On Tue, Oct 28, 2025 at 1:35=E2=80=AFPM Dag-Erling Sm=C3=B8rgrav <des@freebsd.org> wrote:

Thomas Schweikle <tschweikle@gmail.com>= ; writes:
> Dag-Erling Sm=C3=B8rgrav <des@freebsd.org> writes:
> > Thomas Schweikle <tschweikle@gmail.com> writes:
> > > Dag-Erling Sm=C3=B8rgrav <des@freebsd.org> writes:
> > > > How was this system installed?
> > > By compiling 15-STABLE from latest available 14.3 getting sources via=C2=A0 =C2=A0
> > How did you install 14.3?
> Same way. It was Upgraded from 13-STABLE. And this was upgraded from
> 12-STABLE. If I remember it right, the system installed from
> disquettes was 5.0-RELEASE some way back in time ...

Do you understand the difference between the words =E2=80=9Cins= tall=E2=80=9D and
=E2=80=9Cupgrade=E2=80=9D?

What did you use prior to etcupdate?=C2=A0 When was /etc last updated?

=C2=A0

The last time mergemaster was available. Later on it was done manually, since etcupdate did not work.

My findings: etcupdate just does not work, because right after cloning etc is not in a working state. It is, after running

make _legacy

in "/usr/src". Then it runs

etcupdate extract

etcupdate=C2=A0diff

without this "Failed to build new tree.", but it t= hen fails run

etcupdate -p

right after building world, kernel and installkernel, exhausting: "No previous tree to compare against, a sane comparison is not possible." jus= t because there is no tree to compare against, or better: "etcupdate extract" created an empty tree without a= ny files within.It is just "make _legacy" creates all = the folders, etcupdate expects, but not the files. It seems all those advices given within the handbook or at various places within the internet all give it the wrong=C2=A0way:

clone

ettupdate extract

etcupdate diff

make buildworld

make buildkernel

make installkernel

etcupdate -p

reboot

make installworld

etcupdate -B

reboot

But

clone

make buildworld

make buildkernel

etcupdate extract

etcupdate diff

make installkernel

etcupdate -p

make installworld

etcupdate -B

reboot

because you will never have a working etc before building world and kernel. And in tune you'd never will have anythin= g you could extract. You are assuming something to extract, but there isn't anything before building. mergemaster did get this right (comparing the fresh build /usr/src etc against /etc). etcupdate does not -- at least if it is used the way the handbook advises. It would only work this way, if you did not clone the working tree right fresh into an empty directory (or after "git reset hard" -- removin= g anything from /usr/src what was created after the last "gi= t pull" simulating "git clone" as far as possible)= .

It=E2=80=99s a bit risky and not entirely in line with the Handbook = to skip the first reboot, but if you=E2=80=99re upgrading from a relativ= ely recent 14.3-STABLE and your root filesystem is on ZFS, you can create a backup Boot Environment (BE) as a safeguard in case something goes wrong.=C2=A0
Even better, you can create a testing BE and perform the installation into that BE after mounting it by using DESTDIR. Just remember that both mergemaster and etcupdate must also be executed with respect to this DESTDIR path.

If you=E2=80=99ve used etcupdate in the past and weren=E2=80=99t sat= isfied with its behavior, and therefore continued using mergemaster, I recommend cleaning the cruft by running the following command before starting the final transition to etcupdate: