modified: CONTRIBUTING.md

new file:   flake.lock
	new file:   flake.nix
This commit is contained in:
GU Junhua committed 2026-10-11 16:07:46 +08:00
1 parent 0398a813c8
commit 75e9472a47
3 files changed
+200

No files matched your search

+40
View File
@@ -28,6 +28,46 @@ HTTPS or localhost, because it uses the Origin Private File System.
Windows from Linux: `cargo xwin build --release --target x86_64-pc-windows-msvc -p solvecraft`. Windows from Linux: `cargo xwin build --release --target x86_64-pc-windows-msvc -p solvecraft`.
### NixOS: `nix develop`
NixOS has no `/usr/lib`, so nothing is on `LD_LIBRARY_PATH` and a plain `cargo run` finds none of
the libraries the app loads at run time: the Wayland ones winit opens the window through, the EGL
loader wgpu renders through, and libdbus-1, which rfd's file dialogs reach xdg-desktop-portal
through (a dialog then never appears and the pick returns no path):
```
ERROR solvecraft: winit EventLoopError: ... The wayland library could not be loaded
ERROR eframe::native::run: Exiting because of error: WGPU error: Failed to create surface for any enabled backend: {}
ERROR rfd::backend::xdg_desktop_portal::portal::libdbus: Can't connect to a portal: libdbus-1.so not found
```
`nix develop` (the [`flake.nix`](flake.nix) in the repository root) puts the Wayland, X11,
EGL-loader and D-Bus libraries on `LD_LIBRARY_PATH`, so `cargo run --release -p solvecraft` works
from that shell as on any other distribution, file dialogs included. `nix run` builds the app and
wraps the binary the same way. The Mesa drivers stay the system's, from `/run/opengl-driver`
(`hardware.graphics.enable`): that is the driver stack the compositor runs on, and wgpu reaches the
integrated GPU through the EGL loader. The shell deliberately brings no Rust toolchain — the one on
`PATH` keeps its build cache, where a second `rustc` would make every build recompile the workspace
from scratch; `nix develop .#rust` is there for a machine with no Rust at all.
**Anti-aliasing on the OpenGL backend.** NixOS keeps the Vulkan ICD descriptions
(`/run/opengl-driver/share/vulkan/icd.d`) off the Vulkan loader's search path, so wgpu picks its
OpenGL backend on the integrated GPU. On that backend any multisampling above 1× — the default is
4× — draws a black window: the UI still lays out and the control channel still answers, but every
frame, `ui.screenshot` included, comes out empty. Set **Preferences ▸ Graphics ▸ Anti-aliasing** to
1× and restart (the setting applies on the next start), or write it before the first run, since a
black window is no way to reach the dialog:
```sh
mkdir -p ~/.local/share/solvecraft
echo '{"msaa": 1}' > ~/.local/share/solvecraft/graphics.json
```
`apps/solvecraft/src/main.rs` reads that file (`graphics_path()`, eframe's storage dir) before the
window opens; 2× and 4× are the same black window. The Vulkan backend is no way out of it either:
`libvulkan.so.1` is not on `LD_LIBRARY_PATH`, so `WGPU_BACKEND=vulkan` exits with a
`CreateSurfaceError` instead.
## How the code is laid out ## How the code is laid out
The [README](README.md#architecture) has the layer table. The short version: The [README](README.md#architecture) has the layer table. The short version:
Generated
+27
View File
@@ -0,0 +1,27 @@
{
"nodes": {
"nixpkgs": {
"locked": {
"lastModified": 1791594108,
"narHash": "sha256-XK4KceonkwD8LuMAt/vNrz982mXJM5jeHAxMwsUXXAQ=",
"owner": "NixOS",
"repo": "nixpkgs",
"rev": "6510408d147d8c5e6d76043e899e9251756f6a75",
"type": "github"
},
"original": {
"owner": "NixOS",
"ref": "nixos-unstable",
"repo": "nixpkgs",
"type": "github"
}
},
"root": {
"inputs": {
"nixpkgs": "nixpkgs"
}
}
},
"root": "root",
"version": 7
}
+133
View File
@@ -0,0 +1,133 @@
{
description = "SolveCraft — native app development shell and package for Linux";
inputs.nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
outputs =
{ self, nixpkgs }:
let
systems = [
"x86_64-linux"
"aarch64-linux"
];
forAllSystems = f: nixpkgs.lib.genAttrs systems (system: f nixpkgs.legacyPackages.${system});
