On Mon,  4 Dec 2006 22:03:03 -0500
"Stephen M. Rumble" <stephen.rumble@utoronto.ca> wrote:

> Quoting George Harvey <fr30@dial.pipex.com>:
> 
> > On Thu, 30 Nov 2006 22:37:36 +0000
> > George Harvey <fr30@dial.pipex.com> wrote:
> > [ snip ]
> >
> > After several disk swaps, re-installs and Perl builds, I think I'm
> > finally getting somewhere. The combination that's currently working
> > is 3.1 userland with a -current (4.99.3) kernel. I've successfully
> > built Perl 5.8.8 from pkgsrc, with pkgsrc NFS mounted over the G130
> > 100Mb Ethernet interface, and I'm currently building a number of
> > other packages that I need. Looks like it was the combination of a
> > 4.0 BETA userland (before the recent re-branch) and a -current
> > kernel that was causing the kernel hangs.
> 
> Glad to hear that you're having success. Now if only we knew why,
> exactly.

My guess would be something different in the shared libraries but I know
that doesn't help much.

> I was hoping that somebody more knowledgable would comment regarding  
> the "amap_wipeout" panic you saw, but I imagine the right eyes aren't 
>  watching this quiet list. When you get a chance, could you send the  
> R5000 revision? It should be in the dmesg. Also, is your O2 an R5000  
> as well? I'm wondering if our kernel or toolchain has any problems  
> with possibly older revisions of those parts.

Mu Challenge S has a R5000 rev. 1.0 (0x2310) and my O2 has a R5000 rev.
2.1 (0x2321). The O2, running 3.1, has shown no problems building
packages from pkgsrc but does sometimes hang after a warm reboot. No
problems with a cold boot. Since re-installing with a 3.1 userland, the
Challenge S has been churning away doing package builds with no further
problems.

> When I get a chance in a few weeks I'll have to run some builds and  
> see if I can reproduce these problems on my R4400 machines.

With the 4.0 BETA userland, the Perl build always failed, though it
could take a while before it happened (a successful build takes around
2.5 hours on my box). The 3.1 install kernel panicked every time as
well.

Thanks,
George