| === robbiew is now known as robbiew_ | |
| === juergbi_ is now known as juergbi | |
| === [2]sassyn is now known as sassyn | |
| === ev1 is now known as ev | |
| === ev1 is now known as ev | |
| === robbiew_ is now known as robbiew | |
| === [3]sassyn is now known as sassyn | |
| [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 | |
| === robbiew is now known as robbiew-afk | |
| === robbiew-afk is now known as robbiew_ | |
| [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.) | |