Description of problem: Annoying startup and rc1, rcS, rcS-sulogin message glitches, man page typos. Version-Release number of selected component (if applicable): event-compat-sysv-0.3.9-13.fc9.noarch initscripts-8.70-1.i386 NetworkManager-0.7.0-0.9.2.svn3566.fc9.i386 upstart-0.3.9-17.fc9.i386 How reproducible: very Steps to Reproduce: 1. boot to single user mode & exit see rcS killed message 2. boot normally see some [OK] on the second line, and rcS killed message 3. telinit 1 and see rc1 and rcS-sulogin killed messages 4. read events(5) and initctl(8) man pages see typos. Actual results: user gets annoyed and anxious Expected results: startup messages shouldn't look like errors Additional info: Files patched: /usr/share/man/man5/events.5.gz /usr/share/man/man8/initctl.8.gz /etc/rc.d/init.d/NetworkManager /etc/event.d/rc1 /etc/event.d/rcS /etc/event.d/rcS-sulogin /etc/rc.d/rc.sysinit /etc/rc.d/init.d/single They belong to packages: event-compat-sysv-0.3.9-13.fc9.noarch initscripts-8.70-1.i386 NetworkManager-0.7.0-0.9.2.svn3566.fc9.i386 upstart-0.3.9-17.fc9.i386 Patch will be attached. Bug will be cloned and attached to those four components.
Created attachment 303642 [details] Patch which fixes those annoyances. The change to /etc/rc.d/init.d/single makes it completely dispensable. All it does is clear the screen. /etc/event.d/rc1 now does the telinit S. But you wouldn't want this version of init.d/single installed without the matching event.d/rc1.
I'd rather just nuke S99single for the upstart case. (And also move event-compat-sysv into initscripts, as it would make things like this a lot simpler.)
Created attachment 303655 [details] revised patch to rcS-sulogin to fix a different bug. Noticed another bug: if the last command you execute in bash has a nonzero exit code, and you exit by typing "exit" rather than "exit 0", you can't do anything except CTL-ALT-DELETE to reboot.
I'm not seeing why you need a kill timeout/trap in rc1.
In fact, unless I'm missing something, all the 'kill timeout/trap' can be elided by simply doing 'exec telinit ...'
Hm, that just appears to shorten the race. That being said, the kill/trap just seems wrong as a way to avoid the message.
Moving the telinit machinations to post-stop should do the trick. Will generate a new patchset momentarily.
Created attachment 303679 [details] patch for this issue Here's an incremental version of what I'm committing.
Will be in 0.3.9-14.fc9 and initscripts-8.72-1.
s/echo "Telling INIT/echo $"Telling INIT/ since it is a translated string. rcS-sulogin post stop should only do telinit if runlevel is still S. Otherwise, user must have already asked for a different runlevel and he won't like being overruled. ------- man event(5) says: post-stop Execute the command/script after stopping the process. The format is similar to pre-start. That seems to imply that the command/script only runs if upstart stopped the process, not if the process exited. s/stopping the process/the process has exited./ Hmm, looking at the SRPM, I see events.5 is very new and I am speaking to the author... so I guess I don't need to attach this comment to the upstart package bug...
s:/bin/bash || true:exec /bin/bash: Cleaner: /bin/bash is the rcS-sulogin main process. Presumably upstart will report nonzero exit code.
I'm not sure we *care* about the exit code of bash, though. In any case, will do some poking along the lines of comment #10.
http://git.fedorahosted.org/git/?p=initscripts.git;a=commit;h=d78b58c6fb9110d78f8750b4d3779a401d2ebf92 http://git.fedorahosted.org/git/?p=initscripts.git;a=commit;h=89388cd39edcc42ae355ff854f14649822591263 Should be solved in initscripts-8.73-1.
Changing version to '9' as part of upcoming Fedora 9 GA. More information and reason for this action is here: http://fedoraproject.org/wiki/BugZappers/HouseKeeping
This seems to all be sorted. Can we close this?
Closing this, since nobody protested.