Date: Wed, 3 Jul 2002 10:58:46 -0400 (EDT) From: doug@safeport.com To: Doug Barton <DougB@FreeBSD.org> Cc: freebsd-bugs@FreeBSD.org Subject: Re: conf/40102: mergemaster should use the currend working directoryas the default for -m Message-ID: <Pine.BSF.4.21.0207031003100.2331-100000@pemaquid.safeport.com> In-Reply-To: <3D22FAD2.CEECC961@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
I understand your answer and your philosophy. I just disagree. I think
mergemaster's defaults should be compatible with the build process. I
also think it should not do the wrong thing when it is discernible that
is the case, especially when the resulting errors are so esoteric. It
seems to me that the core system (and installation??) is moving toward
the direction of trading ease of upgrading for a more complex process. I
think as the system does more and more for (to :) the user, checks
should be included to try to ensure the right stuff is being done. This
debate can cease at this point. If I have not made a glimmer of a point,
it is not possible for me to do so. I am not pushing this debate for me,
I am (now) beyond this problem.
The system I discovered this on is my build system. It started out from
a 4.1 CD and has been cvsup'd and built along from there. The last build
before this was 4.5 from /usr/src. If I have a mergemaster rc file it is
because one of the upgrades put it some where. On my workstations and
gateways, I have paid no attention to changing the defaults for the
periodic functions or logging or much /etc for that matter. I do crawl
over the production servers pretty much. Except for this last build, the
system I found the problem on has been updated from /usr/src. So the 4.5
base I built it from should have been ok?? Or so I would think/hope.
The only reason I am keeping this going (and this is my last attempt),
is that the resulting errors are very "far a field" from the cause. If
FreeBSD is to spread to the masses (as I would hope), it needs to be on
more desktops or it ultimately will not have to base of support to
survive. I know few if any in the BSD community believe (or care about?)
this but I think from my 35 years in computing I have seen enough
technically superior systems fall away from lack of user support or
perhaps a lacking of understanding of what to do to succeed. GE,
Burroughs, Xerox, to name a few. Maybe even AT&T.
My only thing directly from your answer:
> > 4) Lastly (and the killer for me) is that the process,
> > mergemaster'ing, works if /usr/src is empty.
> This is one thing that makes me believe that you have something
> non-standard in your environment, since this definitely won't work. If
> you have no /usr/src/etc/Makefile, mergemaster will fail without at
> least -m.
Quite possible. Note that the failing system is the only one with a
source tree in /usr/src. On the other systems I get the right stuff.
I have enough systems in this state if you feel this is worth pursuing.
If you give me a clue of what to look for, I am happy to do this. If it
is only a mergemaster rc file, I will just report back on that. I know I
did not put one anywhere.
BTW I think what I am doing (NFS builds) will (is??) more the norm for
companies using FreeBSD. Most of my nameservers and specialty servers
are between 75 - 133 MHz. Building on these systems is unacceptably slow
but they will have a computing life until they die or fry.
_____
Douglas Denault
doug@safeport.com
Voice: 301-469-8766
Fax: 301-469-0601
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0207031003100.2331-100000>