Graphical glitches with Raspberry Pi 400 + LinuxCNC 2.9.4 (arm64)

The official LinuxCNC 2.9.4 image for arm64 is installed on a Raspberry Pi 400. There are strange partial graphical glitches in browsers, such as some pages or partial horizontal section of the page not displaying properly, while the rest is fine.

This was tested with Firefox ESR, but Chromium was also installed, and for example, the settings page for Chromium never displays properly. It would appear the underlying text and buttons/links are present but not visible, except by sometimes mousing over them. 

The problem appears to be only in browsers, not other applications. Youtube in either browser is also unusable, with graphical glitches on the player controls (YT buttons like play/pause/settings are either invisible or look corrupt).

Similarly, when opening PDF documents in browsers, the control buttons along the top of the document for Zoom/Download/Save etc are often invisible, but still present if moused over.

Also, Youtube video playback is extremely laggy, and that is including after removing the isolcpu=2,3 option. The goal on the Pi is just to familiarize with LinuxCNC and set up configs, not to run machines on it, so I'm not too concerned about the latency penalty in this case.. 

Has anyone else experienced this problem? 

This is really more of a Debian/RPi isue, not a LinuxCNC issue.

A Raspberry Pi has a weak GPU, why would you use it for YouTube playback?

Use a Mac Mini, an old PC, or something else with an actual GPU (a dedicated GPU is better, but not essential; a good iGPU can be good enough).

Post a screenshot so we know what you are talking about.

To tell the truth, whilst I haven't done any long term testing, I haven't seen this issue of my RPi-400.
When I build the image I put it through the basic tasks, and if I had of seen this issue arise I wouldn't have released the image, I would have deemed it defective and not fit for purpose.
Now we are going to require some more info, what monitor (brand model) screen resolution and such. And just to be sure I would like to know if this is a clean install without any extras added. If you can remember when was the last time you did an update.

And yes the RPi400 should give reasonable playback on youtube and reading pdfs shouldn't pose an issue. When I get the time I'll recheck things but as I said before I haven't seen this issue arise during testing and so far in the 6 months it has been out in the wild no one else has mentioned this.

One thing I would suggest is look at this thread, forum.linuxcnc.org/38-general-linuxcnc-q...l-images-only#325007 and take note of the video issue link & zswap link.

Thanks for the replies. I finally got a chance to check out the suggestions on the RPi400. For testing I made a new fresh image on a new SD card, and the graphical glitches could be reproduced, but it may be just an issue in chromium browser, which is not installed by default. I did not reproduce the issue in Firefox, and though it is painfully slow to load pages, Youtube does play... eventually.

The SD cards used are 64GB Kingston Canvas Select Plus. Nothing fancy, but I doubt it is the issue, since I've used multiple, and other images (default RPi OS, Ubuntu 24.04) don't have the same issue.

For reference, here are some screenshots of the glitches in chromium. Btw, I had to install xfce4-screenshooter to be able to take screenshots, as it is not installed by default in the LinuxCNC rpi image.

The first screenshot shows a horizontal section of the web page with corrupted graphics. The second is chromium's setting page which never displays correctly. The third is a pdf document opened in chromium, with the controls (download, zoom, etc) at the top not showing. The fourth shows the graphical glitches in the Youtube player controls (buttons for play, pause, settings, etc).

They work in Firefox (very slow though, which is why I installed chromium) so perhaps this is not an issue specifically with LinuxCNC's image, though I can't tell. I don't have these issues with chromium on the other RPi OS images, but I have not tried debian 12 directly. I imagine that also precludes hardware issues. The monitor is a 22" Dell at 1680x1050, though I doubt that is too relevant.

Aside from isolcpu, were there other optimizations made which may impact graphical or general performance?
 

Ok the screenshots are interesting. Unfortunately I haven't had a chance to check as life has got in the way.
To tell the truth when I test the images I use an old 19" 4:3 monitor. Another thing to be aware of is that the images aren't meant to be for a daily driver, the sole purpose is to run Linuxcnc which is what is concentrated on. So yes there will be packages left out that are part of a standard Desktop install.
The kernel was built using the official RPi sources, with the only mod being enabling the RT section in the kernel config. So there is really nothing fancy in that area.
It very well be an issue related to resolution, in config.txt you should be able to adjust the memory allocated to the graphics system, something you'll have to look at the config.txt documentation, 2 things to be aware of, there is no raspi-config utility so the config.txt will need to be hand edited and the location of it is different to the Raspberry OS, the main thread for the images has this info.
If you just want to explore what Linuxcnc has to offer without running a machine or connecting to any hardware there is also the option of running the amd64 version as a live session on a PC or installing in a virtual machine such as VirtualBox. RT capabilities will be non existent but it will run the simulations fine.
Another option would be to install RaspiOS and build Linuxcnc from source, a real time kernel would not be required in your case and use at ad a Run In Place install. How to do this can be found in the docs.

Those screenshots remind me of these screenshots of Minecraft looking really strange sometimes on my Dell XPS L702X with a NVIDIA GeForce 550M GPU.

www3.langdonstaab.ca/Minecraft%20bug%20screenshots/

Why is accessing Chrome necessary?

Check chrome://gpu

Consider submitting a bug report to Chromium, maybe they'll fix this.

You can also submit a bug report for Firefox, though in my experience they don't actually fix bugs.

How is the monitor connected?

VGA? DisplayPort? HDMI? DVI?

Maybe try a different resolution?

Good point about checking chrome://gpu, it shows a number of problems with graphics features, such as hardware decode being software only.

Graphics Feature Status
Canvas: Hardware accelerated
Direct Rendering Display Compositor: Disabled
Compositing: Hardware accelerated
Multiple Raster Threads: Enabled
OpenGL: Enabled
Rasterization: Hardware accelerated on all pages
Raw Draw: Disabled
Skia Graphite: Disabled
TreesInViz: Disabled
Video Decode: Software only. Hardware acceleration disabled
Video Encode: Software only. Hardware acceleration disabled
Vulkan: Disabled
WebGL: Hardware accelerated
WebGL2: Hardware accelerated
WebGPU: Disabled
WebNN: Disabled

 

Problems Detected
WebGPU has been disabled via blocklist or the command line.
Disabled Features: webgpu
Accelerated video encode has been disabled, either via blocklist, about:flags or the command line.
Disabled Features: video_encode
Accelerated video decode has been disabled, either via blocklist, about:flags or the command line.
Disabled Features: video_decode
Disable partial swaps on Mesa drivers (detected with GL_VERSION): 339493
Applied Workarounds: disable_post_sub_buffers_for_onscreen_surfaces
Disable KHR_blend_equation_advanced until cc shaders are updated: 661715
Applied Workarounds: disable(GL_KHR_blend_equation_advanced), disable(GL_KHR_blend_equation_advanced_coherent)
Expose WebGL's disjoint_timer_query extensions on platforms with site isolation: 808744, 870491
Applied Workarounds: enable_webgl_timer_query_extensions
Some drivers can't recover after OUT_OF_MEM and context lost: 893177
Applied Workarounds: exit_on_context_lost
Avoid waiting on a egl fence before swapping buffers and rely on implicit sync on Broadcom GPUs: 938286
Applied Workarounds: rely_on_implicit_sync_for_swap_buffers
Disable GL_MESA_framebuffer_flip_y for desktop GL: 964010
Applied Workarounds: disable(GL_MESA_framebuffer_flip_y)

Full report attached below.

I don't think the issue is hardware or resolution, since it is just this image that has these problems.