- specfile: conflict systemd-bash-completion and systemd-mini-bash-completion
Otherwise the build system detects that systemd-bash-completion and
its mini variant are conflicting at files level even though those
packages can't be installed on the same system.
- specfile: clean up nss-* plugins descriptions and drop
nss-myhostname-config script for now.
Currently /etc/nsswitch.conf is supposed to be edited by the
sysadmin to enable the modules. However for some reasons only
nss-myhostname is removed from the conf file when the corresponding
package is uninstalled. This is inconsistent so let's remove it.
Actually I'm wondering if we shouldn't make those NSS plugins part
of the main package and get rid of all those sub-packages...
- specfile: remove old comments and unneeded sed command
- specfile: no need to create systemd-update-utmp-runlevel.service symlinks anymore
The symlinks in /usr/lib/systemd/system/<target>.target.wants/systemd-update-utmp-runlevel.service
are created in Makefile.am since commit d5d8429a12c4b1.
'reboot' and 'poweroff' targets initially had the symlinks but
there's not point since the latter conflicts shutdown.target whereas
the 2 targets pull it in.
See: https://github.com/systemd/systemd/pull/4429
- specfile: remove the following warnings:
[ 256s] warning: File listed twice: /usr/lib/systemd/system/dbus.target.wants
[ 256s] warning: File listed twice: /usr/lib/systemd/system/default.target.wants (forwarded request 440233 from fbui)
OBS-URL: https://build.opensuse.org/request/show/440243
OBS-URL: https://build.opensuse.org/package/show/openSUSE:Factory/systemd?expand=0&rev=242
- specfile: conflict systemd-bash-completion and systemd-mini-bash-completion
Otherwise the build system detects that systemd-bash-completion and
its mini variant are conflicting at files level even though those
packages can't be installed on the same system.
- specfile: clean up nss-* plugins descriptions and drop
nss-myhostname-config script for now.
Currently /etc/nsswitch.conf is supposed to be edited by the
sysadmin to enable the modules. However for some reasons only
nss-myhostname is removed from the conf file when the corresponding
package is uninstalled. This is inconsistent so let's remove it.
Actually I'm wondering if we shouldn't make those NSS plugins part
of the main package and get rid of all those sub-packages...
- specfile: remove old comments and unneeded sed command
- specfile: no need to create systemd-update-utmp-runlevel.service symlinks anymore
The symlinks in /usr/lib/systemd/system/<target>.target.wants/systemd-update-utmp-runlevel.service
are created in Makefile.am since commit d5d8429a12c4b1.
'reboot' and 'poweroff' targets initially had the symlinks but
there's not point since the latter conflicts shutdown.target whereas
the 2 targets pull it in.
See: https://github.com/systemd/systemd/pull/4429
- specfile: remove the following warnings:
[ 256s] warning: File listed twice: /usr/lib/systemd/system/dbus.target.wants
[ 256s] warning: File listed twice: /usr/lib/systemd/system/default.target.wants (forwarded request 440233 from fbui)
OBS-URL: https://build.opensuse.org/request/show/440243
OBS-URL: https://build.opensuse.org/package/show/openSUSE:Factory/systemd?expand=0&rev=242
- specfile: conflict systemd-bash-completion and systemd-mini-bash-completion
Otherwise the build system detects that systemd-bash-completion and
its mini variant are conflicting at files level even though those
packages can't be installed on the same system.
- specfile: clean up nss-* plugins descriptions and drop
nss-myhostname-config script for now.
Currently /etc/nsswitch.conf is supposed to be edited by the
sysadmin to enable the modules. However for some reasons only
nss-myhostname is removed from the conf file when the corresponding
package is uninstalled. This is inconsistent so let's remove it.
Actually I'm wondering if we shouldn't make those NSS plugins part
of the main package and get rid of all those sub-packages...
- specfile: remove old comments and unneeded sed command
- specfile: no need to create systemd-update-utmp-runlevel.service symlinks anymore
The symlinks in /usr/lib/systemd/system/<target>.target.wants/systemd-update-utmp-runlevel.service
are created in Makefile.am since commit d5d8429a12c4b1.
'reboot' and 'poweroff' targets initially had the symlinks but
there's not point since the latter conflicts shutdown.target whereas
the 2 targets pull it in.
