Continuation of thread about startx failing to close
Summary so far.
I have a problem with startx.
Initially I missed out the step in the beginners guide for testing x and went straight to installing xfce. XFCE does not load and I am left with a blank screen and nothing to do but reboot. I have gone back a few steps to testing x and that seems to work as expected, a couple of terminals and a clock. Following the instructions in the unofficial beginners guide I type exit into the terminals on the last one the screen blanks and does not return to a text page, once again the only option is hard reset.
Log files are empty of xorg messages and .xinitrc does not exist yet. Until the vanilla xwindows works there is no point in chasing the xfce install, or so I think.
Is there any other information I can provide?
The beginners guide does say to run startx without an .xinitrc file, it even suggests removing the .xinitrc file if it is there.
I have read the xfce page several dozen times, it states early on that startx must be installed and configured correctly first, which is logical. I think there is something wrong with startx as it causes the system to hang when trying to close it.
Running startxfce4 from the cli or creating .xinitrc file with the start command in it does not work as expected, it switches to a blank screen and system hangs totally, no option other than a hard reset.
Running as a user with xterm in .xinitrc brings up the xterm and only freezes after I type 'exit' into that xterm.
Sorry to say once it freezes it is totally unresponsive, I cannot get back to TTY1 and any error messages.
Last edited by weedfreak (2013-01-18 18:10:40)
Weedfreak, it may be good for your to edit your first post and (while keeping the link) summarize the problem and the tests so far.
Do you also have no Xorg logs after the lock up? How do you "break" the lockup, do you have to do a hard shutdown?
If you have no logs after the lock up, one thing that might give some information is to start x as in the last test (minimal ~/.xinitrc with just "xterm", start from tty1 as a regular user with `xinit`) except now append a redirection on the command line when you launch x:
xinit >~/my.log 2>&1
Then when X starts and you have a working xterm you can cat or tail -f the log file to at least get the log information up until the lock up.
Point taken about summary.
The only way to break the lockup is a hard shutdown or hard reset.
I am currently moving the system to be next to my working box to enable easier testing on my part. I will report back on the next test as soon as I can.
Couple of things to try in .xinitrc:
Explanations in the xinitrc wiki page.
I have now managed to get to Xorg.0.log.
Near the end I have some EE errors.
NVIDIA(0) Failed to initiaize DMA
NVIDIA(0) Error recovery failed.
Fatal server error:
Failed to recover from error!
running xinit >~/my.log 2>&1 initializes a load of extensions up to DRI2 then loads GLX, NV-GLX, NV-CONTROL and XINERAMA.
Rebooting after exiting from xinit lockup and rechecking my.log shows lines connection to x server lost and waiting for xserver to shut down ......
Leaving the system for 10 minutes does not change the last line, the xserver is not shutting down.
I would try replacing nvidia with the nouveau driver. Or you could boot with an Ubuntu Live CD and see if the DE loads up correctly at all.
The system has been running various Linux flavours for a couple of years with xfce and nvidia. In fact last week it was running Arch with xfce so the hardware should not be the problem. The only reason I started the install again was because of a need to change the / disk. I would for several reasons prefer to keep the nvidia driver but if the consensus is to try nouveau then I will.
Just found a xubuntu live cd and that loads ok, no problems with DE or shutdown.
Last edited by weedfreak (2013-01-18 09:15:43)
I went back a couple of steps and have reinstalled Xorg and nouveau.
Xinit now fails to start.
EE errors from Xorg.0.log show
[drm] KMS not enabled
No devices detected
Fatal server error:
no screeens found
From what I see KMS is supposed to be enabled by default.
I am not sure if this new error is connected with the nvidia error.
Are you sure you followed all steps for moving from nvidia to nouveau (there's an article on the wiki)? It sounds like you might have something weird in /etc/mkinitcpio.conf that's not quite right, or perhaps you forgot to create a new linux image (all in the wiki).
If the system used to work with nvidia you probably messed up the nvidia configuration, or an update might have broken the module for you. So if you want to keep nvidia try downgrading to something that used to work, configuring it right so that it runs and then upgrading to the newest version. Otherwise, nouveau works just the same in my humble opinion, so maybe just switch to it.
yes quite correct I borked the linux image creation all now works with nouveau so I will not retry nvidia.
thank you all.
You should be able to get nvidia working if you want. But now at least you know that the original problem lays somewhere in the configuration of nvidia.