Just to send a "me, too!", I also recently bumped udev to the systemd version and also had a non-booting machine.
Now, most of my machines had a sufficiently well-populated static /dev that I could boot and actually didn't notice the lack of udevd. On the non-booting machine, it was getting somewhere into the init scripts, but I noticed some "command line syntax error" complaints from mount when I hit scroll lock at the appropriate time. Fortunately, I managed to boot with init=/bin/bash and MAKEDEV or mknod the necessary device nodes. So now it's running without udev either. I spent a long time digging through /etc/init.d/mountkernfs trying to find the problem before I came across this bug. The question before me is whether to proceed to fix udev or just remove the PoS from my ststem. I've been running Debian on this machine with hand-compiled kernels since the late 1990s, updating unstable at least weekly, and this is the first time an update has failed to boot to at least a single-user shell. I've had sshd break, which was pretty bad, but this is the worst. The initial introduction of udev didn't cause this kind of mess. When an essential early-boot utility is adding a kernel config dependency, I expect *prominent* notice in changelog and the NEWS file. But I notice, despite the reference to it in the 204-1 changelog.Debian entry, no such file is packaged with udev. And finding it in the systemd source tree, there's no mention of the new requirement there. Hunting through the systemd git tree logs, I find that the requriement was added in commit 220893b3cbdbf8932f95c44811b169a8f0d33939 (udev version 176), and the old udev NEWS file which documented it was deleted in commit 3e2147858f21943d5f4a781c60f33ac22c6096ed. Gosh, maybe that could be hidden more thoroughly? Because Debian jumped from 175-7.2 to 204, there was literally no opportunity to see it. (This is primarily upstream's fault. While it would have been highly deisrable for the Debian maintainer to notice and reinstate things, this cavalier destruction of important history is reprehensible. Cc: to Kay Sievers to vent my ire in the correct direction.) The init.d/udev problem causing the mount syntax errors appears to be someone using `-o $dev_mount_options` rather than `-o "$dev_mount_options"`, which works fine with a null options string. -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
