[00:16] how are you reproducing? [00:17] unfortunately, I don't have a core file [00:18] reproducing is easy [00:18] open /etc/init.d/rc [00:18] and add "runlevel" after line 72 [00:19] it also seems, I was on the wrong track with startpar [00:19] Debian or Ubuntu ? [00:20] Ubuntu [00:20] after "export runlevel previous"? [00:20] yeah [00:21] who is responsible for setting the env variables PREVLEVEL and RUNLEVEL in rc.conf? [00:21] nobody [00:21] telinit sets them in the runlevel event [00:22] does telinit read /var/run/utmp to get that information? [00:22] yes [00:22] and its environment [00:23] I changed rc.conf a little [00:24] script [00:24] echo "RUNLEVEL: $RUNLEVEL" [00:24] echo "PREVLEVEL: $PREVLEVEL" [00:24] exec /etc/init.d/rc $RUNLEVEL [00:24] end script [00:24] hmm, runlevel just works for me there [00:24] sometimes PREVLEVEL is just empty on reboot [00:24] can't you do [00:24] ulimit -c unlimited [00:24] sometimes it's 2 [00:24] runlevel [00:24] ? [00:24] looks like a race somewhere [00:25] ulimit -c unlimited in /etc/init.d/rc ? [00:25] before calling runlevel [00:25] to get a coredump [00:26] oh, I know [00:27] it's that env RUNLEVEL= [00:27] should probably be =N ? :) [00:30] the core files should be in /? [00:30] aye [00:30] hm, there isn't any [00:31] even with ulimit [00:31] bah [00:31] if you change it to =N does it work? [00:32] rc-sysinit.conf? [00:32] yeah [00:34] or try 1165 [00:36] more in 1166 [00:42] Ok, if I set RUNLEVEL=N [00:42] I no longer get any asserts [00:42] runlevel then reports: N N [00:43] after boot? [00:43] immediately after the startup event [00:43] or for rcS? [00:43] ok [00:43] Not sure, if the legacy runlevel reported just "N" in that case [00:44] anyway, I guess runlevel should be more fault tolerant, when RUNLEVEL is not set [00:44] regarding the reboot problem [00:45] What I figured out so far is this: [00:45] I think I fixed that now [00:45] If I just hit strg+alt+del, I get 2 6 [00:45] When I login and type reboot, I get N 6 [00:50] that implies your /var/run/utmp is mucked up? [00:51] reboot seems to break it, [00:52] hah [00:53] by writing a shutdown line [00:55] and now I get the assert \o/ [00:57] try 1168 [00:57] so this should only happen if you *don't* use tmpfs for /var/run? [00:57] no, this was just a bug [00:58] reboot writes the shutdown time, which is a special runlevel record [00:58] that overwrites the runlevel record in utmp [00:58] I mean the assert in runlevel on boot? [00:58] but it did that before calling /sbin/shutdown, which meant it couldn't get the runlevel ;) [00:59] that was just bad sanity checking [01:09] Keybuk: looks good now [01:09] no more asserts, no more problems on reboot \o/ [01:10] btw., I asked on #vim if there is a standard way of installing vim addons [01:11] apparently there isn't a straighforward procedure [01:11] include it in the vim source ;) [01:11] So best is probably, to just ship a README how to install it manually [01:11] oh yeah, when the upstart syntax is frozen [01:11] then this is probably the best idea [01:11] I mean in Debian/Ubuntu === Keybuk changed the topic of #upstart to: PLEASE TEST TRUNK! | Upstart 0.5.3 "Britain's Flag Carrier" | Upstart 0.3.11 "For Friday, June 19th 2009, I'm Jon Masters" | http://upstart.ubuntu.com/ [01:12] anyway bed [01:12] Keybuk: ah [01:12] I shall release in the morning if there's no show-stoppers [01:12] Keybuk: caught you [01:12] :) [01:12] Keybuk: just filed the state transfer patch [01:12] sadmac: yup, quickly! [01:12] Keybuk: look at it in the morning if you like :) [01:12] cool [01:12] will do [02:04] Keybuk: make check no longer passes :-/ [02:11] fyi, I started packaging trunk: http://debs.michaelbiebl.de/upstart/ [02:11] in case you are interested [09:07] mbiebl: which revno did make check fail on? [09:08] keybuk: Does trunk handle the Ubuntu jobs as-is – that is, can i just install to /sbin/init.temp and test with init=/sbin/init.temp? [09:08] ion: no, they're in different paths and slightly different format [09:08] I decided it was better to make it easy for 0.10 to be backwards compatible with 0.6 [09:09] Good :-) [09:17] % autoreconf -i [09:17] autopoint: *** cvs program not found [09:17] le sigh [09:21] sadmac: fyi, i've just finished rpm from todays trunk - http://plautrba.fedorapeople.org/upstart/ [09:47] An Upstart build non-benchmark on a dualcore laptop: http://pastebin.com/mbb57e6b [09:53] keybuk: How about tagging the job files with a format version number? Enforce the file to begin with “upstart 0.6\n” and parse the rest as usual perhaps. It would also make file(1)’s job easier, as well as editor syntax highlighter defintions’. [09:59] I thought about it [09:59] but I don't really like those kinds of things [10:08] http://heh.fi/tmp/vim-highlight-upstart-rcS.conf http://github.com/ion1/vim-highlight [10:09] ion: is that using mbiebl's stuff? [10:10] I just copied contrib/vim/* to ~/.vim and ran vim-highlight -t upstart conf/rcS.conf foo.html. [10:10] Yeah, contrib/vim seems to be written by him. :-) [10:11] -t upstart wouldn’t have been needed, had it been able to automatically recognize the file type. ;-) [10:29] virtualbox-ose-source doesn’t seem to work with 2.6.31-2-generic-pae. Let’s not test Upstart in a VirtualBox VM then. Let’s see how ubuntu-vm-builder works. [12:22] wow [12:22] the test suite is much faster when compiled -O2 ;) [12:25] make check 55.36s user 4.29s system 63% cpu 1:34.17 total [12:42] ion: wanna help with polishing the vim syntax file? [12:45] Keybuk: is the message_iter_abandon_... patch for dbus already in Ubuntu/karmic? [12:45] yes [12:46] was it accepted upstream? [12:46] not yet [12:48] I need to write a test case for it [12:58] Keybuk: what about a small configure check and a ifdef? [12:58] mbiebl: #if YES_I_WANT_TO_ASSERT_ON_ENOMEM ? :) [12:59] something like that, yes :-) [13:00] only as long as the patch is not in the official dbus upstream [13:01] In which can you can bump the pkg-config check [13:01] I think it's better to subversively force everyone to apply that patch [13:01] then ./configure should simply abort [13:02] which would be trivial to do [13:02] together with a helpful error message, pointing to the fd.o bug report [13:04] want a patch :-) [13:07] actually, that is a good point [13:07] configure should check for that function and abort [13:07] yes please [13:08] ok, will do [13:51] mbiebl: which revno did make check fail on? [13:51] 1172 [13:51] http://paste.debian.net/41435/ [13:59] hhe [14:00] isn't that because upstart is already running? [14:08] Keybuk: my "make check" fail - http://pastebin.dqd.cz/939 [14:08] but i have to leave now, bb at monday [14:09] plautrba: do yoiu not have -lrt in ../nih/libnih.la ? [14:12] no [14:14] that's weird [14:14] don't support you mind tarring up your build directory for me so I can examine it? [14:15] Keybuk: could be [14:15] when make distcheck passed yesterday [14:15] I was still running 0.3.11 [14:16] Keybuk: http://plautrba.fedorapeople.org/upstart/upstart-0.6.0-1172-build.tar.bz2 [14:16] Keybuk: it's build from http://plautrba.fedorapeople.org/upstart/upstart-0.6.0-0.1172.fc11.src.rpm [14:21] ok [14:22] plautrba: what revno of libnih? [14:22] ah, 963 [14:22] -r968 fixed that [14:23] make check 61.55s user 4.41s system 65% cpu 1:40.97 total [14:23] (with -Os) [14:23] that means -O2 was slightly quicker than -Os [15:53] mbiebl: If i ever get anything done, i’ll post a patch. :-) [16:02] argh, manpage-de ships init.8 [16:03] so I alway have to run LANG=C to get the correct man page [16:05] mbiebl: did you have that configure.ac patch to hand? [16:10] yeah, could you quickly point me to the fd.o bug again? [16:11] bugzilla so sucks [16:16] found (via google) ;-) [16:23] http://paste.debian.net/41446/ [16:24] wait