On Thu, Feb 21, 2008 at 10:15 AM, Roman Kyrylych <roman.kyrylych@gmail.com> wrote:
2008/2/21, Dan McGee <dpmcgee@gmail.com>:
On Thu, Feb 21, 2008 at 9:33 AM, Roman Kyrylych <roman.kyrylych@gmail.com> wrote:
I've just rendered my system unusable after installing udev-118-2 but not the latest initscripts, because initscripts in core check for start_udev, and when it's not available - falls back to static /dev which obviously has only a couple of files, and no /dev/sda*. User has to remount / in rw mode (to make changes to local db possible), then downgrade udev or update initscripts.
This happens only when user upgrades udev without upgrading to initscripts first. I don't know how this should be solved in a clean way. Will a big fat warning in udev's pre_upgrade with 'read' command and Ctrl-C possibility be enought? The situation when upgrading initscripts and using older udev is already covered by depends=(... udev>=118 ...) in initscripts' PKGBUILD.
This is exactly why I wanted to do the udev upgrade a bit more gradually...
start_udev sucks, but we can leave it in for an iteration or two.
I think we could release udev-118-3 with start_udev in it. Older initscripts would use that and don't fail to static /dev, newer initscripts would use udevadm. Then in next version of udev we can safely remove it, I think.
Ugh, good catch. It's annoying that we have to do this, but meh. I don't have a problem with it for a few versions. I will fix this at some point today, and rebuild for x86_64 - I will need someone else to build for i686 though.