| 1 |
Neil Bothwick <neil@××××××××××.uk> wrote: |
| 2 |
|
| 3 |
> On Wed, 02 Apr 2008 19:40:37 +0200, Michael Schmarck wrote: |
| 4 |
> |
| 5 |
>> > Neil even proposed ALT + |
| 6 |
>> > SysRq + EISUB, to be sure everything is killed, sync'd and |
| 7 |
>> > unmounted. |
| 8 |
>> |
| 9 |
>> Which might or might not work. But note that I was also talking |
| 10 |
>> about applications being in a corrupted state (the database example). |
| 11 |
> |
| 12 |
> E sends a SIGTERM to all applications. Any well behaved application |
| 13 |
> should shut down cleanly on this. |
| 14 |
|
| 15 |
No doubt :) But if the app hangs, it might not respond to TERM. |
| 16 |
|
| 17 |
> I sends a SIGKILL, but it only affects |
| 18 |
> programs that were so locked up they ignored E, so you have nothing to |
| 19 |
> lose by then. |
| 20 |
|
| 21 |
Correct. |
| 22 |
|
| 23 |
But nonetheless, there's still the risk that the KILL has |
| 24 |
destroyed the application database (sort of - more correctly: |
| 25 |
that the application and its database was in a "non consistent" |
| 26 |
state when it received the signal). |
| 27 |
|
| 28 |
Michael |
| 29 |
|
| 30 |
-- |
| 31 |
gentoo-user@l.g.o mailing list |
Archives