Page 1 of 1

[SOLVED]xfwm problem (?)

PostPosted: May 3rd, '21, 17:07
by p68
hello,

after careful online search i couldn't find a hint for this problem:

I installed the Mageia8 XFCE spin to an old Laptop and everthing seems fine until the blue Mageia background occurs and the login is done, but imediately after this the screen begins to blink a B/W pattern and the blue background.

On a command line I killed one subprocess of xfce4-session after the other, until the blinking stops. Seemingly the problems has to do with xfwm. If I kill xfwm only, usability is hampered, but everything seems to work so far.

I realized two zombie processes, namely qt5-check-openg and ionize.

Any help was gratefully appreciated. - Peter

inxi says:
Code: Select all
System:    Host: localhost Kernel: 5.10.16-desktop586-1.mga8 i686 bits: 32 Console: tty 2 Distro: Mageia 8 mga8
Machine:   Type: Portable System: Dell product: Inspiron 8100
           Mobo: Dell model: Inspiron 8100 serial: N/A BIOS: Dell v: A15 date: 03/03/2003
CPU:       Info: Single Core model: Intel Pentium III Mobile bits: 32 type: MCP L2 cache: 512 KiB
           Speed: 733 MHz min/max: 733/1000 MHz Core speed (MHz): 1: 733
Graphics:  Device-1: NVIDIA NV11M [GeForce2 Go] driver: nouveau v: kernel
           Display: server: X.org 1.20.10 driver: nouveau unloaded: fbdev,modesetting,vesa tty: 200x75
           Message: Advanced graphics data unavailable in console for root.

Re: xfwm problem (?)

PostPosted: May 3rd, '21, 20:27
by doktor5000
p68 wrote:I realized two zombie processes, namely qt5-check-openg and ionize.

That should read qt5-check-opengl and ionice. The former is called to detect if there's OpenGL acceleration or if software rendering should be used by Qt.
The script is run from /etc/X11/xinit.d/qt5-check-opengl2.sh and it shouldn't really cause a loop. What happens if you run glxinfo from within your session ?

Also, is there a difference if you boot from the XFCE medium and select "Boot with non-free nvidia driver" or the regular default option (which uses the free nouveau driver, same as you currently have active)? Just for comparison ...

Re: xfwm problem (?)

PostPosted: May 4th, '21, 11:26
by p68
thank you for your interest.

I run glxinfo by means of these steps:
when the blinking on the X begins, I enter CTRL ALT F2 entering a cli console
login as root
pstree -p|grep xfwm
kill -9 <pid>
repeat the last two commands once: now there is no process of xfwm running
CTRL ALT F1
blinking stopped, right-click in the background, start a terminal and enter
Code: Select all
$glxinfo
name of display: :0.0
display: :0  screen: 0
direct rendering: Yes
server glx vendor string: SGI
server glx version string: 1.4
server glx extensions:
    GLX_ARB_create_context, GLX_ARB_create_context_no_error,
    GLX_ARB_create_context_profile, GLX_ARB_fbconfig_float,
    GLX_ARB_framebuffer_sRGB, GLX_ARB_multisample,
    GLX_EXT_create_context_es2_profile, GLX_EXT_create_context_es_profile,
    GLX_EXT_fbconfig_packed_float, GLX_EXT_framebuffer_sRGB,
    GLX_EXT_import_context, GLX_EXT_libglvnd, GLX_EXT_no_config_context,
    GLX_EXT_texture_from_pixmap, GLX_EXT_visual_info, GLX_EXT_visual_rating,
    GLX_INTEL_swap_event, GLX_MESA_copy_sub_buffer, GLX_OML_swap_method,
    GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_pbuffer,
    GLX_SGIX_visual_select_group, GLX_SGI_make_current_read,
    GLX_SGI_swap_control
client glx vendor string: Mesa Project and SGI
client glx version string: 1.4
client glx extensions:
    GLX_ARB_context_flush_control, GLX_ARB_create_context,
    GLX_ARB_create_context_no_error, GLX_ARB_create_context_profile,
    GLX_ARB_create_context_robustness, GLX_ARB_fbconfig_float,
    GLX_ARB_framebuffer_sRGB, GLX_ARB_get_proc_address, GLX_ARB_multisample,
    GLX_EXT_buffer_age, GLX_EXT_create_context_es2_profile,
    GLX_EXT_create_context_es_profile, GLX_EXT_fbconfig_packed_float,
    GLX_EXT_framebuffer_sRGB, GLX_EXT_import_context, GLX_EXT_swap_control,
    GLX_EXT_swap_control_tear, GLX_EXT_texture_from_pixmap,
    GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_INTEL_swap_event,
    GLX_MESA_copy_sub_buffer, GLX_MESA_multithread_makecurrent,
    GLX_MESA_query_renderer, GLX_MESA_swap_control, GLX_OML_swap_method,
    GLX_OML_sync_control, GLX_SGIS_multisample, GLX_SGIX_fbconfig,
    GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group, GLX_SGI_make_current_read,
    GLX_SGI_swap_control, GLX_SGI_video_sync
GLX version: 1.4
GLX extensions:
    GLX_ARB_create_context, GLX_ARB_create_context_no_error,
    GLX_ARB_create_context_profile, GLX_ARB_fbconfig_float,
    GLX_ARB_framebuffer_sRGB, GLX_ARB_get_proc_address, GLX_ARB_multisample,
    GLX_EXT_create_context_es2_profile, GLX_EXT_create_context_es_profile,
    GLX_EXT_fbconfig_packed_float, GLX_EXT_framebuffer_sRGB,
    GLX_EXT_import_context, GLX_EXT_swap_control, GLX_EXT_texture_from_pixmap,
    GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_INTEL_swap_event,
    GLX_MESA_copy_sub_buffer, GLX_MESA_query_renderer, GLX_MESA_swap_control,
    GLX_OML_swap_method, GLX_OML_sync_control, GLX_SGIS_multisample,
    GLX_SGIX_fbconfig, GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group,
    GLX_SGI_make_current_read, GLX_SGI_swap_control, GLX_SGI_video_sync
Extended renderer info (GLX_MESA_query_renderer):
    Vendor: Nouveau (0x10de)
    Device: Mesa DRI nv11 x86/MMX/SSE (0x112)
    Version: 20.3.4
    Accelerated: yes
    Video memory: 21MB
    Unified memory: no
    Preferred profile: compat (0x2)
    Max core profile version: 0.0
    Max compat profile version: 1.2
    Max GLES1 profile version: 1.0
    Max GLES[23] profile version: 0.0
OpenGL vendor string: Nouveau
OpenGL renderer string: Mesa DRI nv11 x86/MMX/SSE
OpenGL version string: 1.2 Mesa 20.3.4
OpenGL extensions:
    GL_AMD_shader_trinary_minmax, GL_ANGLE_texture_compression_dxt3,
    GL_ANGLE_texture_compression_dxt5, GL_APPLE_packed_pixels,
    GL_ARB_clear_buffer_object, GL_ARB_compressed_texture_pixel_storage,
    GL_ARB_copy_buffer, GL_ARB_debug_output, GL_ARB_draw_buffers,
    GL_ARB_get_program_binary, GL_ARB_get_texture_sub_image,
    GL_ARB_half_float_pixel, GL_ARB_invalidate_subdata,
    GL_ARB_map_buffer_alignment, GL_ARB_multi_bind, GL_ARB_multisample,
    GL_ARB_multitexture, GL_ARB_parallel_shader_compile,
    GL_ARB_program_interface_query, GL_ARB_robustness, GL_ARB_sampler_objects,
    GL_ARB_separate_shader_objects, GL_ARB_shader_objects,
    GL_ARB_shading_language_100, GL_ARB_shading_language_include,
    GL_ARB_texture_compression, GL_ARB_texture_env_add,
    GL_ARB_texture_env_combine, GL_ARB_texture_env_crossbar,
    GL_ARB_texture_env_dot3, GL_ARB_texture_mirrored_repeat,
    GL_ARB_texture_rectangle, GL_ARB_texture_storage, GL_ARB_transpose_matrix,
    GL_ARB_vertex_array_object, GL_ARB_vertex_attrib_binding,
    GL_ARB_vertex_buffer_object, GL_ARB_window_pos, GL_ATI_draw_buffers,
    GL_EXT_EGL_sync, GL_EXT_abgr, GL_EXT_bgra, GL_EXT_blend_color,
    GL_EXT_blend_minmax, GL_EXT_blend_subtract, GL_EXT_compiled_vertex_array,
    GL_EXT_copy_texture, GL_EXT_direct_state_access,
    GL_EXT_draw_range_elements, GL_EXT_fog_coord, GL_EXT_framebuffer_blit,
    GL_EXT_framebuffer_object, GL_EXT_multi_draw_arrays,
    GL_EXT_packed_depth_stencil, GL_EXT_packed_pixels, GL_EXT_rescale_normal,
    GL_EXT_secondary_color, GL_EXT_separate_specular_color,
    GL_EXT_stencil_wrap, GL_EXT_subtexture, GL_EXT_texture, GL_EXT_texture3D,
    GL_EXT_texture_compression_dxt1, GL_EXT_texture_compression_s3tc,
    GL_EXT_texture_edge_clamp, GL_EXT_texture_env_add,
    GL_EXT_texture_env_combine, GL_EXT_texture_env_dot3,
    GL_EXT_texture_filter_anisotropic, GL_EXT_texture_lod_bias,
    GL_EXT_texture_object, GL_EXT_texture_rectangle, GL_EXT_vertex_array,
    GL_IBM_multimode_draw_arrays, GL_IBM_rasterpos_clip,
    GL_IBM_texture_mirrored_repeat, GL_KHR_context_flush_control,
    GL_KHR_debug, GL_KHR_no_error, GL_KHR_parallel_shader_compile,
    GL_MESA_pack_invert, GL_MESA_window_pos, GL_NV_blend_square,
    GL_NV_fog_distance, GL_NV_light_max_exponent, GL_NV_packed_depth_stencil,
    GL_NV_texgen_reflection, GL_NV_texture_env_combine4,
    GL_NV_texture_rectangle, GL_OES_read_format, GL_S3_s3tc,
    GL_SGIS_generate_mipmap, GL_SGIS_texture_edge_clamp, GL_SGIS_texture_lod,
    GL_SUN_multi_draw_arrays

72 GLX Visuals
    visual  x   bf lv rg d st  colorbuffer  sr ax dp st accumbuffer  ms  cav
  id dep cl sp  sz l  ci b ro  r  g  b  a F gb bf th cl  r  g  b  a ns b eat
----------------------------------------------------------------------------
0x021 24 tc  0  32  0 r  y .   8  8  8  8 .  .  0 24  8  0  0  0  0  0 0 None
0x022 24 dc  0  32  0 r  y .   8  8  8  8 .  .  0 24  8  0  0  0  0  0 0 None
0x0cb 24 tc  0  32  0 r  . .   8  8  8  8 .  .  0  0  0  0  0  0  0  0 0 None
0x0cc 24 tc  0  32  0 r  . .   
 ---%< snip %< ----------------------


Perhaps I should mention, that some years ago I tested Mageia6. In order to run X properly it was neccessary to switch off the acceleration of nouveau. To do so, I entered the line
GRUB_CMDLINE_LINUX_DEFAULT=video=LVDS-1:1600x1200 noveau.noaccel=1 in /etc/default/grub.old ,
but now with Mageia8 this doesn't help. Certainly I have another problem.
By the way, there are no X-problems with the live-distris antix19.3 and devuan (which have other probs).

I checked the start with "Boot with non-free nvidia driver". This stops with the lines
[ 19.470959] sd 3:0:0:0: [sdc] Assuming drive cache: write through
[FAILED] Failed to start. Run DrakX installer X setup scripts.
Fedora kicked the older NVIDIA-driver out marking them as obsolete and offers only packages Nvidia..340 and 390 for newer cards.
I don't know whether Mageia8 offers an appropriate package, but the Inspiron 8100 has a Geforce2Go and needs the package NVIDIA..304, I assume.

Re: xfwm problem (?)

PostPosted: May 4th, '21, 17:36
by doktor5000
p68 wrote:Fedora kicked the older NVIDIA-driver out marking them as obsolete and offers only packages Nvidia..340 and 390 for newer cards.
I don't know whether Mageia8 offers an appropriate package, but the Inspiron 8100 has a Geforce2Go and needs the package NVIDIA..304, I assume.

Yes, same for us, have a look at https://wiki.mageia.org/en/Mageia_8_Rel ... DIA_driver
Although from what I can tell, /etc/X11/xinit.d/qt5-check-opengl2.sh itself should not be the issue, it should detect that you have only OpenGL 1 and enable software rendering.

This is the relevant part it checks:
OpenGL version string: 1.2 Mesa 20.3.4


So if you run echo $QT_XCB_FORCE_SOFTWARE_OPENGL inside your X session it should say 1.

The question is probably, what runs the loop in the first place. I think the issue could come from mageiawelcome, which is the welcome applet, which is based on Qt's QML and requires OpenGL.

Can you check if the problem persists if you create an empty file .config/mageiawelcome/norun.flag ? So as regular user
Code: Select all
mkdir -p ~/.config/mageiawelcome
echo -n ~/.config/mageiawelcome/norun.flag

That is, if you haven't already unchecked "display at startup" in the welcome applet ...

Re: xfwm problem (?)

PostPosted: May 4th, '21, 21:45
by p68
Following your advice I run 'echo $QT_XCB_FORCE_SOFTWARE_OPENGL' in a terminal inside the X session. It says nothing: empty echo. However, I can do this only after I killed xfwm. The graphical surface is not usable, when it is blinking.

I created a subfolder of ~/.config named mageiawelcome, containing an empty file norun.flag (simply by means of touch ..). This had no effects. I found somewhere the hint, that the norun.flag file has to be placed in the folder ~/.mageiawelcome, but this had no effect, too. I killed the process mageiawelcome, but then the notebook stalled. Finally I renamed /etc/xdg/autostart/mageiawelcome.desktop appending "-". But the blinking took place anyway, though definitely no process mageiawelcome existed.

I checked the names of the zombie-processes again. Yes, indeed, I misspelled 'ionice', but the other zombie is really named 'qt5-check-openg' - and not 'qt5-check-opengl' .

Re: xfwm problem (?)

PostPosted: May 4th, '21, 22:43
by doktor5000
Well, I don't know why it would do that. You can look at and run /etc/X11/xinit.d/qt5-check-opengl2.sh - it's not really complicated - but to check if this really causes the effect you describe.
You can also simply move that file out of the /etc/X11/xinit.d/ folder and it will not be run in your X session.

Re: xfwm problem (?)

PostPosted: May 5th, '21, 13:29
by p68
Yes, I experimented already with that script. The line 5: 'QT5_CHECK_OPENGL_VERSION=..' is kind of artwork. I entered the two lines 'echo $QT5_CHECK..' and 'echo $QT_XCB_FORCE..' before the line 12: 'unset ..' and run the script in a terminal inside the X-session. Both variables are 1. However, if I enter 'echo $QT_XCB_FORCE..' immediately after, the variable is empty. The script exports this variable, shouldn't it be set to 1 still ?

If I disable the script by changing the file-extension to 'sh-', I see a blinking screen again.

Is that script really working correctly, and if so, how to find the reason of the problem ?

Re: xfwm problem (?)

PostPosted: May 5th, '21, 19:25
by doktor5000
QT5_CHECK_OPENGL_VERSION is only used during runtime of the script, while QT_XCB_FORCE_SOFTWARE_OPENGL should be exported to 1 in case only Open GL 1.x is supported.

I'm not sure what the blinking could be. I'd suggest disabling everything in autostart and check if the issue still persists. See https://wiki.archlinux.org/title/xfce#Autostart - I believe the tool is called xfce-autostart-editor but I'm not sure.

Re: xfwm problem (?)

PostPosted: May 6th, '21, 11:54
by p68
Disabling all autostarts does not help.

However, I found this: https://bbs.archlinux.org/viewtopic.php?id=135042&p=2
Though I definitely don't understand what I do (Zauberlehrling), I tried to set vblank=off !

What I did:
as usual, I killed xfwm4, when the blinking starts, then I opened a terminal inside the X-session and typed : xfwm4 --vblank=off . Surprise !! I have a usable screen: the window-manager works like a charm.

I guess, this is a hint to the cause of the problem.

If there are no better means to solve this issue, can you tell me, how to instruct xfwm4 to use this setting as a default ?

Re: xfwm problem (?)

PostPosted: May 6th, '21, 17:47
by doktor5000
Did you try as mentioned in the bugreport in the post you linked to ? Although it should probably be
Code: Select all
xfconf-query -c xfwm4 -p "/general/vblank_mode" -s off


From what I can tell, default value is stored in /usr/share/xfwm4/defaults but if you change it there it would be overriden by the user setting and also overriden on updates.

Re: xfwm problem (?)

PostPosted: May 6th, '21, 18:34
by p68
Yes, I tried already. The command complains
Code: Select all
Die Eigenschaft »/general/vblank_mode« ist in dem Kanal »xfwm4« nicht vorhanden. Um eine neue Eigenschaft anzulegen, bitte die Option --create« benutzen.

Unfortunately there are no man pages, neither for xfwm4 nor for xfconf-query. I don't know, how to use the Option "--create". Because I don't know how to revoke the ghosts, I call, I hesitated to play with the command xfconf-query.

It seems, this is an old bug. Could you please explain some background. Why does it occur only in the 32bit-XFCE-spin of Mageia8 and not in other XFCE-distributions ? Mageia6 didn't show this problem, too.

Re: xfwm problem (?)

PostPosted: May 6th, '21, 18:42
by doktor5000
From what I can tell, that property does exist, and that command also works here to change that value.

Listing all the settings and searching for vblank shows the property
Code: Select all
[live@localhost Desktop]$ xfconf-query -c xfwm4 -l | grep -i vblank
/general/vblank_mode


The current value can also be queried:
[live@localhost Desktop]$ xfconf-query -c xfwm4 -p "/general/vblank_mode"
auto
[/code]

Apart from that, I've only looked at the help:
Code: Select all
[live@localhost Desktop]$ xfconf-query --help
Usage:
  xfconf-query [OPTION…] - Xfconf commandline utility

Help Options:
  -h, --help            Show help options

Application Options:
  -V, --version         Version information
  -c, --channel         The channel to query/modify
  -p, --property        The property to query/modify
  -s, --set             The new value to set for the property
  -l, --list            List properties (or channels if -c is not specified)
  -v, --verbose         Verbose output
  -n, --create          Create a new property if it does not already exist
  -t, --type            Specify the property value type
  -r, --reset           Reset property
  -R, --recursive       Recursive (use with -r)
  -a, --force-array     Force array even if only one element
  -T, --toggle          Invert an existing boolean property
  -m, --monitor         Monitor a channel for property changes


For some examples, check https://wiki.archlinux.org/title/Xfwm#Video_tearing

Re: xfwm problem (?)

PostPosted: May 6th, '21, 19:44
by p68
The file /usr/share/xfwm4/defaults contains a line "vblank_mode=auto". However, if I enter the command 'xfconf-query -c xfwm4 -l', there is no line containing "vblank". I inspected the complete list. The 'grep' gives an empty answer for me.

Of course I looked at the help offered by 'xfconf-query' and I visited https://docs.xfce.org/xfce/xfconf/xfconf-query .
Can I create and set the missing property by means of a command like this:
xfconf-query -c xfwm4 --create /general/vblank_mode -t string -s off

or is '/vblank_mode' the right parameter or something else ? What is the correct incantation ? Please, Magician, help!

Re: xfwm problem (?)

PostPosted: May 6th, '21, 20:10
by doktor5000
p68 wrote:Can I create and set the missing property by means of a command like this:
xfconf-query -c xfwm4 --create /general/vblank_mode -t string -s off


Yes, that looks about right, although you're missing a -p for the property
Code: Select all
xfconf-query -c xfwm4 --create -p /general/vblank_mode -t string -s off


You can also find some examples at https://docs.xfce.org/xfce/xfconf/xfconf-query and https://wiki.archlinux.org/title/xfce

Re: xfwm problem (?)

PostPosted: May 7th, '21, 11:31
by p68
Now it is working. At least for me the problem is solved.

I am happy, that I still can use my good old Inspiron 8100, which I bought twenty years ago.

Thanks to Mageia for offering a 32bit medium.

Thank you, doktor5000, for your helpful patience and perseverance.

Re: xfwm problem (?)

PostPosted: May 7th, '21, 15:11
by isadora
Please, don't forget to mark the topic [SOLVED].
You can do so, by editing the subject/title in the first message in this topic.
Write [SOLVED] to the left of subject/title, thanks ahead. ;)