KDE and plasma5 crashes

… with the end (longer than 15000 characters)

35347.137] (II) No input driver specified, ignoring this device.
35347.137] (II) This device may have been added with another device file.
35347.137] (II) config/udev: Adding input device HDA ATI SB Rear Mic (/dev/input/event5)
35347.137] (II) No input driver specified, ignoring this device.
35347.137] (II) This device may have been added with another device file.
35347.137] (II) config/udev: Adding input device HDA ATI SB Line (/dev/input/event6)
35347.137] (II) No input driver specified, ignoring this device.
35347.137] (II) This device may have been added with another device file.
35347.138] (II) config/udev: Adding input device Eee PC WMI hotkeys (/dev/input/event3)
35347.138] () Eee PC WMI hotkeys: Applying InputClass “evdev keyboard catchall”
35347.138] (
) Eee PC WMI hotkeys: Applying InputClass “system-keyboard”
35347.138] () Eee PC WMI hotkeys: Applying InputClass “evdev keyboard catchall”
35347.138] (II) Using input driver ‘evdev’ for ‘Eee PC WMI hotkeys’
35347.138] (
) Eee PC WMI hotkeys: always reports core events
35347.138] () evdev: Eee PC WMI hotkeys: Device: “/dev/input/event3”
35347.138] (–) evdev: Eee PC WMI hotkeys: Vendor 0 Product 0
35347.138] (–) evdev: Eee PC WMI hotkeys: Found keys
35347.138] (II) evdev: Eee PC WMI hotkeys: Configuring as keyboard
35347.138] (
) Option “config_info” “udev:/sys/devices/platform/eeepc-wmi/input/input6/event3”
35347.138] (II) XINPUT: Adding extended input device “Eee PC WMI hotkeys” (type: KEYBOARD, id 8)
35347.138] () Option “xkb_rules” “evdev”
35347.138] (
) Option “xkb_model” “microsoftpro”
35347.138] () Option “xkb_layout” “gb”
35347.138] (
) Option “xkb_variant” “basic”
35347.138] (II) config/udev: Adding input device PC Speaker (/dev/input/event2)
35347.138] (II) No input driver specified, ignoring this device.
35347.138] (II) This device may have been added with another device file.

Try x11vnc. It is old, but is feature-rich and works in a server computer with nvidia card and nvidia driver.

Hmmm… I realized that there were some error lines in the last Xorg log, AGAIN… I missed them.
I will investigate them an wait for help.

Your nvidia driver installation is broken:

You use/have installed Xorg’s GLX module, nvidia provides its own (which breaks Mesa’s OpenGL support) and doesn’t work without it.
Reinstall the driver.

Maybe an update to Xorg overwrote libglx with Xorg’s version again, which nvidia overwrote in the first place.

This is actually a “flaw” in nvidia’s installer, which is being workarounded in openSUSE’s nvidia driver packages, but those don’t exist for Tumbleweed.

Sorry, but those are the things you have to live with if using Tumbleweed (i.e. reinstall the driver often), and that’s why it’s actually not recommended to use Tumbleweed if you rely on 3rd party drivers like nvidia.

35346.922] (EE) AIGLX: reverting to software rendering
35346.942] (EE) AIGLX error: dlopen of /usr/lib64/dri/swrast_dri.so failed (/usr/lib64/dri/swrast_dri.so: undefined symbol: _glapi_tls_Dispatch)
35346.942] (EE) GLX: could not load software renderer
35346.942] (II) GLX: no usable GL providers found for screen 0

And that shows you that Mesa’s software renderer doesn’t work, because nvidia replaced Mesa’s libGL with its own incompatible version.
Reinstalling Mesa (Mesa-libGL1 in particular) might fix that and even make Plasma5 over VNC work, but then nvidia’s GLX or hardware accelerated OpenGL won’t work.

Dear wolfi323,

If the nvidia stole or use open-source programs without proper permission, go to the court. It is a legal problem.
If above AND no eligible license defense, make boycott for nvidia OpenGL solutions and declare it to the application developers… it is an ethics problem.

BUT if it is an informatics problem also, do not punish the average users.

