNEWS: typo fix v199
Lennart Poettering [Tue, 26 Mar 2013 14:51:57 +0000 (15:51 +0100)]
NEWS: typo fix

11 years agobuild-sys: bump version and .so revisions
Lennart Poettering [Tue, 26 Mar 2013 14:43:43 +0000 (15:43 +0100)]
build-sys: bump version and .so revisions

11 years agocryptsetup: when prompting for password use GPT partition label
Lennart Poettering [Tue, 26 Mar 2013 14:23:54 +0000 (15:23 +0100)]
cryptsetup: when prompting for password use GPT partition label

If there's a GPT partition label set for a LUKS partition, then it's
nicer to show that than the model number, when asking for a passphrase.

11 years agoman: document that the passive units cannot be pulled in mnaully
Lennart Poettering [Tue, 26 Mar 2013 14:16:50 +0000 (15:16 +0100)]
man: document that the passive units cannot be pulled in mnaully

11 years agofstab-generator: drop x-initrd.rootfs mount option
Harald Hoyer [Tue, 26 Mar 2013 14:33:57 +0000 (15:33 +0100)]
fstab-generator: drop x-initrd.rootfs mount option

x-initrd.mount now has different meanings, if fstab-generator is called
in the initramfs.

initrd:/etc/fstab and x-initrd.mount defines mounts for the
initrd-root-fs.target

initrd:/sysroot/etc/fstab and x-initrd.mount defines mounts for the
initrd-fs.target

11 years agounits: disallow manual starting of passive units
Lennart Poettering [Tue, 26 Mar 2013 14:14:27 +0000 (15:14 +0100)]
units: disallow manual starting of passive units

As passive units only are useful for ordering things within the initial
transaction there is no point in ever activating them manually, hence
refuse it.

11 years agofstab-generator: degrade the message about missing "root=" to log_debug
Harald Hoyer [Tue, 26 Mar 2013 13:46:35 +0000 (14:46 +0100)]
fstab-generator: degrade the message about missing "root=" to log_debug

Some installation media (fedora at least) does not have and need a
"root=" argument on the kernel command line.

11 years agoDrop src/login/uaccess.c, dead code
Martin Pitt [Tue, 26 Mar 2013 10:35:17 +0000 (11:35 +0100)]
Drop src/login/uaccess.c, dead code

This moved to src/udev/udev-builtin-uaccess.c a while ago.

11 years agobus: fix missing variable initialization
Lennart Poettering [Tue, 26 Mar 2013 02:42:03 +0000 (03:42 +0100)]
bus: fix missing variable initialization

11 years agomodules-load: there's really no point in mentioning that a certain modules is already...
Lennart Poettering [Tue, 26 Mar 2013 02:13:41 +0000 (03:13 +0100)]
modules-load: there's really no point in mentioning that a certain modules is already loaded

After all, this runs in parallel to udev, so there's quite a chance it
already is....

11 years agobuild-sys: make gcc shut up
Lennart Poettering [Tue, 26 Mar 2013 02:03:13 +0000 (03:03 +0100)]
build-sys: make gcc shut up

11 years agobuild-sys: ship missing unit file
Lennart Poettering [Tue, 26 Mar 2013 01:48:25 +0000 (02:48 +0100)]
build-sys: ship missing unit file

11 years agounits: downgrade dependency on sockets.target/timers.target/paths.target by basic...
Lennart Poettering [Tue, 26 Mar 2013 01:32:20 +0000 (02:32 +0100)]
units: downgrade dependency on sockets.target/timers.target/paths.target by basic.target

There isn't really any need to require any targets but sysinit.target
from basic.target, so downgrade sockets.target, paths.target,
timers.target.

11 years agobus: when we are talking to a bus, SCM_CREDS/SCM_SECLABEL are not very useful
Lennart Poettering [Tue, 26 Mar 2013 00:58:14 +0000 (01:58 +0100)]
bus: when we are talking to a bus, SCM_CREDS/SCM_SECLABEL are not very useful

11 years agobus: automatically generate minimal introspection data to find installed objects
Lennart Poettering [Tue, 26 Mar 2013 00:54:37 +0000 (01:54 +0100)]
bus: automatically generate minimal introspection data to find installed objects

11 years agobus: rename send_hello flag to bus_cient
Lennart Poettering [Mon, 25 Mar 2013 23:36:43 +0000 (00:36 +0100)]
bus: rename send_hello flag to bus_cient

This way we can hide more than just the hello logic behind this flag,
for example, later on automatic match management.

