[20:30] <mistergibson> is this the channel for ubuntu networking stuff?
[20:32] <xopah> Hello could someone help me in a matter where I have hard to connect to a multiple SSID WLAN network?
[20:37] <Keybuk> xopah: that has nothing to do with Upstart, sorry
[20:45] <ion> Two persons from totally different parts thought this channel has something to do with networkingithin a couple of minutes? Interesting occurrence.
[20:45] <ion> networking within
[20:48] <xopah> keybuk: could you please point me in the right direction?
[20:48] <xopah> Wher shuld I turn?
[20:49] <sadmac> xopah: #ubuntu would be a good next stop
[20:49] <xopah> ion: we were recommended to go here from #ubuntu ...
[20:49] <sadmac> Keybuk: ^^WTF?!?
[20:50] <xopah> sadmac: thanks. I'll try there again.
[20:50] <sadmac> xopah: np
[20:50] <sadmac> hmm. Is #ubuntu having troll problems?
[20:52]  * sadmac imagine's Scott's "you're making absolutely no sense" face
[20:52] <Younder> I am having problems with programs due to a pygtk not beeing seen
[20:53] <Younder> there are multiple python versions and it worked under gentoo
[20:53] <sadmac> Younder: how does this relate to upstart?
[20:54]  * sadmac begins to get a sinking feeling
[20:55] <Younder> sorry I'll go back to the python group.. take my chances
[20:58]  * Keybuk wonders whether there's a channel mode flag
[20:58] <Keybuk> to refuse entry to people also in #ubuntu
[20:58] <sadmac> Keybuk: wouldn't that block, say, you for example?
[20:58] <Keybuk> no
[20:58] <Keybuk> I'm not in #ubuntu
[20:59] <sadmac> Keybuk: heh. maybe I don't understand #ubuntu
[20:59]  * sadmac is in #fedora
[20:59] <Keybuk> Ubuntu is where users go to ask questions
[20:59] <Keybuk> err #ubuntu
[21:00] <sadmac> Keybuk: yeah. so is #fedora. I like to meet the people
[21:00] <Keybuk> we have more users than you
[21:00] <Keybuk> a LOT more users
[21:00] <Keybuk> they whine more too
[21:00] <sadmac> Keybuk: yeah, I don't know that I'd like to meet /your/ users. Ours troll more than yours I'd imagine though
[21:00] <Keybuk> at least yours might know something of what they're talking about :p
[21:01] <sadmac> Keybuk: actually I think 3/4 of the people giving help at any time in #fedora just have irssi set up to periodically say "www.justfuckinggoogleit.com"
[21:01] <ion> sadmac: Just for fun, join #ubuntu for a minute. It’ll be an... interesting experience. :-P
[21:02] <ion> “I visited #ubuntu shortly and survived”
[21:02] <sadmac> Keybuk: again though, if you do have so many more users, wouldn't you also be blocking half the relevant portion of the internet?
[21:02]  * raphael__ wonders how #ubuntu, #fedora and #debian compare
[21:02] <sadmac> ion: My God, its full of stars!
[21:03] <ion> It’s full of trolls and the blind trying to lead the blind! rather.
[21:03] <sadmac> #debian is nearly as big and not 1/10 as noisy
[21:04] <notting> ion: that doesn't help me delineate between the three
[21:04] <sadmac> ion: they seem to be under the impression that they're linux users of some kind
[21:04] <sadmac> from what I've heard, #debian has all /their/ advice macro'd to "go install Ubuntu"
[21:05] <raphael__> sadmac: don't think so, it's usually /msg dpkg foo
[21:07] <sadmac> raphael__: the users know what dpkg is? sounds like a nice place.
[21:07] <Keybuk> sadmac: dpkg is a bot on #debian
[21:07]  * raphael__ wonders how long Keybuk is going to wait before setting mode +m to stop this offtopic conversation :)
[21:08] <sadmac> Keybuk: oic
[21:08] <Keybuk> raphael__: this channel rarely gets on-topic :p
[21:08] <raphael__> Keybuk: but from what I've seen off-topic conversations don't last this long ;)
[21:08] <sadmac> Keybuk: fine. on topic: I'm finally edging near a workable nih_parse, which is, of course, making me want to replace nih_error while I'm at it :)
[21:09] <Younder> Yes, it would be terrible if that actual users of upstart got a say :)
[21:10] <sadmac> Younder: what are you talking about?
[21:11] <Keybuk> sadmac: what were you going to change?
[21:11] <Younder> Well at present the boot process, to me at any rate, is very confused. I can't see where sysv ends and upstart begins.
