On 11/30/25 3:36 PM, Robert Alessi wrote:
On Sun, Nov 30, 2025 at 08:46:31PM +0100, Jan Stary wrote:
On Nov 30 12:45:00, [email protected] wrote:
> Yesterday, I wanted to upgrade my VPS to 7.8, but I did
>
> sysupgrade -ns
>
> This was clearly a mistake.
Why? Is there something specific you need
that works in 7.8 but not in a snapshot?
No, but as this is a server, I find it simpler and arguably more
stable to just simply syspatch than to go through the whole process of
sysupgrade to keep my machine updated. Besides, I have binaries and
their associated libraries in the chroot of httpd that need to be
manually replaced when new versions are released. I would gladly do
without this!
unrelated suggestion: script your chroot updates. then it becomes
* sysupdate -s
* pkg_add -u
* updatechroot
done...
yes, getting your updatchroot script working perfectly is a bit of
testing, but SO worth it once done.
back to the topic: well, here's your choices:
1) run -current until next release. Probably no issue, other than you
may need/want to do a sysupgrade -s along the way.
2) push the system backwards to 7.8. May create all kinds of terrifying
issues...or may work just fine.
A couple thoughts...
1) IF you haven't yet upgraded packages, pushing backwards may work
fairly well. IF you have upgraded your packages, I'd suggest riding
the -current wave. Really.
2) I just happened to do this today to troubleshoot a bizarre and
subtle problem(*). I did not have problems with the backwards push,
and in fact, I'm still running packages from the week old snapshot.
However, the "fanciest" package I have on this machine is "rsync",
and rsync is really easy-going in terms of dependencies.
Your millage will differ. But the official answer is "forward only"
and if you wish to push backwards, you get to keep the pieces if
it breaks. But then, you were already semi-resigned to reinstalling.
But personally...I normally only run snapshots.
Nick.
(*) This problem probably won't impact you unless you run a very old
revision of the firmware on an ixl(4) card. But I haven't had a
chance to properly track down the problem yet, so ixl(4) MAY be
innocent.
