[Wayland + NVIDIA + WGPU] Error in Surface::configure: parent device is lost

[uncomfyhalomacro]

What Operating System(s) are you seeing this problem on?

Linux Wayland

Which Wayland compositor or X11 Window manager(s) are you using?

WM: Wayland compositor RiverWM (Wlroots-based like SwayWM)
OS: openSUSE TumbleWeed

Also tested on Plasma Wayland. Same issue.

WezTerm version

20221119.145034.49b9839f+g221

Did you try the latest nightly build to see if the issue is better (or worse!) than your current version?

No, and I'll explain why below.

No relevant commit that can fix it

Describe the bug

Crashes on Wayland. No Wezterm running.

To Reproduce

1. Open a Wayland session -> In my instance, riverwm.
2. Segfaults

Configuration

no config

Expected Behavior

Runs normally.

Logs

23:18:27.081  ERROR  wezterm_gui > running message loop: error during event_q.read_events protocol_error=Some(ProtocolError { code: 7, object_id: 0, object_interface: "<unknown>", message: "" }): Protocol error (os error 71); terminating

Anything else?

Going to test this on Plasma Wayland. I will update later

[uncomfyhalomacro]

Updated. It crashes as well. It seems to have a problem with latest Wayland. Not just the wlroots implementation.

[uncomfyhalomacro]

Tried latest git master. Same issue.

[smokeythemonkey]

Happens for me on Pop os / Gnome with wayland enabled

[pdanpdan]

Set the __EGL_VENDOR_LIBRARY_FILENAMES env variable to /usr/share/glvnd/egl_vendor.d/50_mesa.json (at least for PopOS)

[uncomfyhalomacro]

/usr/share/glvnd/egl_vendor.d/50_mesa.json

That works. This might mean this is an NVIDIA Wayland issue?

[wez]

I'd suggest trying webgpu to see if that behaves any better:
https://wezfurlong.org/wezterm/config/lua/config/front_end.html#webgpu

[carschandler]

I'd suggest trying webgpu to see if that behaves any better:

https://wezfurlong.org/wezterm/config/lua/config/front_end.html#webgpu

Hi wez. We discovered that my issue #3244 is a duplicate of this and I've tried running with WebGpu on which allowed it to boot up but with extremely degraded performance and weird graphics errors. I've got some additional info over on that post.

[supermarin]

I'm observing this on wayland/sway but only when >1 instance is running. i.e. I can normally use 1 instance of WezTerm, but launching the second one crashes both.

[almereyda]

This is also being seen with the Flatpak on Ubuntu 23.04 in OpenGL mode. A window can still open, but doesn't hold for long, e.g. until a few keys are pressed. Opening a second instance instantly kills both.

Enabling WebGpu throws the error immediately here:

$ flatpak run org.wezfurlong.wezterm 
[destroyed object]: error 2: Buffer size (12x613) must be an integer multiple of the buffer_scale (2).
01:30:24.924  ERROR  wezterm_gui > running message loop: error during event_q.dispatch protocol_error=Some(ProtocolError { code: 2, object_id: 0, object_interface: "<unknown>", message: "" }): Protocol error (os error 71); terminating
$

Do we know of versions that are unaffected, and if, is there a way to downgrade the Flatpak?

[swsnr]

I also see this issue on Gnome 44 on Arch Linux. No Nvidia or AMD card, just an Intel iGPU.

[uncomfyhalomacro]

@wez I found out that this is not a wezterm issue, but a crate issue wgpu, Wayland library issue, specifically, wlroots and also NVIDIA issue.

It's an old issue as mentioned here gfx-rs/wgpu#2519

And I am seeing the same errors as well:

thread 'main' panicked at 'Error in Surface::configure: parent device is lost', /home/abuild/rpmbuild/BUILD/wezterm-20230408.112425.69ae8472/vendor/wgpu/src/backend/direct.rs:316:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
warning: queue 0x5567ea017160 destroyed while proxies still attached:
  wl_display@1 still attached
dhat: Total:     89,537,428 bytes in 239,026 blocks
dhat: At t-gmax: 10,375,395 bytes in 2,933 blocks
dhat: At t-end:  7,293,968 bytes in 3,810 blocks
dhat: The data has been saved to dhat-heap.json, and is viewable with dhat/dh_view.html

Similar issues are referenced as well:

• Can't run examples with Nvidia + Wayland gfx-rs/wgpu#3810
• screenshot example crashes on Linux, Wayland with Nvidia GPU bevyengine/bevy#8604

I will keep this open until their side gets resolved.

[ymatsiuk]

while on intel TGL GT2 + older wlroots/sway, the following fixed the crash

wlroots: f4eaab12a056e339b1b7548f211bb463ad064392
sway: 92244c87dbb85080c99eee10db9fed906117e169

~ ❯ grep wezterm ~/.config/sway/config
bindsym Mod4+Return exec wezterm start --always-new-process
exec wezterm start --always-new-process

NOTE: This is no longer working on the latest wlroots/sway revisions and seems to be related to #3996

[uncomfyhalomacro]

Fixed on latest wezterm. Only issue now is to report for performance

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues. If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.