*BSD News Article 32868


Return to BSD News archive

Newsgroups: comp.os.386bsd.questions
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.cs.su.oz.au!metro!physiol.su.OZ.AU!john
From: john@physiol.su.OZ.AU (John Mackin)
Subject: Re: gcc dies on signal 10 and 11 on kernel compile
Message-ID: <1994Jul19.080403.28087@physiol.su.OZ.AU>
Organization: The Land of Summer's Twilight
References: <Ct55AA.MzJ@usenet.ucs.indiana.edu> <Ct5Bou.Cx0@tfs.com>
Date: Tue, 19 Jul 1994 08:04:03 GMT
Lines: 41

In article <Ct5Bou.Cx0@tfs.com> julian@tfs.com (Julian Elischer) writes:

> In article <Ct55AA.MzJ@usenet.ucs.indiana.edu>,
>	Jim Pitts <pitts@mimosa.astro.indiana.edu> wrote:

>> Everything went fine with the install.  It wasn't until I tried to recompile
>> the kernel that I had my first problem.  The compiler kept halting on a
>> signal 10 or a signal 11.

> If it makes you feel any better, ref.tfs.com also suffers from this
> problem. I haven't spent a LOT of time trying to track it down,
> but there is something screwed in the VM system that is triggered by some
> hardware configurations or something.

Yes, that's definitely correct.  I had this when I first put a FreeBSD
machine together.  I wonder: do either of the machines exhibiting
the problem have much memory/swap?  The configuration I got this
problem on was 4Mb RAM and 4 Mb of swap.  I knew it was something
screwy with the kernel, since the signals weren't repeatable:
you could restart the make and it would continue.  I'd see either
signal 11 in gcpp or gcc1, signal 10 in gcc1 or signal 11 in make
itself, or signal 6 (abort) in gcc1.

I assumed the problem was due to exhaustion of some resource, since
the make would always run for ten or fifteen files before dying.
I also noticed that the top-level make, the one that was controlling
the build and hence not exiting (until it died), was actually
growing while the build was going on: something I was rather
surprised to see, and that made me wonder if the machine was
just plain running out of VM.  I could well believe that the
programs wouldn't check properly for success of memory allocation.
I didn't investigate it further since I had other hardware problems
at that stage, due to a flaky NPX.

On my current configuration, with a different motherboard, 8 Mb RAM
and 8 Mb swap, it doesn't die.

-- 
John Mackin <john@physiol.su.oz.au>
Knox's box is a 286.                 Fox in Socks does hacks and tricks
Knox's box is hard to fix.           To fix poor Knox's box for kicks.