11 years agobuild-sys: create kernel/install.d directories
Kay Sievers [Tue, 26 Mar 2013 00:00:14 +0000 (01:00 +0100)]
build-sys: create kernel/install.d directories

11 years agoNEWS: update
Kay Sievers [Mon, 25 Mar 2013 23:41:07 +0000 (00:41 +0100)]
NEWS: update

11 years agocryptsetup-generator: let's be a bit more efficient with strv_extend()
Lennart Poettering [Mon, 25 Mar 2013 22:49:41 +0000 (23:49 +0100)]
cryptsetup-generator: let's be a bit more efficient with strv_extend()

11 years agocryptsetup-generator: add a missing OOM check
Lennart Poettering [Mon, 25 Mar 2013 22:49:13 +0000 (23:49 +0100)]
cryptsetup-generator: add a missing OOM check

11 years agoupdate NEWS
Lennart Poettering [Mon, 25 Mar 2013 22:48:08 +0000 (23:48 +0100)]
update NEWS

11 years agoprepare NEWS for 199
Lennart Poettering [Mon, 25 Mar 2013 22:05:21 +0000 (23:05 +0100)]
prepare NEWS for 199

11 years agoman: properly document the system targets that are also available for the user instance
Lennart Poettering [Mon, 25 Mar 2013 21:32:59 +0000 (22:32 +0100)]
man: properly document the system targets that are also available for the user instance

11 years agoman: rearrange systemd.special(7) to clarify which units are passive and which ones...
Lennart Poettering [Mon, 25 Mar 2013 21:30:51 +0000 (22:30 +0100)]
man: rearrange systemd.special(7) to clarify which units are passive and which ones aren't

This also adds a short explanation paragraph for this.

11 years agounits: there is no point in pulling in ordering 'provides'-style targets
Lennart Poettering [Mon, 25 Mar 2013 21:10:32 +0000 (22:10 +0100)]
units: there is no point in pulling in ordering 'provides'-style targets

Units such as nss-lookup.target, nss-user-lookup.target,
remote-fs-pre.target, local-fs-pre.target, time-sync.target,
rpcbind.target are to be pulled in by the implementing services, and
that's there only purpose. They should not have any 'active component'
otherwise, so let's drop all further deps from these units.

11 years agoupdate TODO
Lennart Poettering [Mon, 25 Mar 2013 21:09:53 +0000 (22:09 +0100)]
update TODO

11 years agounits: introduce remote-fs-setup.target to pull in dependencies from remote mounts
Lennart Poettering [Mon, 25 Mar 2013 21:04:40 +0000 (22:04 +0100)]
units: introduce remote-fs-setup.target to pull in dependencies from remote mounts

This introduces remote-fs-setup.target independently of
remote-fs-pre.target. The former is only for pulling things in, the
latter only for ordering.

The new semantics:

remote-fs-setup.target: is pulled in automatically by all remote mounts.
Shall be used to pull in other units that want to run when at least one
remote mount is set up. Is not ordered against the actual mount units,
in order to allow activation of its dependencies even 'a posteriori',
i.e. when a mount is established outside of systemd and is only picked
up by it.

remote-fs-pre.target: needs to be pulled in automatically by the
implementing service, is otherwise not part of the initial transaction.
This is ordered before all remote mount units.

A service that wants to be pulled in and run before all remote mounts
should hence have:

a) WantedBy=remote-fs-setup.target -- so that it is pulled in

b) Wants=remote-fs-pre.target + Before=remote-fs-pre.target -- so that
   it is ordered before the mount point, normally.

11 years agoRevert "remote-fs.target: want remote-fs-pre.target"
Lennart Poettering [Mon, 25 Mar 2013 20:29:58 +0000 (21:29 +0100)]
Revert "remote-fs.target: want remote-fs-pre.target"

This reverts commit 6bde0b3220e95a08cefb70846f73b2cf24b7734a.

We should not pull in remote-fs-pre.target unconditionally. It's
supposed to be pulled in by the implementors of it, rather then its
users.

