Re: stupid dependencies on update-inetd
md@Linux.IT (Marco d'Itri) writes:
> I don't know exactly how it happened, but a large number of maintainers
> apparently ignored the discussions on this list and added to their
> packages a dependency on update-inetd.
> This is *TOTALLY WRONG* because the /usr/sbin/update-inetd interface is
> guaranteed to be provided by whatever implements the inet-superserver
> virtual package and not by the update-inetd package currently depended
> on by some inetd packages.
> Indeed, the update-inetd package does not depend on a daemon package nor
> it provides one itself.
So what are they *supposed* to depend on, only inet-superserver? I'm
failing to extract a clear guideline from half-remembered debian-devel
discussions (as is clear from the fact that I apparently got the lintian
check wrong). Is this documented somewhere that anyone could expect to
find it?
Currently, lintian allows any combination of dependencies on the following
packages to satisfy the dependency requirement from calling update-inetd
in maintainer scripts:
update-inetd inet-superserver openbsd-inetd rlinetd
I gather update-inetd should be removed from that list. Is it otherwise
correct?
> Some of these packages instead are only slightly less broken and depend
> both on inet-superserver and update-inetd, making impossible to install
> a future xinetd package providing its own /usr/sbin/update-inetd.
Should no package ever depend on update-inetd unless it also provides
inet-superserver? If so, I can add a lintian check for that. If not, is
there something else that lintian can check for?
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: