The update to udev-181 broke my system (sys-process/minit) because udevd moved from /sbin to /lib/udev without warning. Please create a symlink in /sbin in the ebuild. See NEWS and http://marc.info/?l=linux-hotplug&m=131835335207882&w=2 Thanks. Reproducible: Always
uh, no. minit should be fixed to not look for udevd in /sbin
(In reply to comment #1) > uh, no. minit should be fixed to not look for udevd in /sbin Minit was just an example. Every script that expects udevd in /sbin is broken right now. (In my case the symptoms were a non working mouse and keyboard in X. Fun thing to debug...) And all this because moving things around without any reason whatever seems to be en vogue right now.
The symlink would just be "a bandage" while the reverse dependencies should be fixed to not hardcode the path to udevd (or at least look for both /sbin and /usr/sbin) I can see both points but the correct way to go with this one isn't necessarily the easiest.
(In reply to comment #2) > (In reply to comment #1) > > uh, no. minit should be fixed to not look for udevd in /sbin > > Minit was just an example. Every script that expects udevd in /sbin > is broken right now. Like which ones? > (In my case the symptoms were a non working mouse and keyboard in X. > Fun thing to debug...) > And all this because moving things around without any reason whatever > seems to be en vogue right now. I just checked minit source tree and there is no reference to udev there whatsoever.
(In reply to comment #4) > (In reply to comment #2) > > (In reply to comment #1) > > > uh, no. minit should be fixed to not look for udevd in /sbin > > > > Minit was just an example. Every script that expects udevd in /sbin > > is broken right now. > > Like which ones? The one I wrote to handle udevd in minit. > > (In my case the symptoms were a non working mouse and keyboard in X. > > Fun thing to debug...) > > And all this because moving things around without any reason whatever > > seems to be en vogue right now. > > I just checked minit source tree and there is no reference to udev there > whatsoever. I've switched to mdev today and really don't care about udev anymore.
*** Bug 446372 has been marked as a duplicate of this bug. ***
Starting from udev-197-r3, -r4 and 9999 the files are now installed like this: -rwxr-xr-x 1 root root 194640 Jan 19 15:41 /sbin/udevd lrwxrwxrwx 1 root root 11 Jan 19 15:41 /usr/lib/systemd/systemd-udevd -> /sbin/udevd Old and new location. The reason why we did this now is that /lib/systemd path doesn't make any sense and nobody would have accepted patches for such directory. I hope this satisfies all of these requests now. Let me know of any problems. Thanks!
Would it be too much to ask to apply the following patch to /etc/init.d/udev? This is needed in my setup because /sbin/udevd is a symlink to ../mnt/livecd/sbin/udevd. This is a livecd setup created from genkernel where all files in "/" are symlinks to files in /mnt/livecd, which is a squashfs looped FS created using 'lndir'. This setup allows modifications to the rootfs in case of emergencies because symlinks are not at "/" level. I am not sure why that check with ! -L was put in there in the first place. So, please assess if its needed. --- /etc/init.d/udev 2013-02-09 23:20:37.203298266 -0800 +++ /etc/init.d/._cfg0000_udev 2013-02-10 09:16:10.000000000 -0800 @@ -59,7 +59,7 @@ bins="/sbin/udevd /lib/systemd/systemd-udevd /usr/lib/systemd/systemd-udevd" for f in ${bins}; do - if [ -x "$f" -a ! -L "$f" ]; then + if [ -x "$f" ]; then command="$f" fi done
(In reply to comment #8) > Would it be too much to ask to apply the following patch to > /etc/init.d/udev? This is needed in my setup because /sbin/udevd is a > symlink to ../mnt/livecd/sbin/udevd. This is a livecd setup created from > genkernel where all files in "/" are symlinks to files in /mnt/livecd, which > is a squashfs looped FS created using 'lndir'. This setup allows > modifications to the rootfs in case of emergencies because symlinks are not > at "/" level. > > I am not sure why that check with ! -L was put in there in the first place. > So, please assess if its needed. > > --- /etc/init.d/udev 2013-02-09 23:20:37.203298266 -0800 > +++ /etc/init.d/._cfg0000_udev 2013-02-10 09:16:10.000000000 -0800 > @@ -59,7 +59,7 @@ > > bins="/sbin/udevd /lib/systemd/systemd-udevd > /usr/lib/systemd/systemd-udevd" > for f in ${bins}; do > - if [ -x "$f" -a ! -L "$f" ]; then > + if [ -x "$f" ]; then > command="$f" > fi > done that won't work, it would launch udevd with process name systemd-udevd :P need the -L check
(In reply to comment #9) > (In reply to comment #8) > > Would it be too much to ask to apply the following patch to > > /etc/init.d/udev? This is needed in my setup because /sbin/udevd is a > > symlink to ../mnt/livecd/sbin/udevd. This is a livecd setup created from > > genkernel where all files in "/" are symlinks to files in /mnt/livecd, which > > is a squashfs looped FS created using 'lndir'. This setup allows > > modifications to the rootfs in case of emergencies because symlinks are not > > at "/" level. > > > > I am not sure why that check with ! -L was put in there in the first place. > > So, please assess if its needed. > > > > --- /etc/init.d/udev 2013-02-09 23:20:37.203298266 -0800 > > +++ /etc/init.d/._cfg0000_udev 2013-02-10 09:16:10.000000000 -0800 > > @@ -59,7 +59,7 @@ > > > > bins="/sbin/udevd /lib/systemd/systemd-udevd > > /usr/lib/systemd/systemd-udevd" > > for f in ${bins}; do > > - if [ -x "$f" -a ! -L "$f" ]; then > > + if [ -x "$f" ]; then > > command="$f" > > fi > > done > > that won't work, it would launch udevd with process name systemd-udevd :P > need the -L check Why is process name systemd-udevd bad on a system running systemd?