11 years agounits: and also, order all early-boot sockets before sockets.target
Lennart Poettering [Mon, 25 Mar 2013 20:25:06 +0000 (21:25 +0100)]
units: and also, order all early-boot sockets before sockets.target

11 years agounits: order all udev services before sysinit.target, too
Lennart Poettering [Mon, 25 Mar 2013 20:22:52 +0000 (21:22 +0100)]
units: order all udev services before sysinit.target, too

Not that it would matter much, but let's make things a bit more
systematic: early boot services shall order themselves before
sysinit.target, and nothing else.

11 years agoupdate TODO
Lennart Poettering [Mon, 25 Mar 2013 20:21:51 +0000 (21:21 +0100)]
update TODO

11 years agounits: introduce new timers.target and paths.target to hook timer/path units into...
Lennart Poettering [Mon, 25 Mar 2013 20:20:08 +0000 (21:20 +0100)]
units: introduce new timers.target and paths.target to hook timer/path units into for boot

11 years agofstab-generator: add missing strempty() calls
Mantas Mikulėnas [Mon, 25 Mar 2013 19:10:29 +0000 (21:10 +0200)]
fstab-generator: add missing strempty() calls

11 years agobuild-sys: add missing sed substitution for DEBUGTTY
Kay Sievers [Mon, 25 Mar 2013 17:49:55 +0000 (18:49 +0100)]
build-sys: add missing sed substitution for DEBUGTTY

11 years agobuild-sys: configurable debug shell tty path
Umut Tezduyar [Thu, 14 Mar 2013 13:58:40 +0000 (14:58 +0100)]
build-sys: configurable debug shell tty path

11 years agofstab-generator: fix minor memory leak on error path
Lennart Poettering [Mon, 25 Mar 2013 17:19:52 +0000 (18:19 +0100)]
fstab-generator: fix minor memory leak on error path

11 years agofstab-generator: add missing OOM check
Lennart Poettering [Mon, 25 Mar 2013 17:18:57 +0000 (18:18 +0100)]
fstab-generator: add missing OOM check

11 years agofstab-generator: rename x-initrd-rootfs.mount to x-initrd.rootfs
Lennart Poettering [Mon, 25 Mar 2013 17:18:09 +0000 (18:18 +0100)]
fstab-generator: rename x-initrd-rootfs.mount to x-initrd.rootfs

This changes the fstab mount option x-initrd-rootfs.mount to
x-initrd.rootfs, in order to only use a single namespace "x-initrd." for
all mount options of the initrd.

11 years agoupdate TODO
Lennart Poettering [Mon, 25 Mar 2013 17:17:15 +0000 (18:17 +0100)]
update TODO

11 years agojournal: Add sync timer to journal server
Oleksii Shevchuk [Mon, 25 Mar 2013 16:49:03 +0000 (18:49 +0200)]
journal: Add sync timer to journal server

Add option to force journal sync with fsync. Default timeout is 5min.
Interval configured via SyncIntervalSec option at journal.conf. Synced
journal files will be marked as OFFLINE.

Manual sync can be performed via sending SIGUSR1.

11 years agoupdate TODO
Lennart Poettering [Mon, 25 Mar 2013 16:25:48 +0000 (17:25 +0100)]
update TODO

11 years agocore: ensure LSB Provides are handled correctly
Frederic Crozat [Thu, 21 Mar 2013 14:40:45 +0000 (15:40 +0100)]
core: ensure LSB Provides are handled correctly

Let's say you have two initscripts, A and B:

A contains in its LSB header:
Required-Start: C

and B contains in its LSB header:
Provides: C

When systemd is parsing /etc/rc.d/, depending on the file order, you
can end up with either:
- B is parsed first. An unit "C.service" will be "created" and will be
added as additional name to B.service, with unit_add_name. No bug.
- A is parsed first. An unit "C.service" is created for the
"Required-Start" dependency (it will have no file attached, since
nothing provides this dependency yet). Then B is parsed and when trying
to handle "Provides: C", unit_add_name is called but will fail, because
"C.service" already exists in manager->units. Therefore, a merge should
occur for that case.

11 years agoudevd.c: set udev children_max according to CPU count
Harald Hoyer [Mon, 25 Mar 2013 12:02:05 +0000 (13:02 +0100)]
udevd.c: set udev children_max according to CPU count

Setting children_max according to RAM leads to too much concurrent I/O.

11 years agotimer: downgrade time change message to debug
Michal Schmidt [Mon, 25 Mar 2013 11:31:44 +0000 (12:31 +0100)]
timer: downgrade time change message to debug

The manager already prints "Time has been changed" at level info. It
seems too verbose to print the time change message additionally for
every waiting timer unit.

Downgrade the per-unit message to debug.

11 years agobus: split socket related code into bus-socket.[ch], to prepare for kdbus backend
Lennart Poettering [Mon, 25 Mar 2013 02:04:03 +0000 (03:04 +0100)]
bus: split socket related code into bus-socket.[ch], to prepare for kdbus backend

11 years agobus: implement 'unixexec:' protocol
Lennart Poettering [Mon, 25 Mar 2013 01:30:32 +0000 (02:30 +0100)]
bus: implement 'unixexec:' protocol

11 years agobus: make optional whether unix socket passing is negotiated and whether hello is...
Lennart Poettering [Sun, 24 Mar 2013 22:55:03 +0000 (23:55 +0100)]
bus: make optional whether unix socket passing is negotiated and whether hello is sent

This alos gets rid of explicit sd_open_fd() and sd_open_address()
constructors in favour of sd_new() + sd_new_start() where the
negotiation parameters may be set it in between.

11 years agobus: properly handle termination of connections
Lennart Poettering [Sun, 24 Mar 2013 22:54:44 +0000 (23:54 +0100)]
bus: properly handle termination of connections

11 years agobus: implement support for FD passing
Lennart Poettering [Sun, 24 Mar 2013 21:02:05 +0000 (22:02 +0100)]
bus: implement support for FD passing

11 years agoservice: no need to drop rc. prefix anymore
Miklos Vajna [Sun, 24 Mar 2013 09:06:38 +0000 (10:06 +0100)]
service: no need to drop rc. prefix anymore

This reverts commit f5c88ec1330b61787441156de7d764a140774bd2. It is no
longer necessary, and adds unnecessary magic.

11 years agojournald: remove unconditional log_*() use in the main loop
Kay Sievers [Sat, 23 Mar 2013 20:02:59 +0000 (21:02 +0100)]
journald: remove unconditional log_*() use in the main loop

Setting MaxRetentionSec= caused the kernel log to overflow and the
journal daemon to enter an endless loop.

Logging from the journald main loop gets directed to /dev/kmsg,
which wakes up journald again. We skip the import of this message
by checking for our own PID, but this still causes the main loop
to never go to sleep again because we never stopped logging from
there.

11 years agoTODO: add journal MaxRetentionSec= bug
Kay Sievers [Sat, 23 Mar 2013 19:36:05 +0000 (20:36 +0100)]
TODO: add journal MaxRetentionSec= bug

11 years agohwdb: ID_PRODUCT_FROM_DATABASE --> ID_MODEL_FROM_DATABASE
Kay Sievers [Sat, 23 Mar 2013 15:27:28 +0000 (16:27 +0100)]
hwdb: ID_PRODUCT_FROM_DATABASE --> ID_MODEL_FROM_DATABASE

With the conversion from pci-db + usb-db to hwdb, the property
got accitentially renamed.

Move the name back to the long established identifier *MODEL*
instead of *PRODUCT*.

  $ git grep -l ID_MODEL_FROM_DATABASE
  hwdb/20-pci-vendor-model.hwdb
  hwdb/20-usb-vendor-model.hwdb
  hwdb/ids-update.pl
  man/systemd.device.xml
  rules/78-sound-card.rules
  src/core/device.c
  src/cryptsetup/cryptsetup.c

11 years agouaccess: remove needless usb_id call from udev rules
Kay Sievers [Sat, 23 Mar 2013 15:26:49 +0000 (16:26 +0100)]
uaccess: remove needless usb_id call from udev rules

11 years agoudevd: ensure static nodes are created before local-fs mount
Frederic Crozat [Thu, 21 Mar 2013 16:28:13 +0000 (17:28 +0100)]
udevd: ensure static nodes are created before local-fs mount

static nodes (like /dev/loop-control) are created when systemd-udevd
is started and needed to mount loopback devices. Therefore,
local-fs-pre.target should be only started after systemd-udevd is
started.