See: https://github.com/systemd/systemd/pull/4429
- specfile: remove the following warnings:
[ 256s] warning: File listed twice: /usr/lib/systemd/system/dbus.target.wants
[ 256s] warning: File listed twice: /usr/lib/systemd/system/default.target.wants
OBS-URL: https://build.opensuse.org/request/show/440233
OBS-URL: https://build.opensuse.org/package/show/Base:System/systemd?expand=0&rev=961
- Own a couple of directories even if they don't belong to systemd
otherwise the build system will complain.
Some directories (owned by others packages) are simply used by
systemd to ship some scripts or config files to customize others
*optional* components. Since thos components are not build required
by systemd those directories are not owned by any packages and the
BS complains...
- Import commit
15ea716 journal-remote: change owner of /var/log/journal/remote and create /var/lib/systemd/journal-upload (bsc#1006372)
- %sysusers_create and %tmpfiles_create must be called in %post
Calling %pre is broken since the respective conf files are not yet
installed.
- %{_libexecdir}/{tmpfiles.d,sysusers.d}/systemd-remote.conf are
part of systemd-journal-remote package (only).
- systemd-journal-{gatewayd,remote,upload} units are only part of
"systemd-journal-remote" package.
So exclude them from the main package.
- Import commit a1c145e6ad6588555dca64402f9103fb1e02b1a0
7f34037 man: explain that *KeyIgnoreInhibited only apply to a subset of locks
df5798b Revert "logind: really handle *KeyIgnoreInhibited options in logind.conf" (bsc#1001790 bsc#1005404)
f79fee7 Revert "kbd-model-map: add more mappings offered by Yast"
3760c10 manager: tighten incoming notification message checks
d6efd71 core: only warn on short reads on signal fd
6eebd91 manager: be stricter with incomining notifications, warn properly about too large ones
OBS-URL: https://build.opensuse.org/request/show/437403
OBS-URL: https://build.opensuse.org/package/show/Base:System/systemd?expand=0&rev=959
- Import a better fix from upstream for bsc#1001765
- Added:
0001-pid1-more-informative-error-message-for-ignored-noti.patch
0001-pid1-process-zero-length-notification-messages-again.patch
- Updated (no code changes, only patch metadata)
0001-If-the-notification-message-length-is-0-ignore-the-m.patch
0001-pid1-don-t-return-any-error-in-manager_dispatch_noti.patch
- Re add back "udev: don't require nsserv and fillup"
Did this in the wrong project... it was a complicated day today ;)
- Added 2 patches to fix bsc#1001765
0001-If-the-notification-message-length-is-0-ignore-the-m.patch
0001-pid1-don-t-return-any-error-in-manager_dispatch_noti.patch
- Revert "udev: don't require nsserv and fillup"
It's been judged too late for being part of SLE12 final release.
Nevertheless it's part of Factory and will be reintroduced after the
final release is out (ie through an update).
- systemd-sysv-convert: make sure that
/var/lib/systemd/sysv-convert/database is always initialized (bsc#982211)
If "--save" command was used and the sysv init script wasn't enabled
at all the database file wasn't created at all. This makes the
subsequent call to "--apply" fail even though this should not
considered as an error.
- Added patches to fix journal with FSS protection enabled (bsc#1000435)
0001-journal-fix-HMAC-calculation-when-appending-a-data-o.patch
0001-journal-set-STATE_ARCHIVED-as-part-of-offlining-2740.patch
OBS-URL: https://build.opensuse.org/request/show/431464
OBS-URL: https://build.opensuse.org/package/show/openSUSE:Factory/systemd?expand=0&rev=241
- Import a better fix from upstream for bsc#1001765
- Added:
0001-pid1-more-informative-error-message-for-ignored-noti.patch
0001-pid1-process-zero-length-notification-messages-again.patch
- Updated (no code changes, only patch metadata)
0001-If-the-notification-message-length-is-0-ignore-the-m.patch
0001-pid1-don-t-return-any-error-in-manager_dispatch_noti.patch
- Re add back "udev: don't require nsserv and fillup"
Did this in the wrong project... it was a complicated day today ;)
- Added 2 patches to fix bsc#1001765
0001-If-the-notification-message-length-is-0-ignore-the-m.patch
0001-pid1-don-t-return-any-error-in-manager_dispatch_noti.patch
- Revert "udev: don't require nsserv and fillup"
It's been judged too late for being part of SLE12 final release.
Nevertheless it's part of Factory and will be reintroduced after the
final release is out (ie through an update).
- systemd-sysv-convert: make sure that
/var/lib/systemd/sysv-convert/database is always initialized (bsc#982211)
If "--save" command was used and the sysv init script wasn't enabled
at all the database file wasn't created at all. This makes the
subsequent call to "--apply" fail even though this should not
considered as an error.
