systemd removal
I've gone through the patches and identified the key patches that apply systemd bindings on cups. Hopefully these can be reverted cleanly to give us a systemd free version of cups.
All these patches should be reviewed, and reversions should be in order from top to bottom (reverse of the order they are applied). There a couple noted that may be incidental (fixes an issue that was found to effect systemd particularly, but may also have effected non-systemd), and a couple that seemed to mix systemd patches with non-systemd related patches.
The commitish id's from latest to earliest are:
-- 4fdcc872 --
-- dbcc76b8 --
925c3341 - needs review
39313d52 - needs review
3733922f - possible unrelated
-
Is removing a patch which checks for a systemd configuration file really necessary? If systemd isn't installed, none of these bits of code will run. It is only creating more work and a maintenance burden.
I just tried removing the references to systemd from the debian/ files, and cups built fine. I don't have a printer at home, so the most I can do is access the web server.
Are there only problems with versions newer than jessie and cups/systemd? 'Cause jessie seems fine.
-
Jaret is right in asking, removing a patch that checks for a configuration file to add support to systemd at compile time, or even removing a patch that checks for it at runtime, is not necessary. We need to change the less as possible in these packages, to ease future maintainability. What we need to address is exclusive requirement of systemd and default build settings.
If a patch is removed because necessary to change the default build settings in /debian, then it makes sense. The upstream code shouldn't need any modification in such a case.
-
From what I've seen the way systemd is implemented and maintained, it may be easier to start with the cups upstream and the pre-systemd /debian and maintain our our own version. The problem with leaving these bits of code that don't run there is that it leaves room for later rebases to silently turn those bits of code into subtle bugs. I'm happy for others to take this on though and do it their way. I just pulled it in to have a look see and reported what I saw...
-
At a closer look, your scenario seems rather possible, as many changes are occurring.
Still I wouldn't go revert everything, but see if a simple change of configure flags and package deps produces a working cups for sysvinit systems.
-
done
-
Status changed to closed