From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Petazzoni Date: Fri, 20 Mar 2015 14:20:55 +0100 Subject: [Buildroot] RFC: systemd and service files cleanup In-Reply-To: <1426787807-29510-1-git-send-email-mike@mikebwilliams.com> References: <1426787807-29510-1-git-send-email-mike@mikebwilliams.com> Message-ID: <20150320142055.5ba1af4c@free-electrons.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Dear Mike Williams, Adding Steven in Cc. On Thu, 19 Mar 2015 13:56:32 -0400, Mike Williams wrote: > Currently, package service files for systemd are installed by > buildroot in a variety of locations: /etc, /lib, and /usr/lib. For > systemd, split /lib is deprecated and only checked if a split /lib > and /usr/lib is enabled. /etc is meant for local customizations to > the default service files. It is my opinion that buildroot should > install all upstream package and buildroot-provided service files > in /usr/lib, and this series cleans up our packages to do exactly > that. Enabling them by default by linking them > to /etc/systemd/system/multi-user.target.wants has been preserved. > > Upstream systemd's standard installation directory for its binaries > is /usr/lib. It appears that historically, buildroot's systemd > installation location has been switched between /lib and /usr/lib > multiple times. I couldn't find any particular reason for this, so > I've moved it back to /usr/lib along with its service files. I am generally fine with the proposed changes, especially since other systemd users looked at them. However, there's one thing I'm not sure about: you're changing all the relative symbolic links to absolute symbolic links: - ln -fs ../ntpd.service $(TARGET_DIR)/etc/systemd/system/multi-user.target.wants/ntpd.service + ln -fs /usr/lib/systemd/system/ntpd.service $(TARGET_DIR)/etc/systemd/system/multi-user.target.wants/ntpd.service While it causes no problem on the target, it means that when looking at output/target/ on your build machine, now all those symbolic links are broken. I find it quite nice when relative symlinks are used, since it allows to have things in output/target looking somewhat sensible. Is there any strong reason for switching to absolute symbolic links? Also, there are *lots* of systemd related patches sitting in patchwork. Could you and Steven have a look at those pending patches, and let me know which ones can be applied, which ones cannot? Thanks, Thomas -- Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com