On Fri, 1 Jul 2016, Benjamin Drung wrote: > > So why don't you just declare lsb-release required for that > > particular feature to work? > > It's not one particular feature. The salt minion has grains (variables) > and oscodename is one grain that every minion has. So this value is > wrong without lsb-release on all minion. Thus it's not just a broken > feature, but broken for all. > > I like to keep the list of dependencies small. So installing lsb- > release is not my first choice. > > > If you consider this a bug at all, your intended fix (rely on the > > /etc/os-release file) would only half-fix it, because it would still > > be wrong in testing and unstable. > > I know, but we could use other fallbacks there.
Ok, but why reinvent the wheel? The maintainers of lsb-release already took care of this, and lsb-release has an installed size of 62kB. I can understand your desire to keep the dependency list "small", but you should probably take in account the installed size as well. BTW: I use salt myself, on systems running stable and testing, and the package lsb-release is already installed as a dependency from unattended-upgrades. So at least in my case, I would not notice the difference at all. The problem I see in your rationale (lsb-release is optional, while base-files is essential) is precisely that it would make the codename thing essential de-facto even if it's defined as optional in the specs. When one needs a feature, the normal thing is to add a Depends on a package providing the feature, not to ask for the feature to be in an essential package. Thanks.
