Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 17 Nov 2014 22:50:18 -0800
From:      "Chris H" <bsd-lists@bsdforge.com>
To:        Kevin Oberman <rkoberman@gmail.com>
Cc:        "freebsd-stable@freebsd.org" <freebsd-stable@freebsd.org>
Subject:   Re: best overall upgrade from 8.x?
Message-ID:  <3930066eddaa3b17754f5f9cdfba5e6d@ultimatedns.net>
In-Reply-To: <CAN6yY1tqewh_e7D0V-i1DX1rj=h_bNCNdBmki6U-LHs4bdaKzA@mail.gmail.com>
References:  <CA%2BE3k93YhbfCN1La8sJO%2BWA-%2BeBVnSHNO-JjCsd%2BNs5%2BJq6iNw@mail.gmail.com> <CAN6yY1tcgrO4ZgOj4feFA3hcVj%2BgP0Sq5gPcu6xmFgrONaUq6Q@mail.gmail.com> <54673CA1.5020008@netfence.it> <20141115125347.GC16633@swelter.hanley.stade.co.uk> <B40697E4-6CBD-4F88-805F-654E9C3E8238@FreeBSD.org> <CA%2BtpaK37cUgT9MG2t%2BaNfd=1-FMWLnj2fmGX9Pj=JG2BZuTzsQ@mail.gmail.com> <971085e211564c84a0460ca945eda837@ultimatedns.net>, <CAN6yY1tqewh_e7D0V-i1DX1rj=h_bNCNdBmki6U-LHs4bdaKzA@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 17 Nov 2014 22:13:11 -0800 Kevin Oberman <rkoberman@gmail.com> wrote

> This is a bit nit-picky, but... (in-line)
> 
> NOTE personal experience indicates that regression issues introduced
> > by pkg(8), makes >=8.3-* the minimum starting point. YMMV
> >
> > I have used the following for ~10yrs. w/o issue (comments added for
> > *hopeful* clarity).
> >
> > The IMMEDIATELY following, is only relevant RE-building world/kernel:
> > cd /usr/obj
> > chflags -R noschg *
> > rm -rf *
> >
> >
> > cd /usr/src
> > make buildworld
> >
> > NOTES:
> > SINGLE CPU: make -j4 buildworld
> >
> > NOTE:
> > MULTI CPU: make -j(6 through ??) buildworld
> >
> > # pre version 9
> > make kernel-toolchain
> >
> > make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=<your_kernel_name_here>
> >
> I would add -jN to this using the same value for 'N' as used for buildworld
> 
> make -DALWAYS_CHECK_MAKE installkernel KERNCONF=<your_kernel_name_here>
> >
> > reboot (in single user mode)
> >
> > fsck -p (optional, but a good idea)
> > mount -u /
> >
> Redundant before the fact as the next command (mount -a -t ufs) does this,
> too
> 
> mount -a -t ufs
> > swapon -a (most cases; optional)
> >
> > adjkerntz -i (not always necessary, but DO check current time/date, just in
> > case)
> >
> In more fetail, if the hardware clock is UTC, this is not required. If
> /etc/wall_cmos_clock exists, this is needed, but you really should do it
> BEFORE mounting disks. UPDATING shows it right after the fsck. It avoids
> any possibility of something being time-stamped in the future.
> 
> >
> > mergemaster -p
> >
> This is way, way to late. On at least one case, this would have lead to a
> very awkward failure of installworld. The man page is explicit that it
> should be run before buildworld. This is the only thing in this that is
> really, really wrong and can lead to a recoverable, but very annoying
> failure.
10yrs. and a few thousand installs, says otherwise... BUT, the
more I know, the more I know I don't know. :)
> 
> 
> >
> > cp -Rp /etc /etc.old
> >
> Not really needed, but good for the "belt and braces" type.
Yes. Just like recent dump(8)'s. ;)
> 
> >
> > cd /usr/src
> >
> > make installworld
> >
> > mergemaster -F
> > or mergemaster -vF (my personal choice)
> > or mergemaster -cvF
> >
> > make delete-old
> >
> This can be a bit painful as you are prompted for every file. I suggest
> "make check-old" and then "rm -rf" any directories listed as ready for
> deletion. direcctories are listed near the end, just before libs.
> 
> >
> > reboot
> >
> > HTH
> >
> > --Chris
> >
> --
> R. Kevin Oberman, Network Engineer, Retired
> E-mail: rkoberman@gmail.com

--Chris





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3930066eddaa3b17754f5f9cdfba5e6d>