On Tue, 16 May 1995, Andre Skarzynski wrote:

> Michael L. VanLoon -- HeadCandy.com wrote :
> > 
> > 
> > >   I then removed the option CYRIX_CACHE_REALLY_WORKS form my kernel, so that 
> > the
> > >That doesn't surprise me at all.  There's a good reason that option is
> > >not documented anywhere.
> > 
> > To elaborate: that option is intended for people who have motherboards
> > (1) designed *specifically* for DLC-type chips, and (2) which have
> > *correct* and *complete* bus-mastering and cache-coherency support in
> > the hardware.  If you don't have these, chances are your cache doesn't
> > "REALLY_WORK", and you shouldn't have that enabled.
> > 
> 
> What is interesting, is that my M/B worked fine for some time like this even 
> building some large source trees and while under heavy load with X running.
> I only picked up a problem when building a comprehensive GENERIC kernel for
> a new installation...Strange.

which is why i brought up this subject in the first place.  under 1.0 and 
even the early stages of -current, my m/b was perfectly fine while 
compiling the whole system and also running x.  then all of a sudden, 
around the early part of this month, my system kept getting numerous sig 
11's and error code 139's while building anything.  it then seemed to get 
worse, dumping core under any heavy activity.  it got so bad that i 
couldn't even compile a new kernel because i kept getting the errors, and 
since it already takes me about two hours to build a kernel, i wasn't 
about to sit around for two hours straight to continually type "make."  
thus, i just downgraded back to 1.0 and am now stable, but wishing i 
could have all the new -current stuff.

anyways, i'm about to download the latest i386 binary snapshot and try to 
compile a new kernel with the suggested options and see how that works.

-m