Updating multiple computers

[Brandon J. Wandersee]

Mario Lobo writes:

> On Mon, 19 Sep 2016 19:16:33 -0400
> Baho Utot <baho-utot at columbus.rr.com> wrote:
>
>> On 09/19/16 09:43, Brandon J. Wandersee wrote:
>> > Baho Utot writes:
>> >  
>> >> How do I use that release assuming that it is on a network server
>> >> to update/upgrade multiple systems?  
>> > Release images are for clean installs. To update machines over the
>> > network, just run `make buildworld` and `make buildkernel` on the
>> > master system, then mount /usr/src and /usr/obj to the other
>> > systems using NFS. 
>> Ok I will try that
>> 
>> Thanks
>> 
>
> I have one question regarding this alternative.
>
> I already have this scenario set up.
>
> Suppose that I want to save compile time on the remote machines
> and that machine #1 that holds /usr/src and the compiled /usr/obj is
> an intel XEON.
>
> Remote machine #2 that will use this compiled code is an I7 and
> machine #3 is an AMD Phenom II.
>
> If leave CPUTYPE alone, will the resulting code run well on all 3
> machines?

If you don't touch CPUTYPE, generic binaries are compiled for the target
architecture. As long as you're building for the same architecture,
everything should be fine. That's how "official" packages and
installation images are built. ;)

I won't bother with any "debate" about CPU-specific optimizing, but I'll
say that if you're compiling code on a machine other than the one you'll
be running it on, or the machines aren't all identical, just leave
compiler/optimization flags alone. I've had code compiled using one i7
quad-core model fail to run on another i7 quad-core model from only two
years earlier.


-- 
::  Brandon J. Wandersee
::  brandon.wandersee at gmail.com
::  --------------------------------------------------
::  'The best design is as little design as possible.'
::  --- Dieter Rams ----------------------------------