11 years agoTODO: add "dead" device node ACL handling
Kay Sievers [Sat, 23 Mar 2013 13:01:44 +0000 (14:01 +0100)]
TODO: add "dead" device node ACL handling

11 years agoTODO: update
Kay Sievers [Sat, 23 Mar 2013 12:34:13 +0000 (13:34 +0100)]
TODO: update

11 years agoUpdate user session unit template.
Auke Kok [Sat, 23 Mar 2013 04:43:31 +0000 (21:43 -0700)]
Update user session unit template.

While most folks will be using the derivative from user-session-units,
I'm updating this one to reflect some of the fixes and things to note
about user sessions:

- cgroup should be set with "%u" - username instead of %I
- set dbus path with %U explicitly too
- hint to folks that wish to use MEM_CG features in user sessions
- allow unit to be enabled for instances with systemctl enable

11 years agoconf-parser: when we parse a string list, always fill in something
Lennart Poettering [Sat, 23 Mar 2013 03:32:43 +0000 (04:32 +0100)]
conf-parser: when we parse a string list, always fill in something

Some code really wants to know whether there was a string list parsed,
so don't take the shortcut here, and always allocate a string list, even
if it is an empty one.

https://bugs.freedesktop.org/show_bug.cgi?id=62558

11 years agojournal: don't access j->files after use
Lennart Poettering [Sat, 23 Mar 2013 03:11:23 +0000 (04:11 +0100)]
journal: don't access j->files after use

https://bugs.freedesktop.org/show_bug.cgi?id=62605

11 years agoudevadm: place const in the right place
Zbigniew Jędrzejewski-Szmek [Sat, 23 Mar 2013 02:55:05 +0000 (22:55 -0400)]
udevadm: place const in the right place

Otherwise clang at least thinks that both consts apply to char.

11 years agobus: don't free an unallocated var
Zbigniew Jędrzejewski-Szmek [Sat, 23 Mar 2013 02:00:49 +0000 (22:00 -0400)]
bus: don't free an unallocated var

Also remove unused variable.

11 years agobus: move attribute to end of structure, so it is not ignored
Zbigniew Jędrzejewski-Szmek [Sat, 23 Mar 2013 01:58:43 +0000 (21:58 -0400)]
bus: move attribute to end of structure, so it is not ignored

src/libsystemd-bus/bus-message.h:41:1: warning: attribute 'packed' is ignored, place it after
      "struct" to apply attribute to type declaration [-Wignored-attributes]

11 years agoupdate TODO
Lennart Poettering [Sat, 23 Mar 2013 03:03:32 +0000 (04:03 +0100)]
update TODO

11 years agoupdate TODO
Lennart Poettering [Sat, 23 Mar 2013 02:19:50 +0000 (03:19 +0100)]
update TODO

11 years agounit: never retroactively start requisites
Lennart Poettering [Sat, 23 Mar 2013 02:16:17 +0000 (03:16 +0100)]
unit: never retroactively start requisites

Requesites are not supposed to be auto-started afterall, they are just
checks, so don't try to be smarter here than appropriate.

Based on a patch from Michal Schmidt.

11 years agoRevert "units: ignore systemd-sysctl on shutdown"
Lennart Poettering [Sat, 23 Mar 2013 02:10:41 +0000 (03:10 +0100)]
Revert "units: ignore systemd-sysctl on shutdown"

This reverts commit faeffa73a81ab5b59acfadeb571431fb0e42af70.

There isn't really much point in dropping the Conflicts= since shutting
down this service is basically free as it doesn't have anything running.

Also, the patch was incomplete, because shutdown.target was still listed
in Before=.

11 years agoadd --with-telinit=PATH configure option
Cristian Rodríguez [Sat, 23 Mar 2013 00:04:53 +0000 (21:04 -0300)]
add --with-telinit=PATH configure option

Distributions that never shipped upstart do not have
"telinit" in /lib/upstart/..

Defaults to /lib/upstart/telinit so there is no change
for systems existing installs.

11 years agojournalctl: various fixes to the access check logic
Lennart Poettering [Sat, 23 Mar 2013 00:12:22 +0000 (01:12 +0100)]
journalctl: various fixes to the access check logic

- Reword messages a bit

- Correct check whether EACCES is in the set of errors

- Don't complain if no journal files are found

- allocate Set object for errors lazily since in the best case we don't
  need it at all.