[21:11] <sadmac> Keybuk: probably not change so much as a new mechanism, along the lines of the lisp condition/restart thing that I've poked at a few times
[21:12] <Keybuk> Younder: depends which distro you look at
[21:12] <Younder> ubuntu 9.10
[21:12] <sadmac> Keybuk: it would also make a nice replacement for nih_log while we're at it (probably with no API change there)
[21:12] <Keybuk> most actual Linux distributions tend to have the Upstart /sbin/init daemon run the SysV-style rc scripts (/etc/init.d)
[21:13] <Keybuk> some distributions (like Ubuntu) do far more in a Upstart native way than others (like Fedora)
[21:13] <Keybuk> embedded Linux platforms (Chrome OS, Palm Web OS, Maemo, etc.) often use Upstart entirely natively
[21:13] <Keybuk> Younder: read you read "man runlevel" ?
[21:13] <Keybuk> sadmac: I know I'm going to regret this, but ... explain? :p
[21:14] <sadmac> notting: since I'm thinking of it, when do you want 0.6 tagged into rawhide? Will you just do it yourself when initscripts is ready?
[21:14] <Keybuk> Younder: that should be "man 7 runlevel"
[21:14] <sadmac> Keybuk: basically there is a function/macro called nih_sig()
[21:14] <Younder> Keybuk, sure, but gtk is not on any runlevel. Tomcat and apache, however, are.
[21:14] <Keybuk> Younder: that's because GTK+ has nothing to do with Upstart
[21:14] <Keybuk> and Upstart has nothing to do with GTK+ either
[21:15] <Keybuk> my toaster is also not on any runlevel
[21:15] <sadmac> Keybuk: if (some_error) nih_sig (TYPE_OF_ERROR, &some_info, &some_more_info);
[21:15] <notting> sadmac: hrm. i haven't looked at the other apps. so i'll probably work on it tomorrow or thursday
[21:15] <Keybuk> yet I can still make toasted goodness
[21:15] <Keybuk> sadmac: that's basically how nih_error behaves, no?
[21:15] <Younder> and ubuntu has nothing to with upstart either..?
[21:15] <sadmac> Keybuk: well, what nih_sig does is interesting
[21:15] <Younder> They just created it
[21:15] <Keybuk> Younder: Ubuntu uses Upstart, yes
[21:16] <michas> hi, I am trying to debug some ubuntu init.d scripts by adding some echo lines. Unfortunately they print nowhere. :( If I understood everything right, upstart is the one who emulates old init and executes the scripts. - Is there a way to make that echo output visible?
[21:16] <ion> Younder is clearly a troll. Let’s not feed it anymore. :-P
[21:16] <Younder> I guess what I wnat to know is: is there a reference manual for upstart?
[21:16] <sadmac> Keybuk: nih_sig runs a handler for the error, which is a function pointer registered beforehand
[21:17] <Keybuk> Younder: plenty, start at "man upstart" on your system?
[21:17] <sadmac> Keybuk: that function runs right away, and either 1) corrects the error (notice the args passed by ref) 2) longjmps, or 3) aborts
[21:17] <Keybuk> ah
[21:17] <Keybuk> more like dpkg-style
[21:17] <Younder> Keybuk, info?
[21:17] <Keybuk> you have ohshit()
[21:17] <raphael__> btw, what's all the nih_foo? I saw it on ureadahead and seemed like it is being used in other places, but it's sort of strange that it doesn't seem to be a shlib
[21:18] <sadmac> Keybuk: never seen dpkg code
[21:18] <Keybuk> ohshit calls any registered error handlers and longjmps back up
[21:18] <Keybuk> raphael__: it is in lucid
[21:18] <Younder> never mind
[21:19] <sadmac> Keybuk: the nih_log thing comes in under the "corrects the error" option. nih_log just calls nih_sig (LOG_EVENT_$LEVEL, log_text)
[21:19] <Keybuk> oh, I see where you're going with that
[21:19] <sadmac> Keybuk: the default handler does just what nih_log does now, but because there's a handler stack, the log line is now also a sort of tracepoint when you're debugging
[21:19] <Keybuk> it'd be nice if you could thwart the error ;)
[21:20] <sadmac> Keybuk: yeah, that too
[21:20] <Keybuk> e.g. if you raise ENOFILE, an error handler could increase the rlimit, then make you try again
[21:20] <sadmac> Keybuk: precisely the idea
[21:20] <Keybuk> but then that requires while loops and stuff ;)
[21:20] <Keybuk> while (some_syscall (...) < 0)
[21:20] <ion> Yeah, the condition/restart thing is powerful.
[21:20] <Keybuk>   nih_sig (...)
[21:20] <Keybuk> ie. assume that nih_sig solves the problem or longjmps you out
[21:21] <Keybuk> this is basically how dpkg works
[21:21] <sadmac> Keybuk: I'm pulling apart libunwind now to see if it has anything to make some of the uglies around longjmp
[21:21] <ion> More powerful than plain exceptions.
