Subject: Re: Flakiness in build
To: Rick Byers <email@example.com>
From: Raymond Wiker <firstname.lastname@example.org>
Date: 06/11/1997 10:41:51
Rick Byers writes:
> On Tue, 10 Jun 1997, Martin Cernohorsky wrote:
> > > A friend that bought some cut-rate DRAMs saw exactly this sort of
> > > behavior. Things would mostly be ok, but compiles would fail with
> > > intermittent off-the-wall gripes. Replacing the newly acquired junk
> > > DRAM with good Micron DRAM cured the problem.
> > Hmm, is there any possibility to verify the function of the upper part
> > of the RAM simpler way than physically removing the DRAM in question
> > (i. e. something like configuring the size of the RAM during boot
> > time)?
> Hmm... It's too bad it's not more common. It's hard to verify when it's
> working and when it's not. I started a little test program that just
> allocates 40Mb, fills it with a known sequence, and then checks it. So
> far, it hasn't failed. It's not perfect, but if it does fail, I'll know
> it's not a bug in make or anything (I think I allready know that). Oh, I
> should mention, Linux gave a similar compiling error last night as well
> (gcc gave a syntax error, but starting make again worked fine). Now I
> just have to find where the memory is with the problem. I wonder if there
> is any way to make a certain area of memory not-usable (if it is just a
> certain part) - I guess you could just allocate it from the kernel at
> bootup, but you'd have to stop it from being paged.
I had a similar problem when I was setting up a machine with a
genuintel Pentium 133 and 16 megs of 4MB, 80 ns SIMMS. I could get
FreeBSD up and running on it; but programs would sometimes crash for
no reason. Win 95 on the same machine was completely unusable.
I then switched to an AMD K5-133, which worked like a charm,
and later upgraded to 64 megs of 32MB, 60 ns SIMMS. I have been
thinking of trying the Pentium chip again, but I guess I won't bother.
To summarise: I had the same problem, and solved it by
replacing the Pentium chip with a an AMD equivalent. The problem may
have been a faulty processor, or some sort of mismatch between the
Processor, motherboard and memory; in any case, the symptoms were
similar to what you're seeing.