There was one more problem. On machines having this symlink
lrwxrwxrwx 1 root root 9 Apr 27 2021 /lib64 -> usr/lib64 Installing package libc6 2.28-10+deb10u3 removed /lib64/ld-linux-x86-64.so.2 before end of the installation (during some package installation scripts). So the installation won't end correctly. Ending with the same broken system as mentioned earlier. This could be fixed with nasty hack (no warranty guaranteed) perl -e 'mkdir "lib64.new"; symlink "/lib/x86_64-linux-gnu/ld-2.28.so", "lib64.new/ld-linux-x86-64.so.2"; symlink "/lib/x86_64-linux-gnu/libgcc_s.so.1", "lib64.new/libgcc_s.so.1"; unlink "/lib64"; rename "lib64.new", "/lib64"' (You can't use bash or it will gets broken after unlink /lib64) Hope this won't break anything else. Regards, Jan On Thu, 9 May 2024 10:30:36 +0200 Jan Krčmář <honza...@civ.zcu.cz> wrote: > Thanks for your investigation. There is a buster running as a base system. > I didn't notice that the libcrypt1 does not come from buster earlier. > > This situation is probably caused by newer openafs-client/bullseye package, > which depends on libcrypt1/bullseye, which definitely breaks libc6/buster. > There were several bugs in the openafs some time ago, so the decision was > made to switch to the bullseye version, ending with weird distribution state. > > Removing the libcrypt1 package resolves the the problem. > > Cheers, > Jan > > On Mon, 6 May 2024 20:00:10 +0300 > Adrian Bunk <b...@debian.org> wrote: > > > On Mon, May 06, 2024 at 05:37:37PM +0100, Adam D. Barratt wrote: > > > On Mon, 2024-05-06 at 13:02 +0200, Jan Krčmář wrote: > > > > Package: libc6 > > > > Version: 2.28-10+deb10u3 > > > > > > > > Upgrading the system (Debian 10/Buster) causes corrupted system, > > > > ending with kernel panic and unbootable system. > > > > > > > [...] > > > > The following packages will be upgraded: > > > > apt apt-transport-https apt-utils base-files ca-certificates > > > > > > The fact that APT is being upgraded here seems strange - APT hasn't > > > changed in buster for 3 years. What's your base system? > > > > There's also > > > > Unpacking base-files (10.3+deb10u13) over (10.3+deb10u5) ... > > > > That's outdated since September 2020. > > > > > > Unpacking libc6:amd64 (2.28-10+deb10u3) over (2.28-10+deb10u2) ... > > > > Replaced by files in installed package libcrypt1:amd64 (1:4.4.18-4) > > > > ... > > > > > > This, on the other hand, looks like you've done something odd to your > > > system. libcrypt1 doesn't exist until bullseye, so at some point you > > > have partially upgraded your base system. In conjunction with your pre- > > > upgrade system apparently having an APT version that's /older/ than the > > > one in buster, this feels odd. > > > > There is related nastiness with partial upgrades from buster to > > bullseye, see the last two items at [1]. > > > > I am not sure whether it is the root cause here, but it's at least > > plausible that not being able to have the Breaks in libcrypt1 causes > > problems for such partially upgraded systems. > > > > > Regards, > > > > > > Adam > > > > cu > > Adrian > > > > [1] > > https://tracker.debian.org/news/1239066/accepted-libxcrypt-14418-3-source-into-unstable/ > >
smime.p7s
Description: S/MIME cryptographic signature