[21:21] <Keybuk> I spent a long time maintaining dpkg
[21:21] <Keybuk> nih_error was originally written for dpkg2 ;)
[21:21] <Keybuk> it doesn't behave the same way for a reason
[21:21] <sadmac> ion: and, ironically, easier to do in C
[21:21] <Keybuk> a *lot* of problems with dpkg were caused by errant error handlers
[21:22] <sadmac> Keybuk: I think it can be made cleaner than it has been.
[21:23] <sadmac> I'll play with it if I get to it. right now I want to get the parser standing up so I can finish the early triggers stuff and push some of it.
[21:26] <michas> hi, I am trying to debug some ubuntu init.d scripts by adding some echo lines. Unfortunately they print nowhere. :( If I understood everything right, upstart is the one who emulates old init and executes the scripts. - Is there a way to make that echo output visible? (Any hint?)
[21:26] <sadmac> michas: add a line to the jobs that says "console output"
[21:27] <michas> sadmac, well, it is an old init.d script. IIRC "console output" works only in the real upstart scripts.
[21:28] <michas> the first "log_action_begin_msg" prints, but any further echo or log doesn't print anymore. I assume it ist upstart filtering the output or something...
[21:29] <sadmac> michas: that's going to depend on how upstart is being used to run sysv scripts in ubuntu. you'll have to ask in an ubuntu related channel
[21:29] <ion> michas: If the script sources /lib/lsb/init-functions, try log_action_msg "foo" or something like that.
[21:30] <Keybuk> still won't work ;)
[21:30] <Keybuk> michas: write to a different file, e.g.  echo blah blah >> /dev/my.log
[21:30] <michas> ion, it does use init-function. the first log message works. the second does not. - that's why I suspect upstart.
[21:31] <Keybuk> nothing to do with upstart per se
[21:31] <Keybuk> Ubuntu has a "no messages on console during boot" policy
[21:31] <ion> michas: Please ignore what i said and do what Keybuk said. :-)
[21:31] <Keybuk> they get hidden with extreme prejudice
[21:31] <michas> Keybuk, ok, that will surely do. but I'm still wordering wtf is goin on there. ;)
[21:34] <Keybuk> michas: there are ways to make them appear
[21:34] <Keybuk> you can add "console output" to /etc/init/rc.conf and /etc/init/rc-sysinit.conf
[21:34] <Keybuk> and something like console=tty1 to your kernel command-lien
[21:34] <Keybuk> making sure not to run usplash
[21:34] <Keybuk> that brings most of them back
[21:35] <michas> ah, thanks, that looks good. I'll try that.
[21:35] <Keybuk> but if it doesn't work, try #ubuntu ;)
[21:35] <ion> :-D
[21:36] <michas> allright, I'll do. thanks.
[21:38] <michas> one last real upstart question: ist there a way to do something like "start on *" f.e. to monitor all events?
[21:39] <Keybuk> no
[21:39] <ion> Why would you want that?
[21:39] <Keybuk> such a thing would match itself, for a start
[21:39] <Keybuk> including its own stopped event ;)
[21:39] <Keybuk> if you instanced it, you'd end up with a fork bomb
[21:40] <michas> I just started to read about all the upstart stuff. And then I wondered what events will happen at a startup on my system.
[21:41] <sadmac> there used to be a way to dump them as they happened, but we removed it.
[21:42] <sadmac> kind of a dangerous thing. People do horrible things when you give them access to that sort of thing from shell scripts.
[21:42] <michas> :) ok, guess you are right. :)
[21:42] <Keybuk> you can still see them with --verbose
[21:43] <Keybuk> there's thousands in the startup of a typical ubuntu system
[21:43] <michas> where do I have to put that --verbose?
[21:43] <ion> keybuk: --verbose or --debug?
[21:43] <Keybuk> michas: kernel command-line
[21:43] <Keybuk> ion: I tend to recommend --verbose these days
[21:43] <ion> Alright
[21:43] <Keybuk> --debug adds things you generally don't want to see
[21:44] <sadmac> michas: you could also systemtap them out if you have a recent enough version and a little elbow grease to apply
[21:44]  * sadmac makes a note to add stap tracepoints to upstart for doing that sort of thing
[21:45] <michas> are those options documented anywhere?
[21:46] <Keybuk> michas: no
[21:46] <michas> ok :)
[21:46] <Keybuk> (unless you count "init --help" :p)
[21:47] <michas> oops. too easy.
[21:55] <michas> ok, thanks a lot. (That current upstart/rc mix in ubuntu is quite confusing, once things don't work as expected... Hope, they switch to upstart completely soon.)