kfreebsd-i386 d-i/squeeze FTBFS (was Re: Bug#633561: pu: package kfreebsd-8/8.1+dfsg-8+squeeze1)
[tl,dr; these changes broke d-i in stable]
On Sat, 2011-10-01 at 00:25 +0200, Robert Millan wrote:
> 2011/9/27 Adam D. Barratt <firstname.lastname@example.org>:
> >> > - Does this affect which modules end up in the udebs?
> > It looks like this was missed in the original follow-up. As a related
> > query, has a test build of kfreebsd-kernel-di-* been performed in order
> > to find out whether any additional modules get pulled in to the
> > packages?
> I just tried. The resulting file lists are identical.
debian-installer was binNMUed earlier today in preparation for the point
release, and FTBFS on kfreebsd-i386. The log finishes with:
# Move the kernel image out of the way.
mv -f ./tmp/cdrom/tree/boot/kfreebsd.gz ./tmp/cdrom/kfreebsd.gz;
test -e ./tmp/cdrom/tree/boot/zfs || rmdir ./tmp/cdrom/tree/boot/
rmdir: failed to remove `./tmp/cdrom/tree/boot/': Directory not empty
make: *** [stamps/tree-unpack-cdrom-stamp] Error 1
The problem appears to be that the contents of the kernel udeb have
$ debdiff kernel-image-8.1-1-486-di_0.6_kfreebsd-i386.udeb kernel-image-8.1-1-486-di_0.6+b1_kfreebsd-i386.udeb
Files in second .deb but not in first
-rw-r--r-- root/root /boot/acpi.ko
This appears to be due to the fact that the new kfreebsd-8 +squeeze1
kernel on -i386 builds acpi.ko as a module, which is then picked up by
existing "install this module if it exists" rules in kernel-wedge.
A quick fix would be appreciated, whether from the kfreebsd or d-i side,
given the increasingly short period of time we have remaining until the
point release is scheduled.