Please test sysvinit package from experimental
The last few days I have made an effort to fix a few bugs in the boot
system. The changes need some testing, as I do not have all the
affected hardware to test on. I've tested the changes in qemu, but
that do not test some of the changes, like the part that avoid fsck
when a laptop is running on battery.
Here is the list of changes uploaded since the one in testing and
unstable. I am hoping to close even more of the 200 open bugs before
Lenny, but this is how far I got for now. :)
sysvinit (2.86.ds1-42) experimental; urgency=low
* Change XS-Vcs-Svn to Vcs-Svn as the control flags are official now.
* Add Vcs-Browser flag in debian/control, documenting the browsable source.
* Change how init.d/rc call scripts, to make sure sourced scripts
get the correct argument.
* Reduce redundant code in init.d/rc, to make it easier to enable
sourcing of *.sh scripts. It is still not possible to source
scripts, as some of them use exit and terminate the boot. This is
related to bug #339955.
-- Petter Reinholdtsen <email@example.com> Tue, 25 Dec 2007 20:56:44 +0100
sysvinit (2.86.ds1-41) experimental; urgency=low
* Update standards-version from 3.7.2 to 3.7.3. No change needed.
* Remove empty /lib/ directory from the sysvinit and sysvinit-utils
* Remove the empty /usr/share/man/man5/ from the sysv-rc package.
* Remove the registration of of init.d/modutils from the
initscripts/postinst, as the package is removed from Debian.
-- Petter Reinholdtsen <firstname.lastname@example.org> Mon, 24 Dec 2007 13:49:17 +0100
sysvinit (2.86.ds1-40) experimental; urgency=low
[ Petter Reinholdtsen ]
* Rewrite libata shutdown handling to use patch from
Werner Fink at SuSe to handle each disk individually and
automatically instead of guessing in the init.d script how
to handle the collectively (Closes: #426224). This removes
the HDDOWN default/halt variable.
* Improve usage information in update-rc.d by using the same
notation as the one used in the manual page (Closes: #268713)
* Implement noswap kernel boot option to stop automatic
swap activation at boot time (Closes: #388525). Based on patch
from Michael Prokop.
* Rewrite init.d/rc to use shell features instead of calling sed
(Closes: #406393). Patch from Ivan Baldo.
* Fix typo in update-rc.d example code (Closes: #433378).
* Fix minor typo in rc*.d/README files (Closes: #452384).
* Make sure init.d/rc do not complain because the new progress bar
API is unavailable (Closes: #457544).
* New patch 26_last_ipv6 from Fedora to recognize more IPv6
* Extend the 68_init_quiet patch based on patch from Fedora to also
suppress the "Switching to runlevel" message. This patch is
currently disabled because the 'quiet' flag is removed by the
kernel when init is called.
* New patch 69_init_waiting from Fedora to make sure the init
waiting status is preserved across re-exec.
* New patch 41_utmp_64bit from Fedora to avoid writing past the
utmp.ut_tv struct on 64 bit architectures.
* New patch 64_init_init_reexec_env from Fedora to fix typo. Now
passes environment on to the new process when re-exec is used.
-- Petter Reinholdtsen <email@example.com> Sun, 23 Dec 2007 19:44:13 +0100
sysvinit (2.86.ds1-39) experimental; urgency=low
[ Petter Reinholdtsen ]
* Upload to experimental to get more feedback from testers.
* Change checkfs.sh dependency info, make modutils an optional
dependency as it is no longer installed by default.
* Move /etc/init.d/bootclean to /lib/init/bootclean, as it is not an
init.d script but a library.
* Acknowledge NMU. (Closes: #433386)
* Make /lib/init/vars.sh usable when 'set -e' is in effect. Patch
* Rename 20_pidof.dpatch to 80_killall_pidof.dpatch, to group the
killall5 patches together.
* New patch 82_killall_exclude_pids.dpatch adding new option -o
to killall5, to omit killing the pid given on the command line.
Based on patch from Ubuntu. (Closes: #453042)
* Use new killall5 -o option in init.d/sendsigs, reading pids to
omit from /var/run/sendsigs.omit. This make it possible to list
pids there if the process is required to umount the file systems
during shutdown. Make initscripts depend on sysvinit-utils
(>= 2.86.ds1-39) to make sure killall5 support the -o option.
Based on patch from Ubuntu.
* Stop progress bar on ltsp-client-core too. The script was renamed
in recent ltsp versions. Based on patch from Ubuntu.
* Add linux-gnulp to debian/deps-mount, to get the correct behaviour
with regard to initscripts dependencies on lpia. Patch from Adam
Conrad in Ubuntu.
* Add splash screen / progress bar API to initscripts in
/lib/init/splash-functions-base, overridable by the splash
implementations in /lib/init/splash-functions. Based on patch
from David Härdeman. (Closes: #423095) Make initscripts conflict
with usplash (<< 0.5) to make sure a version with PULSATE support
* Rewrite init.d/rc progress bar support to use new API. Move
usplash calls to separate file splash-functions-usplash. This
file should be moved to the usplash package. Based on patch from
* Do not mount gfs2 as a local file system. It is a network file system.
Patch from Guido Guenther (Closes: #454468).
* Modify update-rc.d to refuse to install init.d scripts with illegal
characters in the names. Only accept [a-zA-Z0-9+.-] to match the
limits on package names. Correct handling of init.d scripts with
+ in their name (Closes: #431224).
[ Henrique de Moraes Holschuh ]
* libata shutdown handling fixes:
Check http://linux-ata.org/shutdown.html for information
* init.d/halt: do not issue -h to halt(8) when the kernel supports
* init.d/halt, default/halt: add HDDOWN option to override initscript
halt -h logic
* Add halt(5) manpage, updated for HDDOWN. Thanks to Casper Gielen for
the manpage. (Closes: #407211)
-- Petter Reinholdtsen <firstname.lastname@example.org> Sat, 22 Dec 2007 17:49:42 +0100
Please test, and report to BTS if there are any new problems.