On Sun, Sep 27, 2015 at 2:42 PM, Adam D. Barratt
<a...@adam-barratt.org.uk> wrote:
>
> On Sun, 2015-09-27 at 14:21 -0400, Dave Steele wrote:
> > On Sun, Sep 27, 2015 at 12:24 PM, Adam D. Barratt
> > I don't have a strong reason for making the systemd change. I can
> > revert it, if it is deemed necessary. It means a larger patch against
> > 0.13-1. OTOH, I am getting a fair amount of grief outside of
> > Jessie/testing/sid for requiring systemd.
>
> What do you mean? (Also what's happening outside of Debian isn't really
> relevant in the majority of cases.)
Just that I can go with reverting to sysv-init, or not. Reverting is
more work, which means risk. Ignoring outside influences is fine.
> > > Why does the python version need to be explicitly declared?
> >
> > Jessie+ ${python:Depends} defines a dependency which includes a
> > multiarch reference. This breaks compatibility with outside
> > distributions that could otherwise run the package just fine (e.g.
> > Raspberry Pi/Raspbian is a popular target).
>
> Well, it includes a multiarch-dependency because that's how the Python
> packagers in Debian have arranged things. I have to admit to not being
> 100% sure of the consequences of dropping it, which makes me uneasy
> about doing it in stable.
With my change, the package is no longer, say, compatible with an
i386-compiled python on an AMD64 system. It requires a python using
the same architecture as the installation. I'm saying that I'm fine
with that distinction - I seriously doubt there's any real-world
scenario where this would cause a problem.
> > > Why does this need an explicit dependency on systemd?
> >
> > To make sure I have told e.g. Raspbian users that there is a
> > compatibility issue. Looks like I didn't need it.
>
> I'm not immediately convinced that adding dependencies in Debian to make
> sure that users outside of Debian are aware of something really makes
> sense.
OK
--
"Le mieux est l'ennemi du bien" - Voltaire
