Bug#747073: [cups-daemon] Doesn't work with systemd
On Fri, Oct 24, 2014 at 09:03:44PM +0200, Didier 'OdyX' Raboud wrote:
> Le vendredi, 24 octobre 2014, 10.13:00 Josh Triplett a écrit :
> > On Thu, 25 Sep 2014 22:24:49 +0200 Didier 'OdyX' Raboud
> <odyx@debian.org> wrote:
> > > ii) ListenStream with only the port number.
> > >
> > > This works for the localhost ipv6 [::1] but it doesn't listen
> > > on
> > > 127.0.0.1:631 but on :::631 . This implies that accessing
> > > http://localhost:631/ doesn't spawn CUPS (of course, if you try
> > > accessing http://[::1]:631/ first, then CUPS is spawned and the
> > > IPv4 access works duing the 30 seconds, as cups has taken the
> > > port.
> >
> > This seems like the fundamental bug here. Listening on [::1]:631 with
> > BindIPv6Only=both *should* work equivalently to listening on
> > 127.0.0.1:631 or localhost:631, and accessing http://localhost:631/
> > should then spawn CUPS rather than giving an error.
>
> Did you see Lennart's analysis on
> https://bugs.freedesktop.org/show_bug.cgi?id=84604 ?
Yes, I read that before responding. Between the two bugs, I think the
fundamental problem is that CUPS refuses to support v4mapped IPv6
sockets, which are the only way to handle everything through one socket,
and systemd can't currently work around that by configuring a socket per
*available* address family without failing on unavailable address
families.
> > That said, I'd still like to see systemd capable of producing both a
> > v4 and v6 listening socket, for whichever stacks the local system
> > has, omitting any that cannot exist because the stack doesn't exist.
> > (Ideally, systemd should only allow a ListenStream to fail because of
> > a missing stack, while still erroring on failures such as port
> > conflicts.)
>
> Yes, I concur.
>
> In fact, conducting this analysis and involving both the Apple and
> systemd upstream authors has convinced me that getting this right will
> be possible, but the pieces are not yet available.
Agreed.
- Josh Triplett
Reply to: