| 1 |
On Sun, 26 May 2013 11:45:38 +0200 |
| 2 |
Tom Wijsman <TomWij@g.o> wrote: |
| 3 |
|
| 4 |
> On Sun, 26 May 2013 11:20:25 +0200 |
| 5 |
> Michał Górny <mgorny@g.o> wrote: |
| 6 |
> |
| 7 |
> > It is *easy*. |
| 8 |
> > |
| 9 |
> > ln -s /sbin/newinit /sbin/init.new |
| 10 |
> > mv /sbin/init.new /sbin/init |
| 11 |
> > |
| 12 |
> > Easy and atomic. The inconsistency potential is similar to one given |
| 13 |
> > by init upgrades. Yet we don't do anything magical to defer init |
| 14 |
> > upgrade until reboot, and that's why the upgrades go smoothly. |
| 15 |
> |
| 16 |
> Easy isn't always good. It's not atomic since you can't reboot and |
| 17 |
> because of that I wouldn't call it smooth either. |
| 18 |
|
| 19 |
Can't you? How come? |
| 20 |
|
| 21 |
> > > I think that safest way not using wrapper is to stop all services |
| 22 |
> > > and keep only mounted /, than change things (symlinks,configuration |
| 23 |
> > > update) and reboot. |
| 24 |
> > |
| 25 |
> > This can be done two ways. |
| 26 |
> > |
| 27 |
> > One is hacking into init (RC) reboot procedure. It's fragile, it needs |
| 28 |
> > to be fit into the right moment and I'm not sure if all inits will |
| 29 |
> > handle this without actually needing to patch the code. And in the |
| 30 |
> > end, the init gets replaced before init stops working anyway. |
| 31 |
> |
| 32 |
> You're making things way more complex than a wrapper would do. I'm not |
| 33 |
> a fan of using the words "hacking", "fragile" and "not sure" for |
| 34 |
> selling an approach; so, why were you suggesting the symlink approach? |
| 35 |
|
| 36 |
Don't mix the two mails. I am showing how fragile the solution |
| 37 |
suggested by Robert is. While you seem to be replying to every mail |
| 38 |
possible to repeatedly advocate your idea. |
| 39 |
|
| 40 |
-- |
| 41 |
Best regards, |
| 42 |
Michał Górny |
Archives