[13:33] <SpamapS> Hmm.. didn't somebody provide a workaround script for bug #406397 ?
[13:55] <ion> http://heh.fi/tmp/workaround-upstart-snafu
[14:08] <SpamapS> ion: ty
[15:36] <apw> jhunt, yo ... i have a mad situation where gdm seems to start when its pre-dependancies have not been run (or at least logged as such) ... any suggestions on how to diagnose
[15:37] <jhunt> apw: Adding, "set >> /tmp/gdm.log" in a script section might be useful. What scenario is it starting in where you would not expect it to?
[15:39] <apw> jhunt, it is dependant on startup and (graphics device or stopped fallback-job)
[15:39] <apw> jhunt, i have rm my drm modules and there are none of those events, and fallback-job also does not get logged at all
[15:40] <jhunt> apw: If you use the "set" as above, you can see what the value of UPSTART_EVENTS was (ie what triggered the start on condition to become true).
[15:40] <apw> so gdm should not have started ... ok will do
[15:41] <ion> exec >/tmp/gdm.log 2>&1
[15:41] <ion> set -x  # perhaps
[15:41] <SpamapS> apw: sometimes the started event is fired before a service is actually ready unfortunately
[15:43] <apw> SpamapS, hmm, but in this case the service never starts
[15:44] <apw> its never even mentioned in --verbose output
[15:45] <apw> jhunt, so what the heck does 'filesystem started stopped' in U_E tell me?
[15:45] <apw> start on (filesystem
[15:45] <apw>           and started dbus
[15:45] <apw>           and (drm-device-added card0 PRIMARY_DEVICE_FOR_DISPLAY=1
[15:45] <apw>                or stopped fallback-framebuffer))
[15:46] <jhunt> those were the events that triggered gdm to start: "filesystem + started (dbus) and stopped (fallback-framebuffer)"
[15:47] <jhunt> looks like the fallback-framebuffer is your problem there.
[15:47] <jhunt> did it ever start? You might need to use the technique I mentioned to you the other day.
[15:47] <apw> jhunt, but fallback-frambuffer doesn't even appear in the --verbose output
[15:47] <jhunt> right, it never started: hence it was stopped.
[15:48] <apw> jhunt, that should not be correct... these are events not states
[15:48] <apw> it would have to actually start and then stop to get a stopped event surely
[15:51] <apw> jhunt, is there soem synthetic event at upstart init
[15:51] <jhunt> apw: "startup" is what kicks the whole thing off
[15:52] <apw> jhunt, no i think i am miss understanding
[15:52] <apw> jhunt, how in an event based system is something which never run emitting a stopped event
[16:07] <jhunt> I'm just looking to see how it is doing the matching...
[16:22] <jhunt> apw: is there a "stop on" condition for this job?
[16:22] <apw> jhunt, for gdm or fallback-*
[16:23] <apw> jhunt, fallback-* not as it is a 'task'
[16:23] <jhunt> apw: gdm. is it just runlevel?
[16:24] <apw> stop on runlevel [016]
[16:24] <apw> yep
[16:32] <jhunt> apw: can you send me your *.conf files, along with your syslog + gdm.log?
[16:34] <apw> jhunt, can do
[16:40] <apw> jhunt, location in privmsg
[16:45] <apw> jhunt, its pretty big :)
[16:46] <apw> jhunt, this 'stopped udevtrigger' form is the one i followed here ... and it makes little sense if that can occur before it starts too
[17:51] <apw> jhunt, that _may_ have been a 'few' off, very confusing