| Keybuk | KDesk: hi | 09:51 |
|---|---|---|
| keesj | Morning | 09:53 |
| keesj | today is d-day for me and my upstart state machine... | 09:55 |
| keesj | my goals is to get events queued so I can emit multiple events to a task scripts and be able to handle then all | 09:57 |
| keesj | something like http://www.paste-it.net/public/tf5a413/ | 09:58 |
| keesj | but I guess "start" should be renamed to "queued start" | 09:59 |
| * Keybuk still doesn't know what you're doing | 10:02 | |
| keesj | I have one upstart "job" that acts like a state machine its' called change_handler and listens to many external event (emit usb_detected stuff like that) , that state machine is used to support swiching between different "runlevels" (boot|mass_storage|main_app) | 10:07 |
| Keybuk | the built-in one doesn't work for you? | 10:07 |
| keesj | next to that I have emtpy state_boot state_mass_storage,,, jobs and services that depend on those state jobs | 10:07 |
| keesj | and all the change_handler does is determin the currently running state and the event and stop one state and start a different state | 10:08 |
| keesj | built-in state machine? | 10:09 |
| Keybuk | well, pretty much the point of upstart is that it does that kind of thing internally | 10:13 |
| keesj | well I do use it a lot for simple service depenencies but that is not enough | 10:15 |
| keesj | (trying to explain why..) | 10:17 |
| keesj | well it's also what the sysv replacement does | 10:20 |
| keesj | would it help if I send a mail to the mailing list explaining the problems I have? | 10:24 |
| Keybuk | sure | 10:24 |
| keesj | I really boils down to the http://upstart.ubuntu.com/getting-started.html on bounce example not wokring | 10:25 |
| Keybuk | it doesn't? | 10:26 |
| keesj | no | 10:27 |
| Keybuk | I don't see a bug about that anywhere | 10:28 |
| keesj | well it's not like there is a definitive list of the syntax of the files so I would know what to expect to be honest | 10:29 |
| Keybuk | http://upstart.ubuntu.com/wiki/Stanzas | 10:30 |
| keesj | This page corresponds to upstart 0.3.9. The syntax and job control mechanisms is greatly improved in 0.5.0. | 10:30 |
| Keybuk | ...if only it were a wiki ;) | 10:31 |
| keesj | I did edit some parts of that | 10:32 |
| keesj | on that page the console also had non existent options | 10:33 |
| Keybuk | seriously though, if something isn't working, then a bug should be filed | 10:35 |
| keesj | ok suppose the on bounce would work and would take 5 seconds to run and I emit 5 bounce withing one second , what should the behaviour be? | 10:37 |
| Keybuk | depends on what you do in bounce | 10:37 |
| Keybuk | you either want just one running at a time | 10:37 |
| Keybuk | so use the default behaviour | 10:37 |
| Keybuk | you want one running for each bounce | 10:37 |
| Keybuk | so add "instance ..." to your job | 10:37 |
| Keybuk | or you want one running at a time, but if bounce occurs while it's running, it needs to be run again (but only once no matter how many bounces occur) | 10:38 |
| Keybuk | so instance and exec watershed | 10:38 |
| keesj | if I would like a single instance to be run 5 times would that be possibe? | 10:39 |
| Keybuk | sure, in 0.3.x just add "instance" to the job | 10:39 |
| keesj | I am on 0.5 | 10:41 |
| keesj | 0.5.1 even | 10:41 |
| Keybuk | that's a bit more tricky, you need to add something to the bounce to instance on | 10:42 |
| Keybuk | which is a bug in 0.5 | 10:42 |
| Keybuk | well, more of a misdesign | 10:42 |
| keesj | hm. can you explain a little more? | 10:46 |
| Keybuk | in 0.5, you have to have something to define the limits of the instance | 10:46 |
| Keybuk | it doesn't let them be unlimited | 10:47 |
| Keybuk | so in the bounce example, you'd have to include (for example) a BOUNCENUM variable that changed for each one | 10:47 |
| Keybuk | then the job could have instance $BOUNCENUM in it | 10:47 |
| keesj | but woudl that run the jobs in parallel? | 10:48 |
| Keybuk | yes | 10:49 |
| keesj | what I reall woudl like is that this task gets executed 5 time | 10:51 |
| Keybuk | that works in 0.3 and will work again by 0.10 | 10:56 |
| keesj | writing the mail.. | 10:56 |
| Jc2k | wg 8 | 10:59 |
| Keybuk | bah, that one's less fun than "win" | 11:00 |
| keesj | I guess what needs doing for the Stanzas is to Create 2 pages one for 3.8 and one for 0.5 | 11:21 |
| keesj | https://lists.ubuntu.com/archives/upstart-devel/2007-March/000241.html tels that "on EVENT" was removed | 12:29 |
| Keybuk | yes, in all versions of upstart since 0.2, "on" has been identical to "start on" | 12:29 |
| keesj | I am trying to implement some hack/proof of concept for the handler. From my current understanding what needs to happen is that I need to block events to a jobs instance while it is running as handler | 12:52 |
| keesj | so increasing the blocker on start and decreasing it on stop when that job is defined as ... "handler" | 12:53 |
| Keybuk | I'm not sure what you mean by "block events" | 13:01 |
| Keybuk | jobs don't "receive" events | 13:04 |
| Keybuk | but they can have their state changed by them | 13:04 |
| keesj | what I want to happen (I guess) is that in event_poll (in event.c) the event gets skipped because there are "bockers" | 13:04 |
| keesj | blockers | 13:04 |
| Keybuk | then it would be missed | 13:05 |
| keesj | and would be removed from the queue? | 13:05 |
| Keybuk | right | 13:05 |
| Keybuk | events can be held in the queue while things are processing them | 13:05 |
| Keybuk | but that's after job states have been changed | 13:06 |
| keesj | do you understand the difference between "my" state machine and the upstart states?(I think that mainly the logic to perform a state change is put in a script) | 13:11 |
| Keybuk | no | 13:21 |
| Keybuk | I shall read the mails later ;) | 14:32 |
| === mbiebl_ is now known as mbiebl | ||
| sadmac2 | Keybuk: around? | 20:03 |
| Keybuk | briefly | 20:03 |
| Keybuk | what's up? | 20:03 |
| sadmac2 | Keybuk: was anyone working on the dbus properties code? | 20:04 |
| Keybuk | me | 20:04 |
| Keybuk | right now | 20:04 |
| Keybuk | that depends on variant code | 20:04 |
| Keybuk | which is what I'm trying to figure out | 20:04 |
| sadmac2 | Keybuk: ok. I'll leave that to you then | 20:04 |
| sadmac2 | Keybuk: variant code? | 20:04 |
| Keybuk | properties use variants | 20:05 |
| Keybuk | right, bbl | 20:05 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!