Success updating stable/11 to /12; a couple things to note

[Ian Lepore]

On Thu, 2018-12-27 at 08:30 -0800, David Wolfskill wrote:
> On Thu, Dec 27, 2018 at 09:13:09AM -0700, Ian Lepore wrote:
> > 
> > On Thu, 2018-12-27 at 05:53 -0800, David Wolfskill wrote:
> > > 
> > > * I found that I actually needed to create the ntpd user on the
> > >   running system prior to "make installworld" -- having run
> > >   "mergemaster -U" against the target (DESTDIF) was insufficient.
> > The correct update sequence involves running mergemaster twice,
> > once
> > with the -p option, then again later without it. It's detailed at
> > the
> > bottom of UPDATING. People get in the habit of skipping the -p step
> > because it's only really needed once every few years, such as when
> > a
> > new user is added to the base system.
> > 
> > -- Ian
> > ....
> Yes, but the one after "make installworld" isn't likely to affect the
> "make installworld".  :-)
> 
> The sequence of events:
> 
> mount -u -r /
> mount -u -r /usr
> mount /dev/ada0s1a /S1
> mount /dev/ada0s1d /S1/usr
> mount -u -w /S1
> mount -u -w /S1/usr
> ln -fhs /var /S1/var
> mount -o ro freebeast:/usr/src /usr/src
> mount -o ro freebeast:/usr/obj /usr/obj
> id
> mount
> cd /usr/src
> uname -a
> date
> make LD=ld.lld installkernel DESTDIR=/S1
> date
> mergemaster -U -u 0022 -p -D /S1
> date
> rm -fr /S1/usr/include.old
> date
> mv -f /S1/usr/include /S1/usr/include.old
> date
> rm -fr /S1/usr/share/man
> date
> make installworld DESTDIR=/S1
> date
> mergemaster -F -U -u 0022 -i -D /S1
> date
> make delete-old DESTDIR=/S1
> date
> date
> df -k
> date
> 
> 
> My point was that running "mergemaster -U" against the new image
> (/S1,
> in the case above) is not sufficient for "make installworld
> DESTDIR=/S!"
> to work: it is necessary that the *running* system be aware of the
> "ntpd" user (I presume, to allow ownership of files to be set by the
> "ntpd" name, vs. the numeric "123").
> 
> Peace,
> david


I must have missed the part where you were using DESTDIR= during the
install. In that case, the right magic is to add DB_FROM_SRC=yes to the
make installworld command. There were a couple proposals to automate
that somehow, but I didn't see anything proposed that didn't basically
subvert the entire purpose of that check for whether the user exists.

-- Ian