Got a base system running
Hi all,
ok, at last I got a base system running, and so far I have only had a
few problems. Here's what I did:
a) The basedisks.sh script from boot-floppies (I used 1.1.1.9; I know .10
already exists) required only a few minor changes to run ok on m68k. Here
is a little diff. The only changes were for the new naming style (_ instead
of -), for NOT writing the base package to disk, writing 'ldso' instead
of 'ld.so', and leaving out some packages out of the list that don't
exist on m68k.
---
*** basedisks.sh.orig Tue Jul 11 00:14:46 2000
--- basedisks.sh.new Thu Aug 1 15:11:43 1996
***************
*** 30,36 ****
make_paths() {
local PATHS=""
for i in $*; do
! p=`last $archive/$i-[^a-zA-Z]*.deb`
if [ ! -f $p ]; then
echo "Can't find package" $p 1>& 2
exit -1;
--- 30,36 ----
make_paths() {
local PATHS=""
for i in $*; do
! p=`last $archive/$i\_[^a-zA-Z]*.deb`
if [ ! -f $p ]; then
echo "Can't find package" $p 1>& 2
exit -1;
***************
*** 99,105 ****
(cd $B; chroot $B sbin/ldconfig*)
trap "umount $B/proc" 0
chroot $B mount -t proc proc /proc
! (cd $B;dpkg --root=. --install `make_paths base/ld.so`)
(cd $B;dpkg --root=. --force-depends --install `make_paths base/dpkg`)
(cd $B;dpkg --root=. --install `make_paths base/libc5`)
--- 99,105 ----
(cd $B; chroot $B sbin/ldconfig*)
trap "umount $B/proc" 0
chroot $B mount -t proc proc /proc
! (cd $B;dpkg --root=. --install `make_paths base/ldso`)
(cd $B;dpkg --root=. --force-depends --install `make_paths base/dpkg`)
(cd $B;dpkg --root=. --install `make_paths base/libc5`)
***************
*** 122,127 ****
rm -f $B/var/lib/dpkg/status-old
(cd $B; tar clf - .)| gzip --best > base1_1.tgz
rm -f -r $B $P &
! ./utilities/floppy_split base1_1.tgz base14 1440
! ./utilities/floppy_split base1_1.tgz base12 1200
exit 0
--- 122,127 ----
rm -f $B/var/lib/dpkg/status-old
(cd $B; tar clf - .)| gzip --best > base1_1.tgz
rm -f -r $B $P &
! #./utilities/floppy_split base1_1.tgz base14 1440
! #./utilities/floppy_split base1_1.tgz base12 1200
exit 0
*** scripts/basedisks/PACKAGES.orig Sat Jun 15 23:54:49 1996
--- scripts/basedisks/PACKAGES Tue Jul 11 01:05:07 2000
***************
*** 1,5 ****
base/base
! base/ld.so
base/libc5
base/ncurses-base
base/ncurses3.0
--- 1,5 ----
base/base
! base/ldso
base/libc5
base/ncurses-base
base/ncurses3.0
***************
*** 7,14 ****
base/libdb1
base/libgdbm1
base/dpkg
- base/dpkg-ftp
- base/lilo
base/adduser
base/ae
base/bash
--- 7,12 ----
***************
*** 23,29 ****
base/gzip
base/hostname
base/kbd
- base/mbr
base/tar
base/modules
base/modconf
--- 21,26 ----
***************
*** 32,40 ****
base/procps
base/sed
base/setserial
! base/sh-utils
base/sysklogd
- base/syslinux
base/sysvinit
base/textutils
base/timezone
--- 29,36 ----
base/procps
base/sed
base/setserial
! base/shellutils
base/sysklogd
base/sysvinit
base/textutils
base/timezone
*** scripts/basedisks/PARTIAL_PACKAGE_FILES.orig Tue Jul 11 00:18:33 2000
--- scripts/basedisks/PARTIAL_PACKAGE_FILES Tue Jul 11 00:18:59 2000
***************
*** 15,18 ****
usr/lib/perl*/auto/Socket/autosplit.ix
usr/lib/perl*/strict.pm
usr/lib/perl*/vars.pm
! usr/lib/perl*/i486-linux/5.002/Config.pm
--- 15,18 ----
usr/lib/perl*/auto/Socket/autosplit.ix
usr/lib/perl*/strict.pm
usr/lib/perl*/vars.pm
! usr/lib/perl*/m68k-linux/5.002/Config.pm
--- cut here ---
Any suggestions how to merge this together with the i386 versions of
those scripts? Should we introduce scripts/i386/rootdisk and
scripts/m68k/rootdisk, with the prototype stuff being in scripts/all
or alike?
Adapting the script to create the root disk (rootdisk.sh) is harder; the
largest problem is that it makes heavy use of fdisk for determining
swap, Linux native and other partitions, where we don't yet have an fdisk
for m68k. So what I did here was to manually perform all the steps the script
would usually do: creation of fstab, hosts, init.d/network, setup of keyboard
and timezone etc.
After installing this modified base system on a partition and booting
from it, almost everything worked - with a few exceptions:
- As soon as init.d/boot calls init.d/network, I got a data read fault
from somewhere in the 0xfffe.... range. Funny thing is that this was
my fault - my self-created network file did not have the initial
line "#!/bin/sh". But should this really cause a kernel error? (although
it was non-fatal).
- syslogd doesn't yet want to work (1.3-8). It starts, but as soon as it
wants to write something into a file, it exits. I'll have to check older
buglogs for this one..
- /tmp and /var/tmp were not set rwxrwxrwt - but that seems to be my
fault again, because the base_1.1.0-13 (and above) packages DO have
correct permissions (I just checked).
Well, overall - things look pretty promising. Ok, one thing annoyed me a bit:
dselect is pretty slow when you go through the list of packages. I suppose
this is not as bad on fast 486 or Pentium machines, but as far as I'm
concerned, I'll stick with dpkg for now..
Frank
Reply to: