CVSUP and 5.2.1 RELEASE

[Jason Dusek]

Well, my system is fine actually. I am running GNOME and Mozilla on it right now.

I think the only *good* solution to hosing one's system is to insist on having 
two bootable partitions - an 'old' one and a 'test' one. They would have 
everything but /usr and /home in them. Then have a 'make 
everything-and-i-mean-everything' target that mounts the test partition as 
/test, and puts the new system - kernel and all - into it. When you reboot, use 
the test kernel and system. If they work, then run 'make update', which causes 
the test system to mount the other partition and simpy copy itself over. Then 
you wouldn't find yourself in the situation of not knowing whether your kernel 
works till it doesn't.  And you could run a sort of 'make 
everything-and-i-mean-everything' tinderbox in the /test partition. If CURRENT 
was really sucking one day, you just try again next week.

My two cents,

- Jason


Barney Wolff wrote:
> On Tue, Jul 13, 2004 at 08:28:32PM +0200, Poul-Henning Kamp wrote:
> 
>>In message <20040713182351.GA72492 at pit.databus.com>, Barney Wolff writes:
>>
>>>On Tue, Jul 13, 2004 at 01:05:45PM -0500, Jason Dusek wrote:
>>>
>>>>I ran make world this morning. I ran make kernel as well, but the kernel is 
>>>>broken, so I kept my old kernel. Does this mean that I have a RELEASE 
>>>>kernel but a CURRENT world? Am I headed for trouble?
>>>
>>>To core:
>>>How many users do we have to sabotage with "make world" before it gets
>>>removed from the make targets?  Is it really that hard in the very rare
>>>case when "make buildworld && make installworld" is what's wanted to
>>>type exactly that?
>>
>>And your argument here is that people are reciprocally less likely
>>to hose (or as it may be: not hose) their systems because the have
>>to type 27 characters more to do so ?
> 
> 
> Yes.  That's why there are safeties on guns.  Or, since I'm using US
> metaphors, "make world" could be considered an attractive nuisance,
> like an unfenced swimming pool.
>