- Added patches to fix journal with FSS protection enabled (bsc#1000435)
0001-journal-fix-HMAC-calculation-when-appending-a-data-o.patch
0001-journal-set-STATE_ARCHIVED-as-part-of-offlining-2740.patch
OBS-URL: https://build.opensuse.org/request/show/431464
OBS-URL: https://build.opensuse.org/package/show/openSUSE:Factory/systemd?expand=0&rev=241
is always initialized (bsc#982211)
If "--save" command was used and the sysv init script wasn't enabled
at all, the database file wasn't created at all. This makes the
subsequent call to "--apply" fail even though it's a valid scenario.
OBS-URL: https://build.opensuse.org/package/show/Base:System/systemd?expand=0&rev=954
Split systemctl and tmpfiles into a separate package
The backstory is that %{?systemd_requires} in all kinds of .spec files
pulls in systemd, which is not always desirable. Upstream has come up
with some solution where they make use of a rpm tag (available in
4.10+) that is only about ordering - cf
2424b6bd71
. The commit explicitly states though, that "installing systemd
afterwards [...], does not result in the same outcome."
With this proposed change to our systemd.spec, we could have the cake
and eat it too:
* symlinks in /etc/systemd can be created at install time, permitting
the administrator to install/use systemd as init system in a container
at a later date (enabling "same outcome")
* mandatory tmp directories will be created at install time already,
so the administrator does not have to run `tmpfiles --root=/mycontainer`
before launching the (systemd-less) container.
* the only protruding dependency of systemctl.rpm is libgcrypt20,
i.e. comparatively light.
* the OrderWithRequires feature can be done independently
Sounds like a plan? @mpluskal too #boo980389
OBS-URL: https://build.opensuse.org/request/show/415263
OBS-URL: https://build.opensuse.org/package/show/Base:System/systemd?expand=0&rev=942
- Fix patch patch handle-numlock-value-in-etc-sysconfig-keyboard.patch
that is do not close a file descriptor twice (boo#973907)
- Add two patches which address logind/networkd disappearing from
dbus (and busctl) even while the units and processes continue
running.
0001-core-fix-bus-name-synchronization-after-daemon-reloa.patch
0001-core-re-sync-bus-name-list-after-deserializing-durin.patch
- drop all compiler/linker option customizations:
- -pipe option is used by default since day 0
- get rid of cflags() function which is not needed
- --hash-size has no impact specially in runtime
IOW, use the default options for the compiler and the linker,
there's no point in making systemd different from other package in
this regards.
- use %make_build instead of 'make %{?_smp_mflags}'
- be more strict on own lib version requirements
- systemd should require udev with the exact same version
- Modify patch handle-numlock-value-in-etc-sysconfig-keyboard.patch
to allow that open, seek, and read of /dev/mem may fail e.g.
on XEN based virtual guests (bsc#961120)
- require curl and bzip2 to build importd
- curl also causes building of journal-upload. That one has rather
unusal certificate usage, set it's ca root to /etc/pki/systemd
OBS-URL: https://build.opensuse.org/request/show/384339
OBS-URL: https://build.opensuse.org/package/show/openSUSE:Factory/systemd?expand=0&rev=232
- Fix patch patch handle-numlock-value-in-etc-sysconfig-keyboard.patch
that is do not close a file descriptor twice (boo#973907)
- Add two patches which address logind/networkd disappearing from
dbus (and busctl) even while the units and processes continue
running.
0001-core-fix-bus-name-synchronization-after-daemon-reloa.patch
0001-core-re-sync-bus-name-list-after-deserializing-durin.patch
- drop all compiler/linker option customizations:
- -pipe option is used by default since day 0
- get rid of cflags() function which is not needed
- --hash-size has no impact specially in runtime
IOW, use the default options for the compiler and the linker,
there's no point in making systemd different from other package in
this regards.
- use %make_build instead of 'make %{?_smp_mflags}'
- be more strict on own lib version requirements
- systemd should require udev with the exact same version
- Modify patch handle-numlock-value-in-etc-sysconfig-keyboard.patch
to allow that open, seek, and read of /dev/mem may fail e.g.
on XEN based virtual guests (bsc#961120)
- require curl and bzip2 to build importd
- curl also causes building of journal-upload. That one has rather
unusal certificate usage, set it's ca root to /etc/pki/systemd
OBS-URL: https://build.opensuse.org/request/show/384339
OBS-URL: https://build.opensuse.org/package/show/openSUSE:Factory/systemd?expand=0&rev=232
