[12:27] <quitte> #define HAVE_SYS_INOTIFY_H 1 is in config.h. I'm out of ideas. maybe you can see a problem in the patch here: http://bugs.uclibc.org/view.php?id=917 ?
[12:47] <Keybuk> you need the functions in your uclibc as well as the headers
[12:55] <quitte> the kernel headers are in place,too. 
[01:00] <Keybuk> yeah, you still need the stub functions though
[01:00] <Keybuk> the header declares the existance of the function, but not the function
[01:01] <Keybuk> you can probably just copy the syscall stuff into a header and inline it
[01:01] <Keybuk> like the nih/inotify.h thing does
[01:02] <quitte> hmm. I hardly understood what you said. but I'll have a look at nih/inotify.h
[01:06] <quitte> ok looking at it didn't give me any clue.
[03:28] <|Lupin|> Hello, everybody
[03:30] <|Lupin|> I am joinning the chan because it looks like upstart does not log the bot messages in /var/lob/boot, I am wondering why...
[03:47] <Keybuk> |Lupin|: we had some last-minute issues with logd
[03:47] <Keybuk> it turns out that when it isn't running, processes that try to output to stdout die
[03:48] <Keybuk> ie. the shell running the shutdown scripts
[03:51] <|Lupin|> so ?
[03:52] <|Lupin|> Actually there is a file called /etc/event.d/logd
[03:52] <|Lupin|> it contains a line saying 
[03:52] <|Lupin|> console output
[03:53] <|Lupin|> Does one have to change this line to have hte output redirected to a file rather than to the console ?
[03:53] <|Lupin|> More in general, I can't find any documentation about the synta of the files under /etc/event.d
[03:55] <Keybuk> the console X line does indeed set where the input/output of a job goes; you set jobs to "console logged" to have them use logd, except this is disabled in the code
[03:55] <Keybuk> the lack of documentation is semi-deliberate, since the syntax isn't stable yet
[03:58] <|Lupin|> ok
[03:59] <|Lupin|> Keybuk: In fact what I would really like is to be able to have everything appearing on the screen at boot gathered in a file so that I can read it. Is this possible ?
[04:23] <|Lupin|> Or perhaps would it be more simple to switch back to sysvinit ?
[04:24] <Keybuk> there's various interesting ways to do that
[04:25] <Keybuk> the logd trick is to set the console to a set of sockets that are intercepted by a daemon
[04:25] <Keybuk> the disadvantage is obviously that without the daemon, the trick doesn't work
[04:25] <Keybuk> an alternate trick is to use TIOCCONS to capture "console output"
[04:25] <Keybuk> the disadvantage to that is that you don't know which service is doing the output
[04:27] <|Lupin|> Keybuk: oh indeed the daemon is not installed...
[04:27] <Keybuk> (the other disadvantage to the logd trick is that with ordinary sockets, you don't have a tty, which can adjust the behaviour)
[04:27] <|Lupin|> Keybuk: I think I don't mind loosing the name of the service, as a temporary solution. But how can this solution be implemented, concretely, please ?
[04:28] <Keybuk> there's not an easy answer available right now
[04:28] <|Lupin|> Hmm...
[04:28] <Keybuk> it turns out to be a more difficult problem
[04:28] <|Lupin|> If the system remains in text mode, will I be able to rean the console with shift-pagep / shift-page-down ?
[04:30] <Keybuk> yes
[04:31] <Keybuk> you're obviously relying on the kernel ring-buffer size there
[04:34] <|Lupin|> yes
[04:35] <|Lupin|> I am wondering how I can keep the system in text mode.. ?
[04:37] <|Lupin|> is it a matter of runlevel ?
[04:49] <|Lupin|> I believed so, but runlevels 2, 3, 4 and 5 seem to run gdm... 
[04:54] <Keybuk> remove gdm symlinks from one of the runlevels
[04:58] <|Lupin|> is it all I have to do ? okay.
[05:45] <|Lupin|> it worked, thantks
[06:21] <|Lupin|> bye every body
[06:21] <|Lupin|> thanks a lot for your assistance
[06:56] <quitte> hi. is it possible that my problems compiling upstart using uclibc are related to the kernel being 2.6.13? that's the release that inotify was introduced.
[06:56] <AlexExtreme> quitte, yes, you need a recent kernel
[06:56] <quitte> :( ok. so it'S impossible to use it on a fritzbox without loosing support for the proprietary modules
[06:56] <Keybuk> Upstart works best when you have the most event support from the kernel
[06:57] <Keybuk> since if you're interested in an event-based init daemon, you're kinda inherently interested in the events to base it off
[06:57] <Keybuk> the kernel's uevent interface didn't reach maturity until about 2.6.17
[06:57] <AlexExtreme> yeah, it's better to have the uevent support for udev which was introduced in .17 IIRC
[06:58] <quitte> ok i see. so i'd best stick with busybox init for now. maybe the vendor will update the kernel version someday, but i doubt it.