Bug#478409: problem with SEGV still persists

[Niko Tyni]

severity 478409 grave
thanks

On Fri, Feb 06, 2009 at 05:45:54PM -0500, Simon wrote:
> I have an in-progress Etch->Lenny upgrade going and the problem still 
> persists with the perl segfaults. I would consider this to be grave 
> since debconf cannot run properly, and makes most of the configuration 
> of packages useless and very broken.

Wow, that's bad news. I'm raising the severity and
cc'ing the debian-powerpc list.

I assume this is triggered by some combination of Etch and Lenny packages
during the upgrade. Can you supply the dpkg and apt/aptitude logs
up until the failure?

Is libc6 upgraded yet? (The perl-base pre-dependency says it should,
but I can't think of much else.)

Which kernel are you running on?

If your system is still broken, does just

# perl -e 'use Fcntl'

segfault, or is there some other simple testcase?

> I've been trying to work around it since early this morning $localtime 
> (7:30 EST, n. america, it is now almost 18:00), and no matter what I 
> do, it crashes. The perl* version I have is 5.10.0-19, the latest as of 
> this writing.  Still seems to have odd crashes, "undefined symbol" 
> errors, Perl_pv_something in Fnctl/Fnctl.so and other compiled modules. 

The ldd errors you mentioned in a followup are normal AFAICS: perl
dlopen()'s the shared objects and so the symbols are defined at runtime.

> Also, the libperl5.10 package seems to be empty? Is that normal?

Yes. See the package description.
-- 
Niko Tyni   ntyni at debian.org