Pinebook Pro display experiences significant graphical corruption

[Bluberry-Kat]

Niri doesn't run on the Pinebook Pro out of the box (similarly to the PinePhone in #320), but can be made to launch by modifying render-drm-device.
When launched the environment looks as follows:

Two interesting notes:

• The issue is not reproduced when running niri in a nested session (i.e Sway).
• The graphical corruption persists until a functioning graphical environment is present (i.e When niri-session is ran via a TTY the corruption still persists after niri exits, but does not persist when being ran via SDDM).

Log output is included here:
niri.log

This issue has been briefly mentioned before in the matrix. My Pinebook is hardly a daily driver, so I'm willing to tinker with it extensively if you need more information.

System Information

• niri version: 25.01-1

• Distro: Manjaro ARM

• GPU: Mali T864

• CPU: Rockchip RK3399

[YaLTeR]

Interesting. Could you do the following:

1. Try if this reproduces on anvil and/or cosmic.
2. Attach drm_info output for both niri (broken) and sway (working).

When niri-session is ran via a TTY the corruption persists once niri exits

This hints at a driver bug I think

[Bluberry-Kat]

Sure. Results in a few.

[Bluberry-Kat]

I haven't had any success getting anvil or cosmic to run. Here's the drm_info outputs for niri and sway while I work that out.

drm_niri.log
drm_sway.log

[YaLTeR]

There is no difference between the two. Well, idk.

How did you try running anvil? You need to pass it the --tty-udev flag, and also you may need to set the render device using ANVIL_DRM_DEVICE env variable.

[Bluberry-Kat]

I ran anvil with cargo run -- --tty-udev as the readme suggested, yeah.
Regardless of ANVIL_DRM_DEVICE (card1 contains the logs of, well, card1, which is known to "work" on niri, but generates the corrupted graphics) anvil failed to launch. Not really sure what to do from there. Will try Cosmic next after a night of sleep.

anvil-card1.log

anvil-card0.log

[Drakulix]

Can you please try to set ANVIL_DRM_DEVICE to the render-node instead of the card-node? So something like ANVIL_DRM_DEVICE=/dev/dri/renderD128

[Bluberry-Kat]

Assigning /dev/dri/renderD128 did work to get the compositor open. I'm not clear what the expected behavior is while running anvil; my screen appears to just be white with a mouse cursor moving around, but damn if it's not a good looking (i.e, not corrupted) mouse cursor.

anvil-renderD128.log

I also attempted specifying the render-node to niri but the behavior wasn't interestingly different. (The incorrect colors were still incorrect, but different colors this time. Fuzziness still apparent.)