> On Thu, 25 Dec 1997, Steven N. Hirsch wrote:
>
> > I don't think that 2.1.76 is purported to be functional on Alphas. I've
> > been fighting with mine for the past two days. After switching to a build
> > that uses one of Donald Becker's newer tulip drivers it gets as far as
> > starting syslogd, then blows up with a kernel panic. I didn't have to
> > update syslogd on any of the Intel/Cyrix boxes, so I don't really believe
> > that it's the root cause.
>
> I've had problems with syslogd since 2.1.60. The machine would make an
> odd beeping sound and lots of exception messages would scroll past on the
> logging console. Interactive response was way down (press some keys and
> they will be actioned all at once in 20 seconds' time). I had a
> suggestion to recompile syslogd against the newer kernel for that problem
> to go away, but I couldn't get it to compile.
>
> I just don't run syslogd/klogd on my Alpha any more, but I've been having
> problems under X with some of the newer kernels (things like nxterm
> zombifying and emacs taking an age to refresh).
Ok, here's what I've found so far:
I was able to boot 2.1.76 by removing syslog/klog startup from the rc3.d
init scripts. Turns out that the alpha variants of 2.1.76 default without
Unix-domain sockets enabled. I knew this had been corrected in the Intel
kernels and frankly didn't think to look. As I write, the UDB is busy
rebuilding with sockets enabled (this also stopped X from coming up, BTW).
The decision to make Unix-domain socket support default off is
ill-considered, IMHO.
Otherwise, 2.1.76 actually seems to work. I'm planning to bring up knfsd
and autofs later and will let you know how it goes.
Steve