nosh version 1.18

From: Jonathan de Boyne Pollard <>
Date: Thu, 20 Aug 2015 14:01:22 +0100

nosh is now up to version 1.18


The big news for this release is the nosh-run-system-manager Debian
binary package. This, and the new additional service bundles in
nosh-bundles, package up everything that is needed for running an
entirely nosh-managed basic Debian system with the nosh system-manager
program as process #1. And so the entry on the roadmap WWW page is
crossed out. Some notes:

* Don't forget that the Nosh Guide has a whole chapter on troubleshooting.
* With that package alone, you get very little running. This is
intentional. You'll have to install other nosh-run packages, or add
presets, for the various other things that you want. To get an OpenSSH
server running, for example, you'll need a local preset file (named,
say, /etc/system-control/presets/20-sshd.preset) with "enable sshd" and
"enable cyclog_at_sshd" before (re-)installing nosh-bundles.
(Re-)Installing the nosh-bundles package (re-)applies all current
presets, including your local ones, and auto-starts all enabled services.
* If you are running the freedesktop services, read the notes
hyperlinked-from the package download page.
* You may have spotted that there's a choice between running udev and
busybox mdev. (You pretty much must run one or the other for a fully
functional system.) The nosh-run-busybox-mdev package is broken. I
forgot to write the adapter tool. I've written it ready for version
1.19. There will be more said on the subject of busybox mdev in the
1.19 announcement, therefore.
* It's also intentional that you don't get System 5 shim commands for
the likes of "telinit" and "halt" unless you install the
nosh-systemv-shims package. "system-control poweroff" works without the
presence of the shims, of course.
* For novices, I recommend starting with nosh-run-kernel-vt .
nosh-run-user-vt still requires a manual step, after re-building the
service configuration each time, of "system-control disable
* The recovery mode misbehaviour is a known problem. I'm
investigating. As a local fix, boot with init=/bin/sh on the kernel
command line and then run "exec /sbin/init -s" or even "exec /sbin/init
-b" from that shell prompt.

This is not the only news, of course. The BSD crowd should not feel
left out, moreover.

There are four long-standing problems with the Linux libkqueue library.
One of those problems causes svscan a.k.a. service-dt-scanner to be
spuriously woken up. This doesn't affect Debian but does affect Linux
operating systems such as Gentoo that have more recent versions of that
library. This has been worked around in version 1.18.

The pre-built mount_at_-, fsck_at_-, mount_at_-usr, fsck_at_-usr, mount_at_-var, and
fsck_at_-var service bundles have been removed. Generation of the service
bundles for mounting and checking volumes is now entirely based upon the
auto-creation system in /etc/system-control/convert/ . If you are
installing from scratch by hand, then you must remember to "redo all" in
that directory. The nosh-bundles package does this for you as part of
its post-install procedures.

The problem with the local-syslog-read service on Linux providing the
wrong socket (the BSD one) has been fixed.

The tools now speak true TAI, rather than UTC-10. There's an
explanation of the consequences of this in the manual pages for cyclog,
tai64n, and tai64nlocal.

The /etc/fstab conversion system now recognizes remote filesystem types
and attaches the generated services to .
Received on Thu Aug 20 2015 - 13:01:22 UTC

This archive was generated by hypermail 2.3.0 : Sun May 09 2021 - 19:44:19 UTC