- don't consider it an error if /run/log/journal doesn't exist (because
  that's the usual case actually, if storage is enabled)

11 years agodbus: Do send out "replies" to signals
Colin Walters [Fri, 22 Mar 2013 13:36:22 +0000 (09:36 -0400)]
dbus: Do send out "replies" to signals

Some parts of systemd (at least the DBus activation codepath) "reply"
to signals, which of course have the no-reply flag set.  We will be
defensive here and still send out a reply if we're passed a signal.

Regression introduced by: c6a818c82035da91e

Reported-by: Mantas Mikulėnas <grawity@gmail.com>
Tested-by: Mantas Mikulėnas <grawity@gmail.com>
11 years agoupdate TODO
Lennart Poettering [Fri, 22 Mar 2013 22:28:36 +0000 (23:28 +0100)]
update TODO

11 years agoexec: Assigning the empty string to CapabilityBoundSet= should drop all caps
Lennart Poettering [Fri, 22 Mar 2013 22:25:54 +0000 (23:25 +0100)]
exec: Assigning the empty string to CapabilityBoundSet= should drop all caps

Previously, it would set all caps, but it should drop them all, anything
else makes little sense.

Also, document that this works as it does, and what to do in order to
assign all caps to the bounding set.

https://bugzilla.redhat.com/show_bug.cgi?id=914705

11 years agoDrop trailing whitespace
Tollef Fog Heen [Fri, 22 Mar 2013 21:51:52 +0000 (22:51 +0100)]
Drop trailing whitespace

11 years agotimedated: extra overflow safety check when doing relative time changes
Lennart Poettering [Fri, 22 Mar 2013 20:35:53 +0000 (21:35 +0100)]
timedated: extra overflow safety check when doing relative time changes

Ensure clients don't overflow usec_t when doing relative time changes.
This is mostly just paranoia and protection against accidents, after all
clients are already authenticated, and they can se the time to any
value they wish anyway, but better be safe than sorry.

https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1152187/comments/14

11 years agoupdate TODO
Lennart Poettering [Fri, 22 Mar 2013 20:17:53 +0000 (21:17 +0100)]
update TODO

11 years agotimedatectl: show CanNTP field
Lennart Poettering [Fri, 22 Mar 2013 20:17:45 +0000 (21:17 +0100)]
timedatectl: show CanNTP field

11 years agoudev: no need to output OOM, if we call log_oom() anyway
Lennart Poettering [Fri, 22 Mar 2013 20:05:42 +0000 (21:05 +0100)]
udev: no need to output OOM, if we call log_oom() anyway

11 years agomain: minor simplification
Lennart Poettering [Fri, 22 Mar 2013 14:05:51 +0000 (15:05 +0100)]
main: minor simplification

11 years agoudev/collect: avoid initalizing memory twice
Zbigniew Jędrzejewski-Szmek [Fri, 22 Mar 2013 00:05:42 +0000 (20:05 -0400)]
udev/collect: avoid initalizing memory twice

11 years agoutil: workaround two gcc warnings
Zbigniew Jędrzejewski-Szmek [Fri, 22 Mar 2013 03:24:30 +0000 (23:24 -0400)]
util: workaround two gcc warnings

gcc does not know that errno cannot be negative, and warns
about unitialized variables later on. Kill the warnings by
returning -errno only after checking that errno is positive.

11 years agojournalctl: be smarter about journal error checks
Zbigniew Jędrzejewski-Szmek [Mon, 11 Mar 2013 22:03:13 +0000 (18:03 -0400)]
journalctl: be smarter about journal error checks

There are many ways in which we can get those checks wrong, so it is
better to warn and then error out on a real access failure.

The error messages are wrapped to <80 lines, because their primary
use is to be displayed in the terminal, and it is easier to read them
this way. Reading them in the journal can be a bit trickier, but
this is a bug in logs-show.c.

11 years agobuild-sys: move acl searching code into libsystemd-acl
Zbigniew Jędrzejewski-Szmek [Fri, 22 Mar 2013 14:35:26 +0000 (14:35 +0000)]
build-sys: move acl searching code into libsystemd-acl

This loop over acls is a bit too much to keep inside
of another loop.

11 years agohostnamed: pretty_string_is_safe() already exists in string_has_cc(), so use that
Lennart Poettering [Fri, 22 Mar 2013 17:01:26 +0000 (18:01 +0100)]
hostnamed: pretty_string_is_safe() already exists in string_has_cc(), so use that