Re: bin/39603: postfix doesn't see all "active" network interfaces

["Martin S. Weber" <Ephaeton%gmx.net@localhost>]

The following reply was made to PR bin/39603; it has been noted by GNATS.

From: "Martin S. Weber" <Ephaeton%gmx.net@localhost>
To: David Holland <dholland-bugs%netbsd.org@localhost>
Cc: gnats-bugs%NetBSD.org@localhost
Subject: Re: bin/39603: postfix doesn't see all "active" network interfaces
Date: Sat, 27 Sep 2008 16:14:10 -0400

 On Sat, Sep 27, 2008 at 06:35:28PM +0000, David Holland wrote:
 > On Fri, Sep 26, 2008 at 03:00:05PM +0000, Martin S. Weber wrote:
 >  >  Yeah well, it was my "broken configuration" (Matthias Scheler) userland. 
 >  >  (mail-index.netbsd.org/pkgsrc-users/2007/09/12/0006.html in response to
 >  >   mail-index.netbsd.org/pkgsrc-users/2007/09/12/0002.html -- a problem 
 > which
 >  >   still exists but given the rest of the thread you're not really 
 > interested
 >  >   in hearing about it anyways).
 > 
 > Yeah, whatever. A system where both userland and kernel are built
 > without IPv6 ought to work. If it doesn't, it's a bug, even if some
 > people might consider it a low priority.
 
 That's what I was thinking, too, especially that when using some of
 the "builtin" MK* variables, your system -never- should come out
 unusable. But on different occassions different people answered me
 like m.s. so at some point you're just giving up.
 
 > 
 > What's happening with IPv6 that causes the symptoms you reported?
 > 
 
 I don't know what's wrong with postfix really, the error message from
 the kernel is the best clue I have and thinking of IPv6 was only the
 last thing I did (because gif(4) mentions IP6). I'd expect 'gif' to
 work for ipv4 over v4 tunneling too, so it really might be considered
 a bug in postfix.
 
 As soon as I install a userland without MKINET6=no, and run on a kernel
 without IP6, I'm fine. IPv6 seems present but it won't work so it's
 just ugly and verbose but not as evil as running with IPv6 supported.
 
 > Also note PR 38114. :-/
 
 At least for me it built; then again I was building natively on a
 i386 for i386, and not on/for evbarm...
 
 The build cluster should imho build with every combination of MK*
 variables possible, even if only once a week, and bugs / build-failures
 rising up there should be considered seriously. Then again I understand
 this is an oss project and if nobody's interested in fixing breakage
 introduced by too deeply embedding IPv6, then tough luck.
 
 I at least don't have the expertise in that area, and surely not the
 time to get it within any considerable amount of time.
 
 the only further thing I could do is opening (another) PR saying
 "IPv6 is embedded too tightly into the system" but I guess I'm only
 hearing a mix of silence and insults for it. NetBSD and IPv6 is picky,
 just as ad@ and any non-ad thread library.
 
 -Martin