# Why this shell exists. winit opens the window by dlopen'ing the Wayland client libraries,
# wgpu renders through the EGL loader (libEGL.so.1), and rfd's file dialogs reach
# xdg-desktop-portal through libdbus-1; a plain `cargo run` under NixOS finds none of them,
# because there is no /usr/lib and nothing puts them on LD_LIBRARY_PATH:
# winit EventLoopError: The wayland library could not be loaded
# eframe: WGPU error: Failed to create surface for any enabled backend: {}
# rfd::backend::xdg_desktop_portal::portal::libdbus: Can't connect to a portal: libdbus-1.so not found
# The Mesa drivers themselves — and the EGL vendor / Vulkan ICD descriptions that name them —
# come from the system in /run/opengl-driver (hardware.graphics.enable), not from here, since
# that is the driver stack the compositor runs on. Verified on an Intel LNL iGPU under
# Wayland: wgpu picks the GL backend through this loader and draws.
# The X11 libraries are the ones the app actually uses on a Wayland desktop: with `DISPLAY`
# set and `SOLVECRAFT_WAYLAND` unset, winit runs through XWayland (file drag and drop exists
# only there). The Wayland ones are for `SOLVECRAFT_WAYLAND=1`, and both are needed under
# `xvfb-run` on a machine with no display at all.
runtimeLibs =
pkgs:
(with pkgs; [
wayland
libxkbcommon
libglvnd
dbus
])
++ (with pkgs; [
# xorg.libX11 and friends are deprecated aliases in current nixpkgs.
libx11
libxcursor
libxi
libxrandr
libxrender
libxext
libxcb
]);
driverLibDir = "/run/opengl-driver/lib";
libraryPath = pkgs: "${pkgs.lib.makeLibraryPath (runtimeLibs pkgs)}:${driverLibDir}";
shellHook = ''
if [ ! -d ${driverLibDir} ]; then
echo "solvecraft: ${driverLibDir} is missing — set hardware.graphics.enable = true so the Mesa drivers and their EGL vendor descriptions exist." >&2
fi
command -v cc >/dev/null || echo "solvecraft: no C linker on PATH, cargo cannot link — use nix develop .#rust" >&2
echo "solvecraft: cargo run --release -p solvecraft -- --sample"
'';
in
{
devShells = forAllSystems (pkgs: {
# The shell to work in. It brings no toolchain of its own, Rust or C: the ones already on
# PATH (a system package, rustup, nixpkgs) keep their build cache and their linker, where a
# second rustc — or mkShell's second gcc, with another glibc behind it — would make every
# `cargo build` produce binaries from somewhere else.
default = pkgs.mkShellNoCC {
name = "solvecraft";
LD_LIBRARY_PATH = libraryPath pkgs;
inherit shellHook;
};
# For a machine with no Rust at all: `nix develop .#rust`. Its own target directory, so
# switching between this toolchain and the system one does not thrash the cache.
rust = pkgs.mkShell {
name = "solvecraft-rust";
packages = [
pkgs.cargo
pkgs.rustc
pkgs.rustfmt
pkgs.clippy
];
LD_LIBRARY_PATH = libraryPath pkgs;
# Under target/, which .gitignore already covers.
CARGO_TARGET_DIR = "target/nix";
shellHook = ''
${shellHook}
echo "solvecraft: $(rustc --version), building into target/nix"
'';
};
});
packages = forAllSystems (pkgs: {
default = pkgs.rustPlatform.buildRustPackage {
pname = "solvecraft";
version = (pkgs.lib.importTOML ./Cargo.toml).workspace.package.version;
src = self;
cargoLock.lockFile = ./Cargo.lock;
# Build and install the app alone: the workspace also holds the CLI, the web app, xtask
# and the crates only the test suites need.
cargoBuildFlags = [
"-p"
"solvecraft"
];
cargoInstallFlags = [
"-p"
"solvecraft"
];
# `cargo xtask ci` is the test gate (it wants a wasm target and a GPU); the sandbox has
# neither.
doCheck = false;
nativeBuildInputs = [ pkgs.makeWrapper ];
buildInputs = runtimeLibs pkgs;
# The app reaches the window through dlopen'd libraries, so the packaged binary needs the
# same LD_LIBRARY_PATH the shell exports.
postFixup = ''
wrapProgram $out/bin/solvecraft --prefix LD_LIBRARY_PATH : "${libraryPath pkgs}"
'';
meta = {
description = "Rust-native parametric 3D CAD app";
homepage = "https://github.com/bherbruck/solvecraft";
license = with pkgs.lib.licenses; [
mit
asl20
];
mainProgram = "solvecraft";
};
};
});
};
}