Something went wrong. Try again.
A Wayland client for Zig 0.16 with no libwayland: protocol scanner, sans-I/O wire protocol, and socket client.
Something went wrong. Try again.
zig-wayland-native flake.nix
4.0 kB · 108 lines
Nix
at main
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109# SPDX-FileCopyrightText: © 2026 Jeffrey C. Ollie <jeff@ocjtech.us># SPDX-License-Identifier: MIT
{ description = "A Wayland client written in Zig, with no libwayland";
inputs = { nixpkgs = { url = "https://channels.nixos.org/nixos-unstable/nixexprs.tar.xz"; }; };
outputs = { nixpkgs, ... }:
let inherit (nixpkgs) lib; makePackages = system: import nixpkgs { inherit system; }; forAllSystems = lib.genAttrs lib.systems.flakeExposed;
# The devshell's Zig, with one line of its own standard library put # right, because without it `zig build fuzz --fuzz` cannot compile. # # Zig 0.16.0's `compiler/test_runner.zig` reports a failing fuzz input by # asking `std.debug.writeStackTrace` to print what `@errorReturnTrace()` # gave it. Those are two different types: an error return trace is a # `builtin.StackTrace`, a ring buffer with a write index, and that # function takes a `debug.StackTrace`, which is a plain slice and a count # of what was skipped. It is a type error, it is on the path taken only # under `-ffuzz`, and it stops *any* project with a fuzz test in it from # building one. The fix is the function next door: `writeErrorReturnTrace` # takes exactly the type in hand and is what the other three places in # the same file use. # # `--replace-fail` is the whole safety of this: the day Zig ships the fix # the pattern will not be found, the build will fail here rather than # patch something else, and this can go. # # It buys the fuzzer and not its coverage. Nothing in this release # populates the table of program counters, so a bounded run ends with # "corrupted coverage file: pcs_len was zero" and an unbounded one # panics in the build runner's coverage thread; neither is a finding, # and a finding says "input saved to" above the report. fuzzableZig = pkgs: let # A farm of symlinks rather than a copy: the library is 217 MB, and # exactly one file of it is being changed. library = pkgs.runCommand "zig-0.16.0-lib-fuzz-fix" { } '' cp -rs --no-preserve=mode ${pkgs.zig_0_16}/lib/zig $out chmod -R u+w $out rm $out/compiler/test_runner.zig cp --no-preserve=mode \ ${pkgs.zig_0_16}/lib/zig/compiler/test_runner.zig \ $out/compiler/test_runner.zig substituteInPlace $out/compiler/test_runner.zig \ --replace-fail \ 'std.debug.writeStackTrace(trace, stderr)' \ 'std.debug.writeErrorReturnTrace(trace, stderr)' ''; in pkgs.symlinkJoin { name = "zig-0.16.0-fuzzable"; paths = [ pkgs.zig_0_16 ]; nativeBuildInputs = [ pkgs.makeWrapper ]; postBuild = '' wrapProgram $out/bin/zig --set ZIG_LIB_DIR ${library} ''; }; in { devShells = forAllSystems ( system: let pkgs = makePackages system; in { default = pkgs.mkShell { name = "zig-wayland-native"; nativeBuildInputs = [ (fuzzableZig pkgs) pkgs.git-pages-cli pkgs.reuse # For trying the examples against a compositor that is not the # desktop: `weston --backend=headless`, and `wayland-info` to # compare what it advertises with what `run-globals` prints. # sway is there for explicit sync, which headless weston does # not offer and headless sway on GLES does. pkgs.sway pkgs.wayland-utils pkgs.weston ] ++ lib.optionals pkgs.stdenv.hostPlatform.isLinux [ pkgs.kcov pkgs.perf ]; }; } ); };}