Bug#533009: lib32ffi5: missing Conflicts: libc6-i386 (<= 2.9-18)
Control: found -1 3.0.9-3
Hi,
I just noticed the
Conflicts: libc6-i386 (<= 2.9-18)
that was supposedly added in 3.0.7-2 has disappeared later on again,
causing incorrect updates from lenny to squeeze by installing the
library over the /usr/lib32 -> /emul/... symlink before that gets
replaced by a directory. So the library actually ends up in /emul/...
instead of /usr/lib32.
It's probably too late to fix this in squeeze (as most of the
lenny->squeeze updates should have been done long ago), the package
should still cleanup the mess it created in /emul, and ideally do this
in wheezy (I don't think this would qualify for s-p-u):
In the postinst:
rm -f /emul/ia32-linux/usr/lib/libffi.so.5
rm -f /emul/ia32-linux/usr/lib/libffi.so.5.0.10
rmdir /emul/ia32-linux/usr/lib \
/emul/ia32-linux/usr \
/emul/ia32-linux \
/emul \
2>/dev/null || true
should do the job.
>From a lenny->squeeze piuparts distupgrade test:
...
Preparing to replace lib32ffi5 3.0.7-1 (using .../lib32ffi5_3.0.9-3_amd64.deb) ...
Unpacking replacement lib32ffi5 ...
dpkg: considering deconfiguration of lib32ffi-dev, which would be broken by installation of libc6-i386 ...
dpkg: yes, will deconfigure lib32ffi-dev (broken by libc6-i386).
Preparing to replace libc6-i386 2.7-18lenny7 (using .../libc6-i386_2.11.3-4_amd64.deb) ...
De-configuring lib32ffi-dev ...
Unpacking replacement libc6-i386 ...
Preparing to replace lib32ffi-dev 3.0.7-1 (using .../lib32ffi-dev_3.0.9-3_amd64.deb) ...
Unpacking replacement lib32ffi-dev ...
dpkg: warning: unable to delete old directory '/emul/ia32-linux/usr/lib': Directory not empty
dpkg: warning: unable to delete old directory '/emul/ia32-linux/usr': Directory not empty
dpkg: warning: unable to delete old directory '/emul/ia32-linux': Directory not empty
dpkg: warning: unable to delete old directory '/emul': Directory not empty
...
1m3.5s ERROR: FAIL: debsums reports modifications inside the chroot:
debsums: missing file /usr/lib32/libffi.so.5.0.10 (from lib32ffi5 package)
Andreas
Reply to: