Bug#202756: I can reproduce this willingly on my smp box

[Adam Heath <doogie@brainfood.com>]

I was going to start doing some work on dpkg.  So, I decided to upgrade to the
latest versions of all it's build-depends.  This of course brought in libc6.

However, when it started to upgrade from 2.3.1-17 to 2.3.2-5, the old postrm
and the new preinst both segfault.  Existing programs work fine, however.

In trying to fix that, I made the very bad mistake of doing dpkg-deb -x
libc6.deb /, which then made *all* programs segfault on the host.  Don't do
that.

Anyways, I found an old potato chroot, which I was able to eventually get
into(using it's ld.so, and ld_library_path).  I then did the following series
of steps:

make woody chroot.
upgrade to testing.
install libc6-{dbg,dev,pic}	# I had these on the host, don't really think
				# it affects the bug.  apt just did an
				# --unpack on them in the same run.
edit sources.list, change to unstable
apt-get install libc6{,-{dbg,dev,pic}}

I then get the following error:

Preparing to replace libc6 2.3.1-16 (using .../libc6_2.3.2-5_i386.deb) ...
Unpacking replacement libc6 ...
dpkg: error processing /var/cache/apt/archives/libc6_2.3.2-5_i386.deb (--unpack):
 dpkg: warning - old post-removal script killed by signal (Segmentation fault), core dumped

dpkg: error while cleaning up:
 subprocess pre-installation script killed by signal (Segmentation fault), core dumped

What's interesting is that I can run the old postrm and old preinst fine.

The machine this is broken on is a dual celeron 333 running linux 2.5.68.  The
host is busted right now(I have sash, so can do some things, on the *single*
ssh window I have open).  I'm going to be installing ssh in one of the
chroots, on a different port, so that others can play with this.

I've just verified that it still fails if I remove libc6-{dev,dbg,pic} before
installing just libc6 from unstable.

I can give out access to this machine, but not until tomorrow.  It's also
behind a firewall, so some bouncing will be required.