RS/6000 7043-140 boot floppies
Carlos E Murillo-Sanchez
ce.murillosanchez at gmail.com
Tue Feb 19 09:15:26 CST 2019
Alan Perry via cctech wrote:
> The system does boot the AIX install on one of its hard disks, but
> this is a recycled system and I don't have usernames/passwords for
> that install.
>
> Does anyone here have a suggestion on how to proceed?
>
> alan
Last year I started working on a 7012-320H that I've had for some time;
it would not boot AIX, but it would stall during boot up before bringing
up the console. People in this list and elsewhere made many useful
suggestions. I made the proper serial cabling, no console. Finally, I
was able to boot in system maintenance mode from boot AIX 3.2.5
diskettes that I found on the net, and got a console. I then
discovered that the hard drive had 3.1.005 in it; still, managed to
replace the /etc/security/passwd file (using just cat and cp, there's
no ls or ed in the maintenance shell) with a version that had the root
password removed. The system booted now, except that it would not mount
/usr, because there wasn't an /etc/mount binary. After examining an
image of the hard drive, I noticed that the binary for /etc/unmount had
error messages for mounting operations, had this hunch that /etc/mount
was the same binary as /etc/unmount, copied the latter to the former,
and voila, I had a booting system. You should be able to do the same if
you can boot from a CD.
Since then I added a second hard drive (the original was just 400mb),
and I have been trying to build a toolchain and numeric processing
programs. I was able to build gcc-2.7.2 after building an early gmake
with the system's cc even though the assembler
in 3.1.005 has a known bug, then I built binutils-2.9.1 (the last
version that will build ld on AIX 3.1.005), and then I succesfully built
gcc-2.95.3 configured to use binutils and not the system's ld and as.
The last version of gcc supported on 3.1.005 is 3.3.6, but the build
process fails early on at the dreaded gengtype stage with the usual
"running out of memory" error (this is something that also happens under
SunOS 4.1.4; my IPXs are at 2.95.3 too).
So, I'm stuck at 2.95.3, but I've managed to build at least some things
with it; right now I have: autoconf-2.59, bash-2.05b, binutils-2.9.1,
bison-1.50, bpmpd-2.30, bvi-1.3.2, bzip2-1.0.2, fftw-3.3.7, flex-2.5.4,
glpk-4.45, glpk-4.65, gawk-3.0.4, gmp-4.2.4, gnuplot-4.4.4, gperf-2.7.2,
groff-1.10, gsl-2.1, gzip-1.3.12, jpeg-8d, lapack-3.1.1, libgd-2.0.33,
libpng-1.4.22, m4-1.4.1, make-3.81, mpfr-3.1.6, ncurses-5.2,
patch-2.5.9, pcre-6.7, perl-5.6.2, qhull-2009, qrupdate-1.1.2,
readline-4.2, rx-1.5, screen-4.0.3, sed-3.0.2, texinfo-4.0, tiff-3.9.7,
vim-5.8, zlib-1.2.3. I had to modify the source in most of them. I have
built a python-2.2.3 executable but right now it fails when importing
modules, and I have been trying to build several versions of octave
without success. Most issues have to do with ancient system libs,
ancient c++ grammar in gcc-2.95.3, a nonstandard dynamic loading
mechanism, lack of UNIX98 features, and general senescence in this
system. But I like it anyway. I wish I had a network card. Sometime
in the future I will configure a SLIP server with an old linux box or a
raspberry and try to network the thing.
carlos.
More information about the cctech
mailing list