--set-startup won't work about optimus-manager HOT 10 CLOSED

It's merged. Going to add a few more fixes and features and then publish a new version.

from optimus-manager.

@kuba1909 Could you please test this branch ? https://github.com/Askannz/optimus-manager/tree/pre_post_exec

I made you a PKGBUILD : https://pastebin.com/XqtL7445

This fix works on my end and is a much much much better approach.

EDIT : Sorry, I created the branch but forgot to push to it 🤦‍♂️ It's good now.

from optimus-manager.

I built the package with your PKGBUILD, it works perfectly now. 😄

from optimus-manager.

Same bug: #3

from optimus-manager.

Same bug: #3

Actually it isn't. If you take a closer look at the systemd log, you can see that the script doesn't even recognize my lightdm setup. In contrast to #3.

from optimus-manager.

you can see that the script doesn't even recognize my lightdm setup.

What do you mean ? The log does mention
Nov 20 17:36:26 arch python[567]: Configuring LightDM

If you are referring to
Nov 20 17:35:59 arch python[567]: Error : Cannot find /usr/share/sddm/scripts/Xsetup
it's a bug that slipped in the latest release. Should not affect your problem.

Could you post the full log (with journalctl -u optimus-manager -b0) instead of the unit status ? There could be something happening early. In fact, it may be related to a similar bug I'm investigating on my setup.

Thank you for your patience ; I can't dedicate too much time to the project this week but I'm definitely working on it.

from optimus-manager.

Good news : I'm 90% sure to have found the source of the problem. And the problem is that I am an idiot.

I configured the optimus-manager systemd service to start before the login manager, so it can configure Xorg and kernel modules in advance. What I didn't realize is that it gives no guarantee that the daemon will have finished its job before the X has started.

So this leads to this weird race condition where sometimes the X server will use the autogenerated config, sometimes not. It may depend on your hardware and that's why I couldn't reproduce it. I can reproduce it now by adding an artificial delay to the daemon.

I have a fix in mind for that but it will have to wait for a bit.

from optimus-manager.

you can see that the script doesn't even recognize my lightdm setup.

What do you mean ? The log does mention
Nov 20 17:36:26 arch python[567]: Configuring LightDM

The log starting from Nov 20 17:36:26 is not related to --set-startup nvidia but --switch nvidia. Thanks for your support. I will try 812f3b9 soon. Btw, journalctl -u optimus-manager -b0 gives me the same log.

from optimus-manager.

The commit 812f3b9 was an attempt at fixing the problem, but it's not really functional yet (hence why it's in a separate branch). I got a better idea that's more in line with the philosophy of systemd, I'll try to implement it asap.

from optimus-manager.

Great ! Gonna merge it then.

from optimus-manager.