[Pkg-sysvinit-devel] Experimental release?

Mon Dec 12 22:14:04 UTC 2005

[Thomas Hood]
> Although the changes are mostly minor and I have tested the results,
> the sheer number of changes makes me think that it would be better
> to upload the next sysvinit release to experimental.

Yes, or at least get a few more positive test reports before we upload
it into sid.

> I don't plan to make any more changes other than fixes for bugs in
> the changes I have already made, and we probably won't find any more
> such bugs until the package starts getting tested, so I would
> suggest that the package be released soon.

I tried to test the svn version now, but am unable to test it.  Its
dependencies (debianutils) are missing in sarge, and I only have sarge
test machines around.  I've tested it in qemu on an LTSP client, but
that only tests a small fraction of the boot process.

It might take a while before I have a sid machine to test on.  The
only etch machine I have do not have a monitor nor serial console, and
it is placed in a machine room far away from my home, so I can't test
there either. :(

Anyway, I am willing to upload into experimental right now, and to sid
when a few positive test reports have been sent to this list. :)

What do the rest of the sysvinit team think about a new release?  The
current changelog is rather massive:

sysvinit (2.86.ds1-7) UNRELEASED; urgency=low

  [ Petter Reinholdtsen ]
  * Add commented-out code to /etc/init.d/rc for sourcing
    *.sh scripts for runlevel 'S'.  This is preparation for fixing
    #339955.  Actually fixing it has to wait until other packages
    remove "exit" from their .sh scripts.

  [ Thomas Hood ]
  * *.sh: Make sure that these do their thing when they aren't given any
    command line arguments (as is the case when they are sourced)
  * initscripts: Improve use of log_* functions  (Closes: #55143, #323749)
    There remains the problem that error messages printed by programs
    that are called by initscripts get printed on the console and thus
    mess up single-line-format messages printed by the log_ functions.
    I asked at http://lists.debian.org/debian-devel/2005/11/msg01309.html
    how this should be dealt with.
  * checkroot.sh: Fix double printing of 'Done checking root file system';
    mountall.sh: Fix chopped-up printing of mount information
    (Closes: #339979, #331397, #341097 and presumably closes: #332309)
  * checkroot.sh, checkfs.sh: Save fsck logs  (Closes: #189356)
    Thanks to Theodore Y. Ts'o.
  * checkroot.sh, mountvirtfs: Eliminate use of dir_writable in order to
    try to please selinux  (Closes: #333836)
  * checkroot.sh: Only run findfs for mount on /  (Closes: #275680)
    Thanks to Cameron Hutchison for the patch.
  * mountall.sh, mountnfs.sh: Split call to bootclean out into separate
    script called at the next sequence number  (Closes: #286479)
  * bootmisc.sh: Shorten motd header  (Closes: #340017)
  * bootmisc.sh: Rotate dmesg log with savelog (Closes: #237074)
    and chgrp adm.
  * umountroot: Change mount command in order to exclude bind mounts of
    the root directory  (Closes: #339023)
  * Replace /etc/rc1.d/20single by /etc/rc1.d/S30killprocs and
    /etc/rc1.d/S99single so that packages can insert scripts to do
    things between the "killall5" in the former and the "exec init -t1 S"
    in the latter.  This may help to address #145280.
  * umountfs: Unmount tmpfs before swapoff and other fs's afterwards
    (Closes: #328582, hopefully without reopening #84782)
  * Make stop-bootlogd a distinct script rather than a symlink to bootlogd.
    Give it a different LSB header.
  * Add stop-bootlogd-single initscript to stop bootlogd in "single" mode
    (Closes: #213028, #220025)
  * checkfs.sh: Implement FSCKTYPES  (Closes: #89481)  Set, e.g.,
    FSCKTYPES="ext2,msdos" to fsck only file system types ext2 and msdos
    See fsck(8) for syntax.  FSCKTYPES="none" disables fsck of file
    systems (other than the root filesystem) altogether.
  * initscripts: Set variables using common script
  * initscripts: Fix usage messages
  * initscripts: Allow VERBOSE to be set via the INIT_VERBOSE=yes kernel
    argument (Closes: #286082)  Experimental.  The name of the kernel
    argument may change in the future if we decide to implement this more
    generally.
  * Improve skeleton initscript
  * all scripts: Clean up and standardize indentation and formatting
  * all scripts: Use the which program to test for executables rather
    than fixed paths
  * initscripts preinst: Remove ancient /etc/init.d/boot conversion code
  * default config files: Clean up comments
  * Make initscripts Depend on debianutils >= 2.13.1 in order to prevent
    #295850.  Note that debianutils also has to be >= 2.12.0 so that the
    "which" program is available in /bin.
  * Remove currently unneeded sysvinit Dependency on coreutils
    (Closes: #316423)
  * pidof: If the program is specified with a path, don't match processes
    of programs run from different paths  (Closes: #160329)
  * init: 40_selinux.dpatch: Remove superfluous printf argument
  * init: Add 63_init_longer_procname.dpatch  (Closes: #336172)
  * init.8: Mention that entering runlevel 1 kills all processes
    (Closes: #238861)
  * update-rc.d.8: Mention that update-rc.d will not create multiple start
    or multiple stop symlinks for a service in a single runlevel directory.
    (Closes: #330155)
  * Update FSF address

 -- Petter Reinholdtsen <pere at debian.org>  Sat, 19 Nov 2005 21:11:34 +0100

Friendly,
-- 
Petter Reinholdtsen