a sec [16:26] We don't want -ldbus-1 added to LIBS [16:27] this one's better http://paste.debian.net/41447/ [16:30] I came up with a better way ;) [16:30] I bumped the D-Bus GIT version to 1.2.15 and just dep on that [16:30] which gives us the timeout patches too [16:37] Keybuk: but that makes it harder for distros which ship dbus+the patch [16:38] mbiebl: D-Bus 1.2.16 will be out in NOW+$SOMETIME [16:39] Is Collin on it? [16:39] yup [16:39] cool [16:39] ok, i'll try to do one by tomorrow [16:40] even better ;-) [17:13] build test, check [17:13] distcheck, check [17:13] package built, check [17:13] * Keybuk reboots [17:13] I wouldn’t mind testing the package. [17:14] ...as long as your system doesn’t fail to boot with it. :-P [17:15] I'm just about to upload it :D [17:15] Ah, cool. [17:16] sweet [17:16] it booted [17:17] That’s always a nice bonus. [17:18] Keybuk: so what is the make-check-fails-when-upstart-already-running about? [17:19] mbiebl: not sure, it doesn't do it for me [17:19] I figured I'll fix that in trunk quickly after the release when I figure it out :) === Keybuk changed the topic of #upstart to: Upstart 0.6.0 "How appropriate, you fight like a cow" | Upstart 0.5.3 "Britain's Flag Carrier" | Upstart 0.3.11 "For Friday, June 19th 2009, I'm Jon Masters" | http://upstart.ubuntu.com/ [17:24] mbiebl: hmm [17:24] you had a previously failing test case right? [17:24] try pkill -f test_ [17:25] the test suite doesn't clean up in case of failure (since I want to be able to debug it usually) [17:26] I don't think I have any running test_ processes [17:28] kooky [17:28] I shall investigate and fix [17:43] Keybuk: I applaude your release codename [17:44] Indeed [17:51] keybuk: How much did you say the test suite takes to run? I built the Ubuntu package for 0.6.0 locally, and it didn’t take very long. It did seem to run the test suite. Unfortunately, the build log doesn’t seem to contain any timestamps. [17:52] Wait. What am i talking about? I’ve had the runtime of the previous command in zsh prompt for a while now. debuild -b -j3 of 0.6.0-1 took 491.5 seconds. [17:53] ion: where's your underscore? [17:53] A bit more than 8 minutes, including the time it took to type my PGP password when debuild signed the package. :-P [17:53] sadmac2: Huh? [17:54] ion: you've been ion_ forever [17:55] Ah. #elsewhere: 2009-06-24 00:25:50 < ion> Woot! I got the nick back. I registered it back in the 1800s and someone immediately complained, claiming he had been using the nick in Freenode before me. I was teh benevolent and gave it to him (i guess i got some good karma, since someone kindly gave the nick back to me in IRCnet, since i had been using it there long before him). It seems he hasn’t been at Freenode for six months and the registration expired. [17:55] pimp! [17:59] ii upstart 0.6.0-1 event-based init daemon [18:00] Let’s reboot and see whether i come back. [18:05] It’s alive! [18:09] keybuk: So, is something wrong with my system or your system (re: test suite runtime)? :-P [18:57] Keybuk: when will the new daemon monitoring code land in trunk? [19:11] notting: how do we feel about 0.6.0 in RHEL? are we going to try to swing it? [19:11] plautrba: ^^ [19:12] gah. [19:13] gah? [19:22] "i would prefer not to deal with the headaches of a rushed migration at this particular exact point in time" [19:23] notting: never heard that acronym... [19:23] not acronym. approximation of an unintelligible verbal noise [19:23] notting: its very meaning-dense [19:23] (see also: 'meh', 'bleah', etc.) [19:24] * sadmac2 should probably port rawhide this weekend [19:29] I'll be at the beach, so nothing better to do :) [19:56] Keybuk: is there an equivalent to the old "initctl events" in 0.6.0? [19:57] mbiebl: last we discussed it it was removed by design choice === michael__ is now known as mbiebl [22:32] Keybuk: so you merged the debian packages again?