Haven't been able to make world in about a year

[Giorgos Keramidas]

On 2006-02-28 08:48, "Donald J. O'Neill" <duncan.fbsd at gmail.com> wrote:
> On Tuesday 28 February 2006 05:46, Giorgos Keramidas wrote:
> > Until you are satisfied that everything works without any problems at
> > all, please don't use scripts to run the builds.
> >
> > For instance, your script above lacks a call to ``mergemaster -p''
> > before the ``make buildworld'' step, which may be necessary.
>
> That's a call to 'script' to run, telling it where to put the resulting
> text file. While it is running, commands are given and executed. Later,
> if something goes wrong, or you saw something you want to check on, you
> have a record of what happened that you can look at, and pass on to
> others, if needed.
>
> Could you explain the logic of running mergemaster -p, when you have
> nothing to run it on. /usr/obj was blown away at the beginning of the
> buildworld sequence. What are you going to check at that point? Please
> take a look at the handbook for the sequence that things should occur
> in. I would say look at /usr/src/UPDATING, but their sequence doesn't
> include installing the kernel anymore.

On 2006-02-28 09:52, Kris Kennaway <kris at obsecurity.org> wrote:
>
> He means to run mergemaster -p before make installworld, as in the
> instructions quoted at the top of this email and in /usr/src/UPDATING.
> Also, the documented sequence *does* include installing the kernel, so
> I don't know what you mean there either.

Yes, thanks Kris.

Sorry for the buildworld/installworld confusion.  I meant right before
'make installworld'.

There are cases where 'installworld' will try to chown files to a newly
added system account (i.e. `_dhcp'), but will fail, leaving a half-installed
system if you don't run ``mergemaster -p'' before ``installworld''.

This is why I suggested *avoiding* a scripted, unattended build and install
cycle, until the OP who started this thread is comfortable that his builds
and installs are indeed going to succeed.