In our case (nvidia cards) you should disable the complete OpenGL and Mesa stolen software if the videocard is hardware capable AND the user wants it. From this point it becomes to nvidia and user responsibility, if the program crashes. These are already 10 year old errors… it your turn.

I realized it today morning. It was the 5. kernel update during 10 days… the program worked before.
And I say so sloooowly as possible: I need nvidia remote solution with hardware rendering. Not me, but for the applications.

The 352.63 works without errors… the 361.18 beta was your advice and I did not downgrade after this :frowning: :frowning:
Would you like to see the result?

It was not my “advice”, it was a suggestion to try.

The bug report I mentioned earlier suggests that some manual intervention is required after the beta driver installation though, apparently nvidia’s beta installer does not correctly install some libraries.

Would you like to see the result?

Not necessary if it works now.

It is not my legal problem.

And I never said that “nvidia stole or use open-source programs without proper permission”.
But again, they replace system libraries with their own versions that only work with their driver and break other things.

If above AND no eligible license defense, make boycott for nvidia OpenGL solutions and declare it to the application developers… it is an ethics problem.

Personally, I do not use nvidia anyway (neither cards nor the driver).

BUT if it is an informatics problem also, do not punish the average users.

Who is “punishing” anybody?
If at all, it’s nvidia for not playing game with open source.

In our case (nvidia cards) you should disable the complete OpenGL and Mesa stolen software if the videocard is hardware capable AND the user wants it. From this point it becomes to nvidia and user responsibility, if the program crashes. These are already 10 year old errors… it your turn.

Again, nothing is “stolen”.
And your problem falls back to the fact that nvidia’s driver doesn’t support software rendering, and the system’s software renderer (part of Mesa) is broken by nvidia’s installer because it replaces system libraries.

Even if you don’t like to hear it:
Use nouveau, that does support OpenGL/GLX too, via Mesa.
If that’s not an option for you, complain to nvidia.
And the nvidia driver does work fine in openSUSE on real (nvidia) hardware as you probably noticed already, but not via VNC. And it breaks other stuff (Mesa), that would work.
Maybe there is a way to make nvidia work via VNC, but I don’t know anything about that.

wolfi323, drink a glass of beer, and do not be so aggressive. It is not your problem. Moreover, you do not understand it, as I see.

Hm?
Where was I aggressive? TBH, I rather see (some of) your posts as being aggressive…

I just tried to address your (IMHO incorrect) statements, especially regarding legal problems and “things stolen”, which this has nothing to do with.

It is not your problem.

Right.

Moreover, you do not understand it, as I see.

Obviously.
You don’t seem to understand most of what I write either.

Anybody who has working hardware accelerated nvidia driver?
(libglx.so point to installed nevidia driver AND works)

I suppose most people who install the nvidia driver do.
Those that do not, come to ask for help anyway…

I thought it works again for you too again meanwhile, no?
After all, you posted this yesterday:

But again, this is only going to work on the hardware directly, not via VNC.

I followed your KDE forum thread, and luebking wrote this:

The nvidia blob doesn’t support either environment. You render directly and in hardware.

=> afaik only x11vnc will work (as bonus you get HW accelerated, direct rendering), tight/tigervnc won’t

The default VNC server in openSUSE is tigervnc.

So try to install x11vnc and experiment with that. It is included in openSUSE, just not installed/used by default.
Or have you tried that already?

Btw, you wrote this in the KDE thread :

I do not know why opensuse operator forces Mesa and noveau when it is absolutely unnecessary in my case.

The problem is not that openSUSE forces Mesa and nouveau on you, but that nvidia does not support indirect rendering or software rendering. See luebking’s post.
Why do you not want to accept that?

The x11vnc WORKS (with ssvnc and realvnc),

Thank you, ZStefan and wolfi323.
x11vnc WORKS with client-side ssh tunneled realvnc andd ssvnc, glxinfo answers.
But there are some errors:

  • geometry problems
  • Wakfu starts with blank screen, Kodi with “segmentation fault”
  • Plasma5 starts also with one screen, then after disconnect the desktop remains to it by plasmashell…
  • if the connection lost, the server crashes sometimes.

I will investigate the options and the applications, but it is not OS or desktop problem yet, I think.