[arch-dev-public] [signoff] syslog-ng 3.2.4-2

[Tom Gundersen]

On Tue, Jun 28, 2011 at 6:08 PM, Andreas Radke <a.radke at arcor.de> wrote:
> Am Tue, 28 Jun 2011 10:33:19 -0400
> schrieb Dave Reisner <d at falconindy.com>:
>
>> syslog-ng 3.2.4-2 is in [testing]. It's a rebuild with the following
>> changes:
>>
>> * add --enable-systemd flag to ./configure
>> * add upstream patch to fix socket acitvation for systemd
>> * package upstream syslog-ng.service
>>
>> There should be zero impact for users still on sysvinit.
>>
>> Signoff both,
>> Dave
>>
>
> Maybe I missed some discussion but Arch way usually means to not add
> more additional features than what is required to support our repo
> packages. Since it's a core repo pkg this policy should be respected.
>
> Any reason why you are adding systemd support that we don't officially
> support so far?
>
> I see we have systemd in community now but wouldn't be the proper way to
> offer additional foo-systemd packages in community replacing the
> packages from core/extra where needed?

Just some additional info:

Enabling systemd support in packages, does not introduce a makedepend
on systemd, nor does it change the behavior of the package in question
if systemd is not installed. It is (in most cases) simply a matter of
installing an additional ini-style file in /lib/systemd/system.

(This package upgrade also includes a backport of a bugfix, so it is
slightly more complicated.)

Enabling systemd support makes it a lot easier for people to test and
evaluate systemd, so if no one is adversely affected I would be in
favor of allowing it (especially in the low-level packages like
syslog, udev, dbus, networkmanager, consolekit). Otherwise, we'll have
a bit a of a chicken and egg problem: How are we to decide whether or
not to add systemd to our repos if it is not widely tested? (Dave's
systmed-arch-units package from [community] is nice, but probably not
something we would ever move to an official repo, so not good for
evalutaing systemd).



Looking ahead a bit: if things pan out they way it looks like they
will (most distros adopting systemd, and gnome depending on it), then
we will probably at some point want to support both sysvinit and
systemd, giving people a choice (maybe one in [core] and one in
[extra]). As this seems likely, I think we should make things simple
for ourselves by encouraging testing of systemd (as long as it does
not cause problems for sysvinit of course).

Just my two cents,

-t