Bug#325258: 3 problems in daily powerpc d-i installs
Package: installation-reports
Debian-installer-version: sid daily-built 2005.08.26 powerpc
uname -a: Linux pegasos 2.6.12-1-powerpc #1 Tue Aug 9 10:36:53 UTC 2005 ppc GNU/Linux
Date: Sat Aug 27 03:02:33 CEST 2005
Method: chrp tftp netboot of vmlinuz-chrp.initrd
Machine: Genesi Pegasos 2
Processor: PowerPC 7447 (G4)
Memory: 2GB
Root Device: ide -> /dev/hda
Root Size/partition table: not on device right now.
Output of lspci and lspci -n:
Base System Installation Checklist:
[O] = OK, [E] = Error (please elaborate below), [ ] = didn't try it
Initial boot worked: [X]
Configure network HW: [1]
Config network: [X]
Detect CD: [X]
Load installer modules: [X]
Detect hard drives: [X]
Partition hard drives: [X]
Create file systems: [X]
Mount partitions: [X]
Install base system: [2,3]
Install boot loader: [X]
Reboot: [X]
Comments/Problems:
[1] -- The first problem encountered was relative to busybox, see :
Bug#325244: busybox FTBFS due to gcc-4.0 on powerpc, mips, mipsel and ia64, breaks d-i.
This is due to busybox 1.00-4/5 FTBFS due to a gcc-4.0 bug, and thus missing
longjmp GLIBC2.0 symbols, since it was not rebuilt against the newer glibc
2.3.5 included in the initrd. The symptom of this problem is :
/bin/sh: relocation error: /bin/sh: symbol longjmp, version GLIBC_2.0 not
defined in file libc.so.6 with link time reference
This was solved with a local rebuilt of busybox including the patch in the
bug report above.
[2] -- Second problem was while installing etch.
debootstrap apparently has this funny notion that it should install
ipchains, which is not present in etch, so the installation fails. sid has
not this problem, since the debootstrap sid rules are good.
I continued with a sid install.
[3] -- Thirdly, base-installer tried to install a kernel-image-2.6.8-powerpc-smp.
Well, this is the wrong kernel, and i was kind of expecting a
linux-image-2.6-powerpc for my arch/subarch/cpuinfo, especially as this is
what base-installers kernel stuff for powerpc should be chosing. Not sure
what went wrong here.
Basically, i believe that as we move to a common kernel for all
architectures for etch, most of the kernel-chooser stuff should be reworked
to take this into account, and made to use the linux-image-2.6-<flavour>
metapackages, at least for all 2.6 flavours. Hopefully the remaining 2.4
kernels will move to the common packaging infrastructure soon, or just plain
dissapear.
Ok, that is all for now, i will try to redo regular powerpc installs in the
future.
Friendly,
Sven Luther
Reply to: