Bug 33324 - Mesa from 23.3.3 to 24.0.9 breaks rendering in Nheko
Summary: Mesa from 23.3.3 to 24.0.9 breaks rendering in Nheko
Status: NEW
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: 9
Hardware: All Linux
Priority: Normal normal
Target Milestone: ---
Assignee: Mageia Bug Squad
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2024-06-24 09:26 CEST by Jose Manuel López
Modified: 2024-07-01 01:52 CEST (History)
4 users (show)

See Also:
Source RPM: mesa-24.0.9-1.mga9.tainted,mesa-24.0.9-1.mga9
CVE:
Status comment:


Attachments
Nheko with Mesa 24.09 (54.83 KB, image/jpeg)
2024-06-24 09:33 CEST, Jose Manuel López
Details

Description Jose Manuel López 2024-06-24 09:26:00 CEST
Description of problem: When updating to Mesa 24.09 I noticed that the Nheko application breaks on my Asus Eeepc Intel Atom computer. Texts throughout the app have been lost and now do not appear. 

I have verified that the error comes with the Mesa 24.09 update by following this process: 

- Installing Nheko on a clean installation of Mageia Lxqt and xfce.
- Nheko starts correctly.
- Update of the entire system except for the xorg and mesa packages and reboot of the system.
- Nheko starts correctly.
- xorg update and system reboot.
- Nheko starts correctly.
- Mesa update to version 24.09 and system reboot.
- Nheko breaks down. App texts and chats are now not visible.


Version-Release number of selected component (if applicable): Nheko, Mesa, Mageia9


How reproducible: Install nheko and then update mesa to version 24.09


Steps to Reproduce:
1. Install Nheko
2. Update Mesa 24.09
3.
Comment 1 Jose Manuel López 2024-06-24 09:33:30 CEST
Created attachment 14577 [details]
Nheko with Mesa 24.09

This is the image of Nheko running in my computer, with Intel Atom N450 and Intel graphics, and Mesa 24.09.
Comment 2 katnatek 2024-06-24 20:52:44 CEST
I think this is more a fail in mesa, so set SRPMs

Source RPM: Nheko and Mesa 24.09 => mesa-24.0.9-1.mga9.tainted,mesa-24.0.9-1.mga9

Comment 3 Morgan Leijström 2024-06-25 10:26:11 CEST
Does the problem appear with both the tainted and non tainted flavour of mesa?

What desktop and login manager do you use?

Do the desktop login and Mageia Control Center display correctly?
(SDDM and MCC have seen problems with some mesa on some systems)

Please report on your system; what do
 $ inxi -SMCG
return?


CC last packager

CC: (none) => fri, ghibomgx

Comment 4 Jose Manuel López 2024-06-25 14:18:31 CEST
[jose@localhost ~]$ inxi -SMCG
System:
  Host: localhost Kernel: 6.6.28-desktop-1.mga9 arch: x86_64 bits: 64               
    Desktop: LXQt v: 1.4.0 Distro: Mageia 9                                         
Machine:                                                                            
  Type: Laptop System: ASUSTeK product: 1005PE v: x.x                               
    serial: <superuser required>                                                    
  Mobo: ASUSTeK model: 1005P v: x.xx serial: <superuser required>                   
    BIOS: American Megatrends v: 1202 date: 06/23/2010                              
CPU:                                                                                
  Info: single core model: Intel Atom N450 bits: 64 type: MT cache:                 
    L2: 512 KiB                                                                     
  Speed (MHz): avg: 1171 min/max: 1000/1667 cores: 1: 1009 2: 1333                  
Graphics:                                                                           
  Device-1: Intel Atom Processor D4xx/D5xx/N4xx/N5xx Integrated Graphics            
    driver: i915 v: kernel                                                          
  Device-2: IMC Networks Integrated Webcam type: USB driver: uvcvideo               
  Display: x11 server: X.org v: 1.21.1.8 with: Xwayland v: 22.1.9 driver: X:        
    loaded: intel,v4l dri: i915 gpu: i915 resolution: 1024x600~60Hz                 
  API: OpenGL v: 2.1 Mesa 24.0.9 renderer: i915 (: Pineview M)                      
[jose@localhost ~]$
Comment 5 Jose Manuel López 2024-06-25 14:20:26 CEST
I am mesa of tainted repo.

I can't change mesa tainted for mesa core, urpmi tells me that thist can break the system.
Comment 6 Morgan Leijström 2024-06-25 15:30:20 CEST
When I QA-test mesa I usually test both flavours.

I switch from tainted to core flavour by having tainted repos disabled, while *downgrading* installed mesa packages.

You can list your installed mesa tainted packages by 
$ rpm -qa | grep mesa | grep tainted

Im my case I downgraded by this command:
$ sudo urpmi --test --downgrade lib64mesavulkan-drivers-24.0.9-1.mga9 mesa-24.0.9-1.mga9 lib64mesagl1-24.0.9-1.mga9 lib64mesaegl1-24.0.9-1.mga9 lib64osmesa8-24.0.9-1.mga9 lib64mesaopencl1-24.0.9-1.mga9 lib64mesaglesv2_2-24.0.9-1.mga9

"--test" parameter included for dry run, remove that to really execute it.
If you do not use sudo, run it as root.

My shell-fu is not advanced enough to automate rpm -qa -> urpmi.


To switch from core to tainted simply enable tainted repos and perform normal system update; it seem to think tainted is an upgrade even if same version.
Comment 7 Jose Manuel López 2024-06-25 16:23:51 CEST
I followed the instructions of comment 6, and I have can downgrade the mesa tainted package.

Now, I have installed the mesa core package, and I have the same result with Nheko. The text of app don't see.

Greetings and thanks for the comments!
Comment 8 Jose Manuel López 2024-06-25 16:25:45 CEST
I have using Lightdm and works fine. Mageia Control Center works too.

I have checked others apps, and the bug only appears with nheko for the moment.
Comment 9 katnatek 2024-06-25 20:10:04 CEST
Put this info https://bugs.mageia.org/attachment.cgi?id=14573 from bug#33310 as it could help to debug the issue

It looks like a hardware/software combination https://community.khronos.org/t/intel-driver-texture-lookup-limitation/63464 but is weird that is not produce until you full update https://bugs.mageia.org/show_bug.cgi?id=33310#c13 

I start to think could be a thing with Qt updates https://forum.qt.io/topic/102105/qopenglshader-link-error-con-gl_position
Comment 10 katnatek 2024-06-25 20:29:30 CEST
(In reply to katnatek from comment #9)
> I start to think could be a thing with Qt updates
> https://forum.qt.io/topic/102105/qopenglshader-link-error-con-gl_position
I was confusing with plasma updates sorry
Comment 11 Jose Manuel López 2024-06-25 20:43:23 CEST
By testing with the help of Katnatek, we have downgraded the version to 23.1.9.1. 

With this version, after restarting, the application works as it should, and all its texts are displayed, both in the chats and in the menus and windows.
katnatek 2024-06-25 20:44:40 CEST

CC: (none) => j.alberto.vc

Comment 12 Jose Manuel López 2024-06-25 20:59:20 CEST
More tests. We have updated again to version 23.3.3-1 to see which version fails next.

After restarting, in this version the application crashes again and the texts are no longer visible.

So the bug must be in the version change from 23.1 to version 23.3 and later.
Comment 13 Giuseppe Ghibò 2024-06-25 21:11:11 CEST
Try using "LIBGL_ALWAYS_SOFTWARE=1 nheko" to use software acceleration.

There is also a newer nheko 0.12 around. And cauldron has it, maybe you could try just to recompile 0.12 in mga9.

Does switching from "intel" driver to "modesetting" change something?

Does the kernel journal contains some report about "illegal instruction"?

Upstream there were similar probs, but for mediatek chip and zink (e.g
https://gitlab.freedesktop.org/mesa/mesa/-/issues/6355).
Comment 14 Jose Manuel López 2024-06-25 23:14:20 CEST
Well,I have tested to use the command of comment 13. Now nheko works fine.

I don't know if this option is relevant to table packaging or only affects my computer (I assume so, because the same thing doesn't happen to me on others).

But with the option "LIBGL_ALWAYS_SOFTWARE=1 nheko" , the application works perfectly and the texts are back again.
Comment 15 Giuseppe Ghibò 2024-06-26 00:21:44 CEST
(In reply to Jose Manuel López from comment #14)

> Well,I have tested to use the command of comment 13. Now nheko works fine.
> 
> I don't know if this option is relevant to table packaging or only affects
> my computer (I assume so, because the same thing doesn't happen to me on
> others).
> 
> But with the option "LIBGL_ALWAYS_SOFTWARE=1 nheko" , the application works
> perfectly and the texts are back again.

It forces to use the llvmpipe software driver 3D accelerator instead of the hardware one. games and intensive application would be slower, but for that 2D application probably is enough.

Other attempt is to edit /etc/X11/xorg.conf and replace Device "intel" with Device "modesetting".
Comment 16 Giuseppe Ghibò 2024-06-26 00:24:02 CEST
(In reply to Giuseppe Ghibò from comment #15)

> Other attempt is to edit /etc/X11/xorg.conf and replace Device "intel" with
> Device "modesetting".

I meant *Driver* "intel" with Driver "modesetting".
katnatek 2024-06-26 00:37:45 CEST

Summary: Nheko breaks with latest Mesa 24.09 update => Mesa from 23.3.3 to 24.0.9 breaks rendering in Nheko

Comment 17 David GEIGER 2024-06-26 06:16:37 CEST
I have the same issue here on Cauldron with blender:

$ blender
Read prefs: "/home/david/.config/blender/4.1/config/userpref.blend"
EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example, a valid context requires buffers not supplied by a valid surface).
EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example, a valid context requires buffers not supplied by a valid surface).
EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example, a valid context requires buffers not supplied by a valid surface).
EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example, a valid context requires buffers not supplied by a valid surface).
Warning: No OpenGL vendor detected.
blender: ../src/dispatch_common.c:872: epoxy_get_proc_address: Assertion `0 && "Couldn't find current GLX or EGL context.\n"' failed.
Abandon (core dumped)

CC: (none) => geiger.david68210

Comment 18 Jose Manuel López 2024-06-26 09:09:08 CEST
Hi!

The "modesetting" option does not work. After reboot the system crashes. 

I have tried configuring the graphical server from the Mageia Control Center, but disabling hardware acceleration, and after rebooting, everything works fine.

Nheko appears correctly.
Comment 19 Jose Manuel López 2024-06-26 09:27:49 CEST
By disabling hardware acceleration, there are several applications that feel heavier, and checking video, smplayer now opens a new mpv window without volume options and others when starting a video.

So for now I run nheko again from the terminal with the previously indicated option "LIBGL_ALWAYS_SOFTWARE=1 nheko", which I cannot indicate in the .desktop file either because it does not work. I can only run it like this from the terminal.
Comment 20 Jose Manuel López 2024-06-26 09:53:38 CEST
Well, looking at some shortcuts on another of my computers with a Plasma desktop, I have seen that for the shortcut located in "/usr/share/applications/" to work, the correct order would be to change the line:

exec= nheko %U

by

exec=env LIBGL_ALWAYS_SOFTWARE=1 nheko

Then, log out and reopen and the application now works for me from the applications menu with this variable.
Comment 21 katnatek 2024-07-01 01:52:36 CEST
(In reply to David GEIGER from comment #17)
> I have the same issue here on Cauldron with blender:
> 
> $ blender
> Read prefs: "/home/david/.config/blender/4.1/config/userpref.blend"
> EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example,
> a valid context requires buffers not supplied by a valid surface).
> EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example,
> a valid context requires buffers not supplied by a valid surface).
> EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example,
> a valid context requires buffers not supplied by a valid surface).
> EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example,
> a valid context requires buffers not supplied by a valid surface).
> Warning: No OpenGL vendor detected.
> blender: ../src/dispatch_common.c:872: epoxy_get_proc_address: Assertion `0
> && "Couldn't find current GLX or EGL context.\n"' failed.
> Abandon (core dumped)

Your issue could be a few different and perhaps should open a new bug
Look if this https://bbs.archlinux.org/viewtopic.php?pid=2141399#p2141399 helps but that thread and a gentoo one https://forums.gentoo.org/viewtopic-t-1169651-start-0.html , make me think that your card not fit the new blender requirements

Note You need to log in before you can comment on or make changes to this bug.