Bug#915261: SysV init regression thanks to broken "container detection" logic
Package: udev Version: 239-14 Followup-For: Bug #915261 I tried to set severity to serious but it was reverted. An unbootable system is not nice, I had to look around for a rescue disk (luckly I still had one around) to change the script (it was obvious by message, and also I roughly remembered the etckeeper diff). It could be a bit more troublesome if combined (no other local machines to help, no USB sticks, issues in remote machines, failures without output...). ---8<--- if [ -w /sys ]; then log_warning_msg "udev does not support containers BUT BUG WORKAROUND" # exit 0 fi --->8--- Bug 915415 is about losing X11 mouse and keyboard, which is bad too. Bug 915361 was reported as grave. Avoiding more people to get into issues when cause is known seems reasonable, it saves everyone time. I don't mind finding new bugs and help solving them, but getting hit by old ones like this is discouraging. Confused about apt-listbugs purpose, GSR
Bug#915261: SysV init regression thanks to broken "container detection" logic
Package: udev Version: 239-14 Followup-For: Bug #915261 Severity: serious Severity should be at least serious, for two reasons. First so it gets listed with apt-listbugs default levels and people can decide what to do. And second, it does not affect every user, but those affected have to handle a broken boot, which is not nice. Per https://release.debian.org/testing/rc_policy.txt it is release critical. ---8<--- * makes unrelated software on the system (or the whole system) break --->8--- GSR
Bug#915261: SysV init regression thanks to broken "container detection" logic
Control: tags -1 pending On Sun, Dec 2, 2018 at 10:51 AM Lorenz wrote: > I confirm the bug. This prevent me to start/restart udev, and i'm not in a > container. > > On Sun, 2 Dec 2018 09:00:07 + Jamie Heilman < > ja...@audible.transient.net> wrote: > > The logic change in the init script to check for a writable /sys is > > backwards; udev *should* start if /sys if writable. > Indeed, it is backwards. Thanks for reporting. I have queued up the change in git: https://salsa.debian.org/systemd-team/systemd/compare/2f6a08f9...c05e4e13 BTW, patches adding autopkgtests that exercise the init script would be welcome. -- Saludos, Felipe Sateler
Bug#915261: SysV init regression thanks to broken "container detection" logic
I confirm the bug. This prevent me to start/restart udev, and i'm not in a container. On Sun, 2 Dec 2018 09:00:07 + Jamie Heilman wrote: > The logic change in the init script to check for a writable /sys is > backwards; udev *should* start if /sys if writable. I don't have experience with container, so can't tell if the logic is backwards, it looks to me that /sys directory is read/execute only in my Debian, but 'test -w' performed by root exit 0.. # cat /home/ombra/udevtest.txt Script started on 2018-12-02 14:36:45+01:00 root@lorenz:~# ls -l / | grep sys dr-xr-xr-x 12 root root 0 Dec 2 14:20 sys root@lorenz:~# test -w /sys root@lorenz:~# echo $? 0 root@lorenz:~# exit exit Script done on 2018-12-02 14:37:37+01:00 I suspect this is breaking the boot sequence in non-systemd installs, can you please revert to using "ps" for this test untill you sort it out? Thanks, Lorenzo
Bug#915261: SysV init regression thanks to broken "container detection" logic
Package: udev Version: 239-14 Severity: important The logic change in the init script to check for a writable /sys is backwards; udev *should* start if /sys if writable. -- Jamie Heilman http://audible.transient.net/~jamie/