[FFe] [MIR] libnss-myhostname

From a packaging, maintainability POV, this is fine. The package doesn't follow Debian policy for library package names. It should be libnss-hostname2. But I suppose that since this is not a library that other programs should link against, and since the first time it bumps SONAMEs, it could also bump the package name then, this isn't a huge problem.

It would be nice to see a bug subscriber.

I'll punt to the security team for a quick security check.

I reviewed libnss-myhostname version 0.3-4 from the Raring archives.

This shouldn't be considered a full security audit, rather an indicator of
maintainability.

- libnss-myhostname adds a new nss module that can be added to nsswitch to
  always provide local hostname<->IP lookups regardless the contents of
  /etc/hosts
- No encryption
- No off-machine networking; will change results of nss-based lookups
- Relies upon netlink to retrieve address information for interfaces
- Does not run as a daemon; it is linked into every process that performs
  nss lookups
- No PIE (library, fine), no Fortify (no functions used, fine), no
  immediate biding (as dynamicly loaded library with few deps, fine)
- Stack protection, read-only relocations
- No initscripts, no dbus services, no setuid, no sudo fragments
- No spawned processes
- Careful memory management
- Careful data handling
- Warnings in build from autotools
- Warnings in build from failed inlining

It'd be nice to have the warnings fixed at some point, but this does not
block main inclusion.

ACK

Override component to main
libnss-myhostname 0.3-4 in raring: universe/admin -> main
libnss-myhostname 0.3-4 in raring amd64: universe/admin/extra -> main
libnss-myhostname 0.3-4 in raring armhf: universe/admin/extra -> main
libnss-myhostname 0.3-4 in raring i386: universe/admin/extra -> main
libnss-myhostname 0.3-4 in raring powerpc: universe/admin/extra -> main
5 publications overridden.

please seed it or add the dependency/recommendation.

Moving status back to New, upload happened before release team approval.

Adding a default NSS module post feature freeze seems rather dangerous to me, especially one that potentially messes with the IP on which services will listen.

I'm not necessarily opposed to libnss-myhostname per say, but I'm opposed to it getting added by default at this point of the cycle.

This is a release team NACK, please revert your change and close the bug in the process.

Thanks

> The package doesn't follow Debian policy for library package names. It should be libnss-hostname2.

No, it's an NSS module not a shared library - the soname shouldn't be part of the package name.

But as far as an FFe is concerned, I don't agree that libnss-myhostname is a technically sound approach, *period*; it works at cross-purposes to the existing Debian handling of the hostname being resolved to 127.0.1.1 via /etc/hosts. There should be a plan for aligning these, across *all* systems, not just desktop systems.

For my money, I believe the right answer is to ensure tools which update the system hostname also update /etc/hosts (via some common tool), and *not* add to the NSS overhead with this module.

Is the Release Team ok with me having ubuntu-gnome-desktop depend on libnss-myhostname in order for hostnamectl to work?

On Thu, Apr 11, 2013 at 06:46:22PM -0000, Jeremy Bicha wrote:
> Is the Release Team ok with me having ubuntu-gnome-desktop depend on
> libnss-myhostname in order for hostnamectl to work?

I reckon that as a flavor lead, it's your call which of the two imperfect
options you want (unusable hostnamectl tool vs. inconsistency across
flavors). In any case, I hope that in S you'll help us work toward a
solution that provides consistency across all Ubuntu installs.

Thanks, I don't really see how I can help for S though: I'm not a systemd developer nor do I have anywhere near the programming skills to code anything better.

I was just hoping that users would finally be able to benefit from a feature introduced in GNOME 3.2 and shipped by the other GNOME distros. I'd be happy to re-land this when S opens so that there's plenty of time to identify any actual issues but I guess we need the Foundations Team to approve that?

Status changed to 'Confirmed' because the bug affects multiple users.

Unsubscribing release team; I don't think there's anything for us to do here at this point.

Status changed to 'Confirmed' because the bug affects multiple users.

There is a conflict, libnss-myhostname has been moved into systemd source tree and this package should now be provided by src:systemd instead of separate source package.

And systemd is in main already I believe, thus no MIR required.