How to build Bevy apps in remote development environment?

[jsjoeio]

Hi! New to Bevy. I use a remote development environment (Ubuntu based Docker container). I connect via VS Code in the browser (code-server). When I try and run a very basic Bevy app, I get this error:

thread 'main' panicked at 'Failed to initialize any backend! Wayland status: "backend disabled" X11 status: 
LibraryOpenError(OpenError { kind: Library, detail: 
"opening library failed (libXcursor.so.1: cannot open 
shared object file: No such file or directory); opening 
library failed (libXcursor.so: cannot open shared object 
file: No such file or directory)" })', 
/home/coder/.cargo/registry/src/github.com-1ecc6299db9ec823/winit-0.26.1/src/platform_impl/linux/mod.rs:619:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Is this possible or does Bevy only work with local development environments?

[alice-i-cecile]

You might be interested in #4700, which sets up remote development using Gitpod. I'm not immediately sure how adaptable it might be, but it's a place to look.

[jsjoeio]

Wow, that has to be the world's fastest reply! I'll take a look 👀

[jsjoeio]

Turns out, it's easier than you think. Bevy can be compiled for WebAssembly. Follow this guide and it will run your game on a webserver.

[mockersf]

if your game can't work in wasm (need more shadows, need more perf, need multithread, ...), there are still options open to you:

• run your game on your server. you'll need a server with a GPU, and something like X11 forwarding or screen sharing
• cross compile your game on your server for your local setup, then download the binary and run it locally. you'll need a GPU locally in this case

[bjorn3]

I believe wayland + waypipe works better than X11 for forwarding programs using hardware acceleration I believe. Waypipe forwards the framebuffer as an (h.264?) compressed video.

[AgustinRamiroDiaz]

Related post to this idea: https://jasoncg.isthecode.ninja/2023/09/20/developing-gui-wayland-app-bevy-in-vscode-devcontainer/

[DeepDoge]

You can also run cargo rustc then run the built from host. So you only have development related things inside the container.
And can test the thing directly inside your host machine. I think this approach is more clean.
You can also use SSH which makes everything easier.

Here how I did it:

• First you gotta have SSH open on your host of course.
• Then inside devcontainer.json add these.
  You will need these with in the container.

"containerEnv": {
    "HOST_USER": "${localEnv:USER}",
    "HOST_WORKSPACE": "${localWorkspaceFolder}",
    "HOST_DISPLAY": "${localEnv:DISPLAY}",
    "HOST_XAUTHORITY": "${localEnv:XAUTHORITY}"
}

• And then inside your project create a file named run or any other name you want.
  Inside the script we first run cargo rustc if it was successful, then we ssh connect to the host machine and run the thing.

#!/bin/bash

# Build the project
cargo rustc

# Check if the build was successful, otherwise exit
if [ $? -ne 0 ]; then
    exit 1
fi

# Get the host user and ip address
REMOTE_ADDRESS="${HOST_USER}@$(ip route | grep default | awk '{print $3}')"

# Check if SSH key pair exists, otherwise generate one
if [ ! -f ~/.ssh/id_rsa ]; then
  ssh-keygen -t rsa -N "" -f ~/.ssh/id_rsa
fi

# Copy the public key to the local host if not already done
ssh-copy-id -i ~/.ssh/id_rsa.pub ${REMOTE_ADDRESS} >/dev/null 2>&1

# Run the program on the local host
ssh ${REMOTE_ADDRESS} \
    "XAUTHORITY=${HOST_XAUTHORITY} DISPLAY=${HOST_DISPLAY} ${HOST_WORKSPACE}/target/debug/$(basename "${HOST_WORKSPACE}")"

If im missing something let me know, just started using rust and bevy today.
This was the best way to get started for me since i use dev containers within flatpak vscode and dont let development related things on my host system.

[cs50victor]

Leaving this here in hopes it helps someone in future : https://github.com/cs50victor/new_media/pull/18

[1mamute]

I've been using VSCode's devcontainer to develop bevy for a few days. There were a lot of fiddling around until I've managed to make it work: mounting GPU, X11 and sound sockets, installing NVIDIA containers toolkit on host machine, installing cuda toolkit and drivers inside container.

Relevant informations about my setup: Ubuntu 22.04, NVIDIA GPU with proprietary drivers, system user mapped to docker group
The exact requirements may depend on your setup, but it may serve as a starting point:

1. Install and configure NVIDIA's Container Toolkit on your host machine: required to mount the GPU into containers

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
&& curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list
sudo apt-get install -y nvidia-container-toolkit
sudo nvidia-ctk runtime configure --runtime=docker

2. Enable X11 forwarding using the command xhost +, remember to disable later xhost -
3. Mount the GPU, X11, sound sockets and set the DISPLAY environment variable inside the container:

// .devcontainer/devcontainer.json
{
	"name": "Rust",
	"build": {
		"dockerfile": "Dockerfile"
	},

	"runArgs": [
		// Map sound devices (doesn't work on Windows, see https://github.com/microsoft/WSL/issues/689)
                "--device=/dev/snd:/dev/snd",
                "--device=/dev/dri:/dev/dri",
		"-e", "DISPLAY=${env:DISPLAY}", // Set the display for X11
                "-v", "/tmp/.X11-unix:/tmp/.X11-unix", // Mount the X11 socket
		"--gpus", "all", // Mount the GPU
                "--privileged" // It may or not be necessary to run the container in privileged mode
	]
}

3. Install required drivers and tools inside the container, for me, the following were required

# .devcontainer/Dockerfile
FROM mcr.microsoft.com/devcontainers/rust:1-1-bullseye

# Install cuda-toolkit
RUN distribution=$(. /etc/os-release;echo $ID$VERSION_ID | sed -e 's/\.//g') && \
    wget https://developer.download.nvidia.com/compute/cuda/repos/$distribution/x86_64/cuda-keyring_1.0-1_all.deb && \
    dpkg -i cuda-keyring_1.0-1_all.deb && \
    export DEBIAN_FRONTEND=noninteractive && \
    apt-get -y update && \
    apt-get -y install cuda-toolkit && \
    rm cuda*

# Install additional packages.
RUN export DEBIAN_FRONTEND=noninteractive && apt-get -y upgrade \
    && apt-get -y install --no-install-recommends \
    g++ \
    pkg-config \
    libx11-dev \
    libasound2-dev \
    libudev-dev \
    libxkbcommon-x11-0 \
    libxcursor-dev \
    libxrandr-dev \
    libxi-dev \
    mesa-vulkan-drivers \
    lld \
    clang

# Cargo tools for nightly must be installed with vscode user
USER vscode
RUN rustup toolchain install nightly && \
    rustup component add rustfmt --toolchain nightly-x86_64-unknown-linux-gnu && \
    rustup component add clippy --toolchain nightly-x86_64-unknown-linux-gnu

Good luck!