Closed adriaandegroot closed 6 years ago
I flashed, got the same behaviour.
Then rebooted and it went straight to plasma Then again and again and it kept working.
Can either of you figure out why this happens? :D I can't do much without pinebook here.
I can't do much without pinebook here.
That seems like a valid excuse.
It's going to be hard to figure out given it means flashing from scratch each time, but I'll give it a go
@Bhushan: any chance to get a pinebook to your place? Ebay? Are you coming to the sprint?
Um, it's terribly hard to get pinebook from here, on ebay I had found an listing : http://www.ebay.in/geb/ImportHubViewItem?itemid=263140303914 but it was 5X cost and when I found it it was already sold out. I had also found a listing on .co.uk but it didn't ship to India and had similar price as one shown above..
I am coming to sprint yes.
It's going to be hard to figure out given it means flashing from scratch each time, but I'll give it a go
plasma crashing only during the initial javascript setup?
will download the image and try if just deleting appletsrc repeats the issue
plasma crashing only during the initial javascript setup?
No, it doesn't even reach plasma.. somehow just SDDM bails out..
is the home existing and complete at that point?
What we know
sddm doesn't start
getty@tty1 starts
sddm is supposed to start at tty1
If we add nottyautologin to the command line it doesn't start the sddm, and also doesn't start autologin we get empty blank screen
There is code in /live/live/config/00??-systemd to enable autologin on tty, removing that code have no effect
Bunch of interesting info from #systemd
18:13 <dreisner> bshah: perhaps graphical.target isn't the default for the first boot.
18:14 <bshah> if I check systemctl get-default , it's graphical.target
18:14 <dreisner> bshah: i'm suggesting that maybe something during the boot process changes the default on first boot, but the actual first boot target is NOT graphical.
18:15 <bshah> ah.. that might be quite possible
I am yet to verify if this is case
Doing systemctl enable sddm in image also doesn't have any effect
For the sddm issue. Just because I just saw it here and it does not come up on the rock64 aswell without it: https://github.com/netrunner-zeronet/rock64-live-build-zeronet/blob/master/customization/hooks/live/02-add-sddm-to-video.hook.chroot
Make sure the live user is part of the video group.
Some more case checking. I tested that there is no /etc/systemd/system/default.target symlink. I've tried adding manual systemctl set-default graphical.target in hook... let's see.
So at end of day.. current investigation
ExecStartPre=/bin/sh -c '[ "$(cat /etc/X11/default-display-manager 2>/dev/null)" = "/usr/bin/sddm" ]'
from sddm.service ... didn't work
Some response:
2018-05-03 16:13:06 <grawity> May 03 10:13:22 localhost.localdomain systemd[1]: sddm.service: Fixing conflicting jobs sddm.service/start,sddm.service/stop by deleting job sddm.service/stop
2018-05-03 16:13:15 <grawity> that was a quick grep.
2018-05-03 16:14:02 <grawity> you've got a dependency loop, involving sddm + plymouth, and they're resolved by skipping a random job
2018-05-03 16:14:18 <grawity> May 03 10:15:13 live-pc systemd[1]: sddm.service: Job sddm.service/start finished, result=canceled
2018-05-03 16:14:19 <grawity> May 03 10:15:13 live-pc systemd[1]: sddm.service: Installed new job sddm.service/stop as 225
After I spent some more time grawity noticed something else...
2018-05-03 18:31:25 <grawity> bshah: more importantly than the loop, it seems that something in your boot process activates plymouth-quit-wait.service
2018-05-03 18:31:36 <grawity> bshah: and sddm directly conflicts with that
2018-05-03 18:32:17 <bshah> oh
2018-05-03 18:32:53 <grawity> actually, I think that's what causes *all* messages, it's not a loop
Later...
2018-05-03 18:43:02 <bshah> oh! ./lib/systemd/system/multi-user.target.wants/plymouth-quit-wait.service (!!!) could it be a thing triggering plymouth-quit-wait.service grawity ?
2018-05-03 18:43:32 <grawity> bshah: yes
2018-05-03 18:49:01 <bshah> grawity: so the way I see is break sddm depends/conflicts to plymouth-quit-wait.service ? since it depends on multi-user.target and ubuntu/bionic multi-user.target.wants already depends on plymouth-quit-wait.service (symlink provided by plymouth package).
2018-05-03 18:49:46 <grawity> tbh, I'm not 100% sure
2018-05-03 18:50:16 <grawity> what I seemed to find in the logs was that the problem kinda happened *twice* during the same boot
2018-05-03 18:50:53 <grawity> conflict gets resolved somehow, boot continues, and suddenly another conflict causes sddm to be cancelled
2018-05-03 18:52:16 <bshah> another problem might be getty service I think
2018-05-03 18:52:22 <bshah> there is conflict on it too in sddm service
2018-05-03 18:52:58 <grawity> or it could be that, yeah... but
2018-05-03 18:53:14 <grawity> well, now that I think of it, somehow it works fine with gdm
2018-05-03 18:53:37 <grawity> so maybe Conflicts= was designed to be deterministic, I guess?
2018-05-03 18:54:00 <grawity> if both are queued at the same time, sddm wins
2018-05-03 18:54:15 <bshah> annoying thing is this breaks ONLY on first boot..
2018-05-03 18:54:20 <grawity> but if sddm is already starting, and something tries to start plymouth (or getty)
So I spent some time trying to disable plymouth dep from the sddm service... which also failed..
meantime i found out that plymouth is not even used in the kernel.. as there is no quiet splash option in kernel command line.
For better debugging i added following to kernel cmdline
console=tty0 systemd.log_level=debug systemd.log_target=kmsg
I am not sure what to do next, but perhaps sleeping over some ideas will give some more thought on what might be wrong.
Okay so I tried something and it seems to be "working" , working in double quotes because plasma is super slow at startup for some reason... BUT..
Things now load instead of completely fail to load, and we have idea of what is wrong...
Optional (if you want to watch logs.. as this sometimes get stuck doing live-config job forever.)
Edit : new link to 0170-systemd ; https://ptpb.pw/PbAm
Try #3 to get rid of long delay : https://ptpb.pw/j2i_
Booted first time \o/
https://files.kde.org/neon/images/pinebook-remix-nonfree/bionic/20180504-1120/
This image fixes it..
Appearantly we were shipping some file which was not needed and was conflicting with something else doing exactly same thing.. https://github.com/pangea-project/pangea-tooling/commit/a7735f01c3ae8951a1ed858bdabfeee5f9e1bd6d which had bugs.. my last attempt was to fix bug but realized this code is no longer needed at all..
Keep in mind tha plasma startup is super slow for some reason and needs debugging, but it is seperate issue.
Latest image (0404-1031) went to text login after boot, not to plasma desktop. From text prompt,
sudo sddm
started the desktop as expected.Previous image (0404-0549) went to plasma desktop once boot was fixed.