[11:18] <jk44> hi, i have noticed a possible bug in upstart
[11:19] <jk44> i wrote this job :
[11:19] <jk44> console output
[11:19] <jk44> pre-start script
[11:19] <jk44> 	echo "**UPSTART : ************* launching RCS ********** "
[11:19] <jk44> end script
[11:19] <jk44> script
[11:19] <jk44> 	echo "avant exec"
[11:19] <jk44> 	exec /etc/init.d/rcS
[11:19] <jk44> 	echo "apres exec"
[11:19] <jk44> end script
[11:19] <jk44> post-start script
[11:19] <jk44> 	echo "will try to launch runlevel 3"
[11:19] <jk44> 	echo "done rlv 3"
[11:19] <jk44> 
[11:19] <jk44> 	echo "gona launch rcS"	
[11:19] <jk44> end script
[11:19] <jk44> and post start is executed before exec script
[11:20] <Keybuk> incorrect
[11:20] <Keybuk> post start is executed alongside the exec script
[11:21] <jk44> ok, it explain what i get
[11:21] <Keybuk> which wins is dependant on such matters as your kernel, and possibly the one that outputs first is running second, and it's the console buffer that's fooling you
[11:21] <Keybuk> post-start makes most sense when the process run by exec is not immediately redy
[11:21] <Keybuk> you can use the post-start script to poll the daemon until it's accepting connections
[11:22] <Keybuk> the job won't be marked running until the post-start script exits
[11:22] <jk44> ok, thk very much
[11:22] <jk44> another dub question :
[11:23] <jk44> when a job fail starting, how can i recover, (as it has not failed)
[11:23] <jk44> and start the other jobs ?
[11:23] <jk44> is "started failed or started"
[11:23] <jk44> the solution ?
[11:24] <jk44> i am trying that, but quite unsuccessfully
[11:29] <Keybuk> start on stopped JOB failed
[11:30] <Keybuk> the stopped event for the job will include the argument "failed", and will include arguments and environment saying what failed, etc.
[11:30] <Keybuk> http://upstart.ubuntu.com/wiki/JobFailedEvent
[11:30] <Keybuk> ^ see that
[11:30] <Keybuk> e.g.
[11:30] <Keybuk> stopped apache failed
[11:30] <Keybuk>   EXIT_SIGNAL=SIGSEGV
[11:31] <jk44> ok, sorry for bothering with such uninterresting question ^^, i had not very well understood the doc (poor english :s)
[11:34] <Keybuk> that's ok
[11:34] <Keybuk> the docs aren't great yet
[02:38] <AlexExtreme> hmm
[02:38] <AlexExtreme> this is weird
[02:38] <Keybuk> hmm?
[02:39] <AlexExtreme> i've got some native jobs written for LFS
[02:39] <AlexExtreme> (just playing around)
[02:39] <AlexExtreme> it boots fine
[02:39] <AlexExtreme> apart from somewhere a bash syntax error gets printed
[02:39] <AlexExtreme> it says the syntax error is in /dev/fd/7 every time
[02:40] <AlexExtreme> i can't figure out why
[02:40] <Keybuk> heh
[02:40] <Keybuk> /dev/fd/7 is the name of the shell script upstart executes if it's > 1024k or so
[02:41] <AlexExtreme> strange
[02:42] <Keybuk> why?
[02:42] <AlexExtreme> i have no job even 2kb
[02:42] <Keybuk> that's a little more unusual
[02:42] <Keybuk> oh, sorry, 1024 bytes
[02:42] <Keybuk> not K
[02:42] <Keybuk> I always do that
[02:42] <Keybuk> 1K
[02:42] <AlexExtreme> ah
[02:42] <AlexExtreme> that limits it to 2 jobs
[02:44] <Keybuk> (the /dev/fd thing is because upstart makes a pipe on which it arranges the script to be executed; and that's the trick to passing it to sh as a file)
[02:44] <AlexExtreme> k
[02:44] <AlexExtreme> the weird thing though, is that i get no "job failed" message from upstart
[02:44] <Keybuk> passing it on "standard input" would mean the script couldn't use exec < or things like that
[02:44] <Keybuk> and passing it in the arguments could exceed available space
[02:49] <AlexExtreme> ah, got it
[02:49] <AlexExtreme> stupid type
[02:49] <AlexExtreme> *typo
[02:50] <Keybuk> dunno why it doesn't fail
[02:50] <Keybuk> maybe you've got a set +e in there?  or an ||/&&, or it's part of an if, etc.
[02:51] <AlexExtreme> yeah, it was part of a ||
[06:35] <AlexExtreme> now *that's* what i call optimization
[06:35] <AlexExtreme> http://www.alex-smith.me.uk/files/bootchart-ud.png
[06:35] <AlexExtreme> :D
[06:37] <ion_> :-)
[07:38] <tale> AlexExtreme, I'm not an upstart expert, but that distro wouldn't do much, right?
[07:39] <AlexExtreme> the boot sequence does actually run more stuff, just it seems that bootchart missed some stuff
[07:39] <AlexExtreme> currently it boots to a text-based multiuser system with networking and ssh
[07:40] <tale> ah
[07:40] <tale> what are some of the fastest boot times you guys are getting with the latest development code?
[07:41] <AlexExtreme> that is the fastest boot i have ever got with any init system
[07:44] <tale> but, what about a normal ubuntu install?
[07:44] <AlexExtreme> i haven't tried with ubuntu, since i don't use it ;)
[07:44] <tale> what distro do you use?
[07:44] <tale> I assumed ubuntu since most distros don't use upstart
[07:45] <AlexExtreme> frugalware atm