Maintainer notes for sysvinit News ---- Last updated 16 May 2009 Stuff to implement ------------------ * Do not allow system to continue booting if both fsck and sulogin fail (#337444)? * Make sulogin try /bin/sh and /bin/sash too (#40144, #43317)? * Port bootlogd to use posix_openpt(3), since BSD PTs are now deprecated and /dev/pts _is_ now available at the time that bootlogd starts (...in Debian)? Stuff to do after etch ---------------------- * mountnfs.sh: Don't start portmap Ubuntu changes to backport -------------------------- (none known ATM) Stuff to send upstream ---------------------- debian/patches/fix-minor-warnings should be applied upstream. So should a lot of the other patches. src/Makefile should not set CC. set -e ------ Note that many scripts do not use "set -e". This is deliberate. Instead, return statuses are checked. PATH ---- * Current policy: + Leave unset if absolutely no pathless commands are run; otherwise... + set to "/sbin:/bin" in /etc/init.d/rc and in scripts that run prior to mountnfs.sh; otherwise... + set to "/usr/sbin:/usr/bin:/sbin:/bin". + Add "/lib/init" to beginning of PATH if readlink command or domount function used. This will disappear once coreutils 5.94 makes it into testing, and we Depend on that. * Question: Should we include /usr/local/sbin and /usr/local/bin too? I don't see this question addressed in Policy. Ref: #227540 Dependencies ------------ To go away: * sysvinit Pre-Depends on initscripts, sysv-rc | file-rc so that the conffiles in the old monolithic sysvinit package are taken over by the initscripts package. (Only a Pre-Depends is strong enough to ensure that the conffiles are taken over without an orphanic hiatus.) Strictly speaking we can drop this dependency now since everyone is either installing anew or else has upgraded to sarge; and Debian does not support skip upgrades. However, I don't see any harm in keeping this dependency around until after etch releases. But then it should go away. * initscripts Depends on mount >= 2.11x-1 because that's when swapon's "-e" option was added. Sarge satisfies this requirement so this dependency could be dropped. We can drop it after etch releases; then both stable and oldstable will satisfy the requirement. * initscripts needs to Depend on debianutils >= 2.12.0 until etch has been released so that which can be assumed to be in /bin (rather than in /usr/bin, as it was before, including in 2.8.4 in sarge) * initscripts Conflicts with and Replaces mdutils which was last seen in potato. Can disappear after etch. * initscripts Conflicts with sysvinit (<< 2.86.ds1-12) because earlier versions of bootlogd didn't have the -c option which the bootlogd initscripts now uses Questionable: * initscripts Replaces various libc packages for reasons described here: http://lists.debian.org/debian-hurd/2004/06/msg00048.html. Anyone know how long this dependency needs to be carried? * sysv-rc (current) Conflicts with and Replaces file-rc. file-rc (0.8.7) Replaces: sysv-rc but does not Conflict with it. I am not sure that sysv-rc needs to Replace: file-rc. Isn't Conflicting sufficient? And why doesn't file-rc Conflict with sysv-rc? Strange. Possible future: * If the mtab initscript is ever changed so that it updates the mtab file without rerunning mountdevsubfs.sh then initscripts should be made to Conflict with udev << 0.080-1. (The latter versions of udev put the udev initscript at S04, _after_ S04mountdevsubfs.sh, and the script unmounts /dev/pts and /dev/shm.) Speed and compatibility ----------------------- People running Debian on low-memory systems say: * Use built-ins * Avoid fork-and-exec, pipelines, $() * Reduce the number of commands within pipelines * Avoid unnecessary code * Avoid commands that don't work the same way in BusyBox: 'nice', 'find', 'mount', 'umount', 'init', 'halt', 'shutdown', 'syslogd', 'klogd', 'hwclock', 'cron', 'anacron', 'crontab'. Regenerate /etc/default/rcS? ---------------------------- The current /etc/default/rcS template just refers the admin to rcS(5). This way we have only one set of docs to maintain, and we don't have to worry about out-of-date comments in /etc/default/rcS. I wonder if we should strip comments out of existing /etc/default/rcS files, leaving behind only the reference to rcS(5).