| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
some are on purpose, so leaving those in. Fixing properly is too much
effort
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
We don't really care about leaving left over cruft here as the
package is getting removed. Dangling symlinks and such like can
just be left there without getting in the way too much.
(and this should be rare anyway)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Due to the fact that systemd is reload by chkconfig, it will still
think the service exists. A hacky patch was included which rm'ed
the sysvinit script and then reloaded systemd again, which worked
but caused errors from rpm when removing packages complaining that
a file it controlled did not exist (which was correct).
It's actually much better to fix this with a filetrigger. We
actually already ship a filetrigger that performs appropriate
daemon-reloads, but it only matched on native units. Simply
changing the regexp to include sysvinit script paths too should
fix the problem equally well, but in a much less hacky way.
|
| |
|
|
|
|
|
| |
This is actually where the non-relative symlinks in chroots might occur
as technically all packges will only include relative symlinks.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When passing just a sysvinit name, we synthesise a systemd unit name via
the amazingly cunning and subtle technique of adding '.service' on to the
name.
This is very effective for referring to units for most commands as even if
an actual unit file of this name exists systemd might be providing the
unit for this service via a different name, but just symlinking to the
sysvinit name in order to prevent the sysvinit script from firing.
This is done with e.g. the dm and networkmanager services.
When using the symlinked name, systemctl will refuse to enable or disable
the unit (see https://bugs.mageia.org/show_bug.cgi?id=3188) and this is
expected behaviour.
Thus we use readlink to attempt to resolve the link to a real name.
|
|
|
|
|
|
|
|
|
|
|
| |
This ensures that services are started/stopped in the most appropriate
way depending on what init system is being used.
Also when upgrading a service, ensure it is properly restarted even
when using systemd (assuming it is currently running).
Testing these scripts without actually installing/uninstalling package
may result in the sysvinit scripts being deleted so be careful.
|
|
|
|
|
|
|
|
|
|
|
| |
The Mandriva code required that all packages were updated to deal with
systemd units whenever they are provided. While packages generally need
updated to include new units, it's also something that requires a lot of
work and thus in the short term we should deal with it with minimal
changes.
By always seting the units variable, some checks become redundant so
skip them or replace with a suitable alternative
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
with Magiea.
This commit consitsts of the following subversion commits:
------------------------------------------------------------------------
r2081 | dmorgan | 2011-10-27 23:29:24 +0100 (Thu, 27 Oct 2011) | 1 line
import rpm-helper
------------------------------------------------------------------------
|
|
|
|
|
|
| |
(or else shorewall will be stopped because moved in another
package, making the install hang for a while)
|
| |
|
| |
|
|
|