Rebuilding kernel/system to a state "back-in-time"?

[Erik Cederstrand]

Jonathan Horne wrote:
> ...
> IMO, (and forgive me, i generally dont spew my opinions where they arent 
> welcome or asked for), RELENG_6_2 is better for a server over RELENG_6 
> (aka, -STABLE), as it doesnt include items that are not critically 
> required for secure and stable operation.  remember, that the true 
> -STABLE branch has items merged in from -CURRENT (call it back-ported?).
> 
> let say, you already know that -p8 is the latest 6.2 revision.  you get 
> on a server, you log in, and it says 6.2-RELEASE-p8.  you already know 
> that this system is up to date.  if you log in, and see 6.2-STABLE... 
> you dont immediately know when this system was last rebuilt without 
> doing some other version checks first.  i have to be honest, when it 
> comes to managing a farm full of servers, i like my "visual version 
> checks"... the same way i like my women:

We're going off-topic now, but you have a point. I'm not going to argue 
if STABLE is better than release branches on servers, but I think it 
would be useful to record the CVS date somewhere by default (I know you 
can do this manually via src/sys/conf/newvers.sh). Sometimes the "p8", 
"prerelease #4" or even kern.osreldate is too low resolution. uname -a 
just exposes the build date of the kernel, not the date of the sources. 
Maybe a sysctl like:

sysctl kern.oscvsdate: 20071105224900

Erik