[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#505829: installation-report: Installation on sparc U10 - success with some issues

reassign 505829 hw-detect 1.70
retitle 505829 [sparc] Tries to install obsolete package libc6-sparcv9
severity 505829 minor

On Saturday 15 November 2008, Frans Pop wrote:
> Install boot loader:    [?]

This was because the system did not reboot into the new system, but that 
could be because I installed to the second hard disk. I do not plan to
investigate this further.

> Comments/Problems:
> 1) The installer tries to install the package libc6-sparcv9, but that
> does not exist anymore in testing as it has been replaced by
> libc6-sparcv9b. However, the old package should still be installed for
> Etch installs.

Hmmm. That initial analysis was incorrect. The code that queues these 
packages for installation is (hw-detect.sh):
        if grep -q '^type.*: sun4u' /proc/cpuinfo ; then
                # sparc v9 or v9b
                if grep -q '^cpu.*: .*UltraSparc III' /proc/cpuinfo; then
                        apt-install libc6-sparcv9b || true
                        apt-install libc6-sparcv9 || true

No idea if this is still valid or that we should always install the 9b 
variant now. Fact is that Etch has _both_ packages while Lenny only has 
the 9b variant.
From the glibc (2.6-3) changelog:
  * Drop SPARC V8 support:
    - control.in/opt: drop libc6-sparcv9.
    - sysdeps/sparc.mk: drop libc6-sparcv9.
    - script.in/kernelcheck.sh: add check for SPARC V8 or earlier CPU.
    - patches/series: disable sparc/submitted-timing.diff.
    - patches/sparc/local-sparcv8-target.diff: rename to
      local-sparcv9-target.diff and default to v9.

So my *guess* is that as v9 is now default, the "else" branch is no longer 
needed, but the v9b variant is likely still useful for the same cases as 
the original test. If that is correct, the "else" branch can be dropped 
post-Lenny, which will ensure we keep compatibility with Etch for the 
Lenny version of D-I.

Alternative would be to drop it now and have etch-support install it, but 
that's more trouble than it's worth IMO.

> 2) I ran the web server to check installer logs. The logs show a kernel
> oops at that point. 

I've reported this oops to the upstream sparc kernel mailinglist:

Will follow up to the kernel team if anything comes out of that.


Attachment: syslog.gz
Description: GNU Zip compressed data

Reply to: