Date: Sun, 19 Jun 2005 23:32:18 +0200 From: Andreas Tobler <toa@pop.agri.ch> To: Matt Emmerton <matt@gsicomp.on.ca> Cc: Tarc <tarc@tarc.po.cs.msu.su>, =?UTF-8?B?77+9?= <des@des.no>, freebsd-current <freebsd-current@freebsd.org> Subject: Re: -CURRENT crashes on compilling Message-ID: <42B5E462.7090501@pop.agri.ch> In-Reply-To: <001b01c57513$9e9265b0$1200a8c0@gsicomp.on.ca> References: <20050617180321.GA1131@tarc.po.cs.msu.su> <867jgskfvd.fsf@xps.des.no><20050618105622.GA723@tarc.po.cs.msu.su><86k6krg0z5.fsf@xps.des.no> <42B5D82E.2090509@pop.agri.ch> <001b01c57513$9e9265b0$1200a8c0@gsicomp.on.ca>
next in thread | previous in thread | raw e-mail | index | archive | help
Matt Emmerton wrote: >>Dag-Erling Smørgrav wrote: >> >>>Tarc <tarc@tarc.po.cs.msu.su> writes: >>> >>> >>>>On Fri, Jun 17, 2005 at 08:17:10PM +0200, Dag-Erling SmЬrgrav wrote: >>>> >>>> >>>>>Bad hardware - most likely bad RAM, possibly a bad CPU. >>>> >>>>Hmm... I thinked about this, but RAM is ok. >> >>Side note, I could not have a look at your config/dmesg etc. >>tarc.po.cs.msu.su could not be found. >> >> >>>How do you know? Most software memory testers don't load the system >>>enough to trip over marginal RAM; 'make buildworld' does. >> >>Yeah, buildkernel and buildworld are good stress tests. The same as a >>gcc bootstrap is. >> >> >>>>How test processor? >>> >>> >>>'make buildworld' with known-good RAM is a pretty good indicator. >> >>Here my side note 2, it's not on x86, but on ppc. I have a powerbook >>with one GB of ram. Doing buildworld and buildkernel ended up in such >>sig 11 failures at random places. Not reproducable. >> >>I know my hw is ok, I do daily gcc bootstraps and the machine works. But >>under fbsd ppc I got the above sig 11 issues. A short talk with Peter >>Grehan made me try to reduce the physical memory software side with >>hw.physmem=512M. Bingo, that was it. I could do buildkernel and >>buildworld with hw.physmem=512M. No problem. So, our thinking is, that >>there is a trouble with physical mem > 512/640MB on fbsd ppc. > > > It more likely means that there is a problem with the memory chip(s) that > hold the upper 512MB of memory on your system. A gcc bootstrap is probably > not exercising this memory; a FreeBSD buildworld is. Well, I should have mentioned that the gcc bootstrap procedure happened under OS-X. And I have the experience that bootstrapping gcc with java takes an enormous amount of memory. So this is a prove for me that the hw is ok. It could still be that there is a hw problem. I just wanted to mention the fact that there is some other world. But going back, I suppose, if you have x86, the issue could be really related to hw issues. Regards, Andreas
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?42B5E462.7090501>