ERROR_OUT_OF_HOST_MEMORY with Qt on Linux/Wayland

[almarklein]

Relates to #470

Trying to use wgpu with PySide6 today, I get the following error:

thread '<unnamed>' panicked at /root/.cargo/git/checkouts/wgpu-53e70f8674b08dd4/9f505e7/wgpu-hal/src/vulkan/instance.rs:395:18:
XlibSurface::create_xlib_surface() failed: ERROR_OUT_OF_HOST_MEMORY
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
fatal runtime error: failed to initiate panic, error 5
Aborted (core dumped)

Things work as expected with glfw. The error seems to be related to obtaining a surface id. A process that indeed goes a little different with the qt backend, because we seem unable to get the actual surface id of a qt widget. I don't yet understand why this is suddenly a problem. It worked fine some weeks ago.

[almarklein]

Seems specific to Wayland.

[hmaarrfk]

I've been using wayland on Ubuntu 24.04 as my daily driving for the last few weeks at least. (Only integrated graphics, no Nvidia graphics card)

I tried it in on Ubuntu 23.10 and it was pretty bad and I would see errors like the one you described above. So my hunch is that likely the some version in the graphics stack is too old.

[hmaarrfk]

And to speicfy, we use the following conda packages for our wgpu/pygfx stack:

 $ mamba list | grep -E "(wgpu|pyside|pygfx)"
pygfx                     0.2.0              pyhd8ed1ab_0    conda-forge
pyside6                   6.7.0           py310h7a3fa89_1    conda-forge
wgpu-native               0.19.3.1             h2b8f863_0    conda-forge
wgpu-py                   0.15.2             pyha804496_0    conda-forge

[hmaarrfk]

ps. i noticed that conda-forge qt6-wayland is likely broken recently, so maybe i'm not using "qt wayland"

[almarklein]

We track this in pygfx/rendercanvas#36