/srv/irclogs.ubuntu.com/2010/03/26/#upstart.txt

Keybukmbiebl: no, hadn't heard anything about that02:25
Keybukcool ;-)02:25
=== pocek_ is now known as pocek
=== pocek_ is now known as pocek
nanonymeAny devs or other people who'd like to spar a few ideas about upstart present?13:39
sadmac2nanonyme: what do you mean?13:39
nanonymeWell, I bumped into the apparently common issue of it being a bit tricky to start programs inside a chroot when dealing with upstart.13:41
nanonymeI was wondering if there is some working solution and if not, whether dbus could be used so that upstart outside the chroot could communicate with processes inside the chroot and supervise them.13:42
nanonymeOpinions or ideas?13:44
Keybukthere isn't a working solution13:44
Keybukjust some fuzzy ideas13:44
Keybukit's not really that high on the urgency list13:45
nanonymeYeah. But would you think something like that could work? That is, that since upstart can use dbus, you'd hook up the directories dbus uses between real system and chroot.13:46
nanonyme(or am I missing something crucial here?)13:47
KeybukI don't see how D-Bus helps here13:48
nanonymePoint. I didn't really think it far enough through.13:54
Keybukthe difficulty mostly comes from the chroot separation13:58
Keybukmaking sure that "start apache" in the chroot starts the apache in the chroot, not in the outside system13:59
nanonymeHrm.14:01
sadmac2_udev's syntax makes me pee a little.14:48
=== sadmac2_ is now known as sadmac
sadmacKeybuk: here's one for you: writing a "hey, look how cool libnih is!" blog post. Wanted a practical but simple code snippet that shows why multiple parents is useful.15:13
Keybukerr, there must be one somewhere in Upstart's source ;-)15:15
sadmacKeybuk: practical, lots of them. The simple is key :)15:15
sadmaci.e. an example that makes sense in 20 lines without having to see the whole rest of upstart's code15:16
Keybukerr, you could use a message?15:17
Keybukwhere a struct has a string in it15:17
Keybuk(string parent is struct)15:18
Keybukand you borrow the string, referencing it, using it in another struct15:18
sadmacwas thinking along those lines15:18
sadmachttp://screwyouenterpriseedition.blogspot.com/2010/03/why-you-should-be-using-libnih.html17:57
sadmacKeybuk: are you against patches to make some of the ignored compatibility options in /sbin/shutdown do what they used to?18:32
Keybukyes18:32
Keybukfor the most part18:33
KeybukIMO shutdown should only accept options that change the way the machine is shut down18:33
sadmacnotting: we still have bugs for making -f, -F, and -n18:33
Keybukall of the other silly hangers-on are better off moved to other tools / GNOME UI / etc.18:33
sadmacnotting: kill these things?18:33
Keybuk-f, -F, etc. wildly assume your filesystem check implementation18:34
Keybuke.g. for us, forcing a fsck is adding a kernel command-line argument, not touching a file18:34
sadmacwe can do it either way18:34
Keybukright, but I don't want to force that kind of thing18:35
Keybukyou may not even fsck on boot18:35
Keybukmaybe it's an embedded system with a read-only filesystem anyway18:36
Keybuketc.18:36
Keybukthat's why I don't want any of those flags in shutdown18:36
sadmacI'd prefer not to carry patches. I think I'll WONTFIX the fedora bugs. If plautrba wants/is forced to do something about it in RHEL, so be it.18:37
nottingKeybuk: well, if it's an embedded system, you get what you deserve if you tell it to fsck on the next reboot18:43
sadmacnotting: are you up for an initscripts-based fix for 464456?18:45
sadmacsee comment 318:45
nottingeh, i'm not sure the current behavior is really wrong18:46
Keybuknotting: put another way18:46
Keybukif shutdown should have a "and check filesystems on next reboot" option18:47
Keybukshouldn't shutdown also have a "reboot into windows" option? :p18:47
nottingoh gah, lilo -R18:47
sadmacnotting: I can close 464456, or give it to you to close. Where should we put it?18:48
nottingi'll close it18:48
sadmacnotting: https://bugzilla.redhat.com/show_bug.cgi?id=441048 , terminal isn't in cooked mode when it should be?18:55
nottingyeah. it goes away, it comes back. usually we blame halfline18:56
sadmacnotting: should I move it to plymouth? Or to initscripts (I'm guessing some sort of echo ^[!@#$%!@#$^! at the start of the single user job could fix it)18:58
nottingplymouth18:58
sadmacKeybuk: do you have a launchpad bug about upstart not working on XFS?19:03
nanonymeHrm, ^M is supposed to be enter too though, me thinks.19:12
halfline^M is carriage return.  tty settings determine whether ^M or ^J or some combination there of is enter19:14
nanonymeRight. Both ^M and ^J are considered enter here though.19:15
nanonymeI've always hated the amount of variation those things have. :p Including backspace.19:16
CaesarKeybuk: you around?21:50

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!