| darius12 | does process handover work with current upstart bzr? (from an initramfs upstart to the "real" upstart) | 10:05 |
|---|---|---|
| Keybuk | no | 10:20 |
| darius12 | what is missing? the ipc protocol between the two instances of upstart? | 10:36 |
| Keybuk | right | 10:40 |
| darius12 | is it just that, or there needs to be any refactoring before this protocol is added? | 10:42 |
| Keybuk | the complete lack of any IPC is a major todo item right now | 10:43 |
| Keybuk | the necessary conf changes have already happened | 10:43 |
| darius12 | I see. | 10:45 |
| Keybuk | a related, but not critical path, item is the atomicity of start & stop | 10:50 |
| darius12 | they are not yet atomic? | 10:51 |
| darius12 | I 'm setting a qemu image to play with boot-time facilities (upstart and grub2) | 10:53 |
| Keybuk | no | 10:54 |
| Keybuk | in the sense that: | 10:54 |
| Keybuk | start JOB with environment FOO=x BAR=y | 10:54 |
| Keybuk | this should start the job, storing those environment variables in it, and those environment variables should persist throughout the life of the job, including its stop scripts | 10:54 |
| Keybuk | if I now do | 10:54 |
| Keybuk | stop JOB | 10:54 |
| Keybuk | start JOB with environment FOO=y BAR=x | 10:55 |
| Keybuk | (immediately after each other - ie. restart) | 10:55 |
| Keybuk | the job should stop with FOO=x and BAR=y and not adjust its environment until it has completely stopped | 10:55 |
| darius12 | I see, interesting | 10:56 |
| Keybuk | actually, it's turned out to be quite a headache | 11:16 |
| darius12 | is "with environment" really needed? | 11:24 |
| darius12 | (I guess it has to do with supporting different instances of a service right?) | 11:24 |
| Keybuk | right | 11:33 |
| Keybuk | and if you don't have environment, you don't know what event started you | 11:33 |
| darius12 | right | 11:34 |
| darius12 | and I guess I need service instances | 11:37 |
| Keybuk | instance cases are even more fun | 11:39 |
| Keybuk | since in theory, the environment attached is similar enough anyway | 11:39 |
| Keybuk | start getty TTY=tty1 | 11:39 |
| Keybuk | stop getty TTY=tty1 | 11:39 |
| Keybuk | start getty TTY=tty1 | 11:39 |
| Keybuk | doesn't really make much difference :) | 11:39 |
| darius12 | and why is it such a problem for the atomicity? (I haven't looked at the source yet) | 11:44 |
| darius12 | you are storing an event together with its environment, not? | 11:45 |
| darius12 | I meant a job instance together with its environment | 11:51 |
| Keybuk | right | 12:11 |
| Keybuk | you'd need to store its current environment | 12:11 |
| Keybuk | and its next environment | 12:11 |
| Keybuk | or... | 12:12 |
| Keybuk | spawn a new instance, and don't let it start until the other stops | 12:12 |
| darius12 | right, I see | 12:20 |
| Keybuk | I'm not sure which is the right solution | 12:57 |
| === kylem_ is now known as kylem | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!