[15:00] Keybuk: hi [15:00] hey [15:00] tested your ubuntu-boot ppa [15:00] noticed two issues: [15:00] With FSCKFIX=no, it fails to boot [15:01] and I get a warning during boot from mountall, that it couldn't write the pid file [15:01] as the fs was ro [15:02] err, as in you *set* FSCKFIX=no? [15:02] mountall also mounts stuff which is later attempted again by the apparmor sysv init script [15:03] Yes, I had set FSCKFIX=no [15:06] With FSCKFIX=yes it boots fine [15:07] I'm getting a warning though: init: mountall main process (927) terminated with status 1 [15:08] oh, that's not good ;) [15:10] mbiebl: Do you have multiple partitions in fstab, btw? Are they on the same disk? [15:10] mbiebl: Oh, actually, do you have multiple pass=2 partitions in fstab? [15:11] interestingly I can't always reproduce this warning message [15:12] The: "mountall: Unable to write pid file: Read-only file system" is reproducible [15:12] The other one not [15:12] keybuk: Does mountall handle the pass=1 / partition first and pass=2 ones later in parallel? I haven’t really looked at that part of the code. [15:12] ion: I tested inside virtualbox installation with a / and swap partition [15:13] no [15:13] Ok, no need to test my patches then. I probably should install virtualbox myself and test different partition configurations. [15:13] pass= is fairly nonsensical for mountall [15:13] Ok [15:14] Keybuk: another issue I observed is, that the upgrade of at hangs [15:15] The "start at" command in postinst does not complete [15:15] so the upgrade process is stuck at this stage [15:15] I manually had to kill the start command so the upgrade process could finish [15:17] I also remember problems with services that use restart in postinst instead of start/stop (like rsyslog), when the initial upgrade to the upstart job is made [15:18] I thought I'd fixed all those issues [15:18] it'll complain but not fail the postinst [15:19] lemme check again. I set a snapshot in virtualbox, so I can easily test an upgrade again. [16:55] Keybuk: looking at the bootchart of the default alpha5 install now [16:56] What is this weird couchdb doing. It spawns xulrunner and takes over 5 secs [16:59] Keybuk: I think FSCKFIX=no is the default [16:59] it is [17:00] mbiebl: couchdb wat? [17:00] sadmac2: no idea what this is doing [17:01] and why this is in a default install [17:01] mbiebl: I'd love to know [17:02] couchdb is a key-value data storage [17:02] I fear that somebody tought it would be a great backend for some desktop application... [17:02] Md: why does it need to run xulrunner for that? [17:02] I know [17:02] I saw that today [17:03] BIG SIGH [17:03] Md: is it just key-value? last I read up on it I thought it was a bit shinier. [17:04] Keybuk: damn, I'm hitting the superblock mount time in the future issue in my virtualbox install [17:04] I guess this is also where the new mountall fails [17:04] mbiebl: your drive has negative write latency? [17:05] hahaha [17:06] could this be a ext4 or util-linux bug? [17:06] did the new mountall fail properly to a shell [17:06] ? [17:06] it's a bug in the ext3 and ext4 filesystem code in the kernel [17:06] no clue about both issues, since it is written in erlang I stand clear from it [17:08] Keybuk: when is it triggered? [17:08] I don't seem to encounter it on every boot [17:09] Md: something particular against erlang, or just not a λ guy? [17:10] * sadmac2 still has haskell on his list of languages to find something to do with. [17:10] mbiebl: east of UTC, hardware clock in local time, system powered off without clean unmount [17:11] It's the clean unmount that bothers me, because I did a regular shutdown/reboot [17:11] sadmac2: I had ejabberd explode puke erlang core dumps on me too many times... [17:12] Md: shame. That's supposed to be one of the better jabber servers. [17:12] and indeed it is. you figure :-) [17:12] (I jave not tried the java ones, but I hate java with a passion) [17:13] Md: you're a good man [17:14] CouchDB seems awesome and Erlang is awesome. :-P [17:15] I'm very drawn to λ programming. I need something to do in ocaml/haskell/erlang [17:15] something that isn't such a wacky idea that I'm afraid of dooming it further by writing it in a language people are hostile to by policy === ion_ is now known as ion [18:12] mbiebl: hah, that FSCKFIX is never going to work, is it? [18:41] Keybuk: what do you mean, setting it to yes? [18:41] no, I mean that if you do FSCKFIX=no it'll still act as if it's yes [18:43] dunno if this is the case with the new mountall command [18:43] With the old initscripts, it made a difference [18:44] huh? :) [18:44] I mean the code in /etc/init/mountall.conf [18:44] [ -n "$MOUNTALL" ] && fsck_fix="--fsck-fix" [18:44] err [18:44] [ -n "$FSCKFIX" ] && fsck_fix="--fsck-fix" [18:44] is wrong ;) [18:44] I just changed it [18:45] [ "$FSCKFIX" = "yes" ] && fsck_fix="--fsck-fix" [18:45] ;) [18:45] ok [21:25] Keybuk: I tried to test an upstart build the other day. It didn't grab the dbus name (because I didn't run it as root and left out the upstart.xml file), but moreover I couldn't look at it at all with d-feet. System logs said something about not having permissions to call the introspection methods. Will putting the permissions file in place and running as root miraculously fix this? [21:40] Keybuk: found a few other issues: at vs atd [21:40] after the upgrade, there is /etc/init.d/atd and /etc/rcS.d/S89atd because the upstart job is called "at" [21:42] Directly after the upgrade, this is the list of initctl list: http://paste.debian.net/46172/ [21:44] As you can see, a few services are not listed as running, like rsyslog, udev or avahi-daemon [21:44] interestingly dbus is marked as start/running without a pid [21:57] There is also still a symlink /etc/rcS.d/S37udev-finish after the upgrade [22:01] on shutdown, I also get the following messages [22:02] init: rc main process (2383) stopped by STOP signal [22:02] init: rc main process (2383) continued by CONT signal [22:02] init: rc main process (2383) stopped by STOP signal [22:02] init: rc main process (2383) continued by CONT signal [22:02] ... === robbiew is now known as robbiew-afk