Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 7 Dec 2002 13:50:26 -0800
From:      Kent Stewart <kstewart@owt.com>
To:        John Mills <john.m.mills@alum.mit.edu>, John Mills <jmmills@telocity.com>
Cc:        freebsd-questions <freebsd-questions@FreeBSD.ORG>
Subject:   Re: buildworld fail
Message-ID:  <200212071350.26164.kstewart@owt.com>
In-Reply-To: <Pine.LNX.4.21.0212070822450.2236-100000@otter.mills-atl.com>
References:  <Pine.LNX.4.21.0212070822450.2236-100000@otter.mills-atl.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Saturday 07 December 2002 05:40 am, John Mills wrote:
> Kent, all -
>
> On Fri, 6 Dec 2002, Kent Stewart wrote:
> > Just remember that a signal 11 in a compile is usually your
> > hardware telling you something is wrong. If it dies at the same
> > spot, it can be options or whatever. If you use strange options,
> > you have to try them first but still remember a signal 11is
> > normally related to memory or heating problems.
>
> Yes, but when several users with different systems in different
> parts of the world report the same message in the same spot, I
> rather suspect a configuration or tool issue. For example, tripling
> the RAM didn't change it.

I agree, if it is exactly the same spot. I have several machines that=20
are not having problems. I have a AMD 2000+ that I will do a=20
buildworld on if someone is having trouble. It was upgraded to=20
RELENG_4 2 days ago. All that means is that it tends to points to=20
your machine more than RELENG_4. I trust RELENG_4 to the point that=20
it has my cvs-mirror on it. It i also fast enough that I won't miss=20
the 20 minutes of cpu time required to do a buildworld.

I also think that if you have a system built on a bad set of options,=20
you can have real problems upgrading to get rid of a buggy compiler=20
and userland. If I am given a choice of a system that runs 10% slower=20
and is absolutely stable, I will run that way rather than add options=20
to get the 10%. I think the time is being spent on gcc-3 as far as=20
cpu options are concerned. I dropped back to defaults when I reached=20
that conclusion.

I have one system that I added an 80 GB HD. Adding it involved=20
rearranging my HDs in the case. I cross mounted 2 of them and the=20
system wouldn't boot. I moved them where they were supposed to be and=20
still had problems doing builds. It was if I broke some links. In=20
order to get a stable system, I had to NFS mount /usr/src and=20
/usr/obj and do my installkernel and installworld from it.

>
> I may well have a bad set of configuration options for these newer
> releases. I am very new to FreeBSD. I followed the Handbook
> configuration instructions to reflect my hardware and didn't
> knowingly set compiler options at all. Your note was the first to
> remark on compiler options in connection with this error message so
> it certainly warrants my attention.
>
> I build RELENG_4_5 seamlessly with the same configuration file (in
> a RELENG_4_5 installation, however), so an environmental or
> generational issue is likely here.

Everytime you upgrade a level, i.e., 4.5 to 4.6, I think you should=20
redo your kernel config file. You never know what has been obsoleted=20
and added.

>
> I _have_ also hit heating problems and swap overruns, but these
> showed different 'footprints'. The partition size issue is my real
> stimulus to reconfigure the setup before I am more invested and
> dependent on it: "A house built on sand ..." etc.

Yes, that footprint is readily apparent once you have hit it. I think=20
people forget that when you have heating problems, a single bad build=20
means more than 5 or 6 good ones.=20

There were so many signal 11's in the past that they wrote a FAQ on=20
it. It is at=20
http://www.freebsd.org/doc/en_US.ISO8859-1/books/faq/troubleshoot.html#SI=
GNAL11
and is a good place to bookmark. Much less trouble than finding it=20
when you need it :).

I have /usr/src and /usr/obj in their own 1.5 GB partitions on their=20
own HDs. Each HD has its own ATA-100 controller. Most of these=20
machines have 256 MB of memory and don't swap. My gateway only has=20
128 MB but I have never looked at swap while I was doing a=20
buildworld. I also have setiathome running in the background while=20
everything is going on. The systems are always at 100% cpu. When you=20
run out of swap, the system used to just panic. I don't think that=20
has changed.

I started out with small /, /var, and /tmp. My later versions have /=20
with 250 MB and /var and /tmp each have 1.5 GB. The large /var lets=20
me log everything from my cvsups (ports, docs, src), builds, and=20
installs. Part of the cvsup update process converts the cvsup.log=20
into an html file that lets my web browser link directly into the=20
cvs-repository. When I have used up 1/2 of /var, I clean out the old=20
stuff.

>
> Thanks for your comments. I need to dig out my error messages and
> compare the compilation options with the other poster's.

Good luck on finding the problem.

Kent

>
> Regards.
>  - John Mills
>
>
> To Unsubscribe: send mail to majordomo@FreeBSD.org
> with "unsubscribe freebsd-questions" in the body of the message

--=20
Kent Stewart
Richland, WA

http://users.owt.com/kstewart/index.html


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200212071350.26164.kstewart>