On Mon, 2020-04-27 at 15:30 +0100, Daniel P. Berrangé wrote:
On Mon, Apr 27, 2020 at 04:16:57PM +0200, Andrea Bolognani wrote:
> On Mon, 2020-04-27 at 15:38 +0200, Pavel Hrdina wrote:
> > I've checked systemd and they are installing these files using meson but
> > GLib for example is not installing them.
> >
> > After checking other libvirt projects we are not installing this files
> > using autotools or other build systems.
>
> I can bring up another counter-example that's close to us: QEMU
> installs these files, along with a bunch of other documentation.
QEMU isn't a good example of best practice as it is a completely
custom written set of build system rules.
The fact that it accomplishes something using a custom set of tools
doesn't automatically make the result not a good idea.
> The packaging for both Fedora and Debian go out of their way to
> include them, and at least for the latter it's strongly recommended
> by policy to include NEWS while including COPYING is mandatory.
> Clearly both projects consider this information important to users,
> and I don't see any disadvantage in installing them even for people
> who build from sources directly instead of using a package, only
> advantages really.
I think these are different situations though. Distro vendors include
the files because the end users don't have any direct visibility of
the source tarball. So the only way users would see the files is if
the binary packages included them. I don't think that extends to
building from source in general & don't see a need for us to do this
specially in the libvirt-dbus module, or indeed our other modules
in general.
The fact that something is not strictly needed doesn't automatically
make it not worth pursuing.
Anyway, I don't care enough about this to keep the discussion going
any further, so I'm just going to go ahead and revert the change.
Can we have a libvirt-dbus release now? :)
--
Andrea Bolognani / Red Hat / Virtualization