Been running 252.6 since forever with custom suspend hooks in /lib64/elogind/system-sleep. After update to 255.5 I noticed that these were no longer invoked. As it turns out the configured hook directory has changed to /usr/lib/elogind/system-sleep without notice. This is especially confusing in light of related bugs: https://bugs.gentoo.org/923258 https://bugs.gentoo.org/939042 Reproducible: Always Steps to Reproduce: 1. update from elogind-252.6 to 255.5 2. fix suspend config (#939042) 3. perform suspend/resume Actual Results: Suspend/resume hooks are no longer invoked.
Sorry, list of related bugs should be: https://bugs.gentoo.org/693384 https://bugs.gentoo.org/939042
ftr there's plans[1] to have nvidia-drivers install its own hook now, it'll be done in both directory for backward compat until 255.5 is stable for a while (I find the new path a lot better honestly). Help is welcome in testing [1], it's on hold until I get enough feedback for it to work right, can't easily test it myself. That aside users should instead be using /etc/elogind/system-sleep for their hooks rather than system paths though. [1] https://github.com/gentoo/gentoo/pull/38482
As discussed in the nvidia-drivers PR I can confirm that with 255.5 either /usr/lib/elogind/system-sleep or /etc/elogind/system-sleep for suspend hooks works.
It probably wouldn't hurt to at least have a ewarn/log message before stabilizing this version, message could suggest /etc/elogind/system-sleep Either once w/ ver_test, or always if /lib64/elogind/system-sleep/* has files. Albeit the aforementioned nvidia PR will install a single temporary "nvidia" symlink in there which users shouldn't be warned about.
Why did you link to bug 923258 which is 252.9 stabilisation?
(In reply to Andreas Sturmlechner from comment #5) > Why did you link to bug 923258 which is 252.9 stabilisation? Just general Saturday morning confusion.
Some personal notes about elogind-255.5 on ~amd64: Xorg+Elogind+Nvidia-drivers here: sleep finally works back after some struggle with sys-auth/elogind-255.5: - elogind-uaccess-command path have changed and so I have need to fix the kernel initramfs to include the new path of this command (in my case dracut). - the default sleep method have changed from deep to s2idle. I have need to configure elogind to explicitly use deep method to really have a sleeping system. - If I use as usual echo "mem" > /sys/power/state: I got a dmesg error that show that the nvidia hook haven't been used and so sleep don't work this way any more. I have to use the elogind way: loginctl suspend that is working.
