nandi/frq · flake.nix

nandi/frqpublic⑂ Fork 0
⑂ b33f288db8fbd9caaa202ce396497c77f76c1c49
flake.nix · 556 lines · 26.8 KBNix    Blame HistoryRaw
Build the client with Nix, not just a shell for it   1
# frq is Jolt source, so "building" it is three things, not one:
#   jolt        the runtime that reads it        (github:jolt-lang/jolt)
#   jolt-native libvidya and libjoltmoq, in Rust (gitlab:nandithebull/jolt-native)
#   frq         this tree, with its deps resolved to store paths
# Jolt resolves deps.edn by running git at startup, which a build sandbox has
# no network for — so every dep is fetched by Nix instead and handed back as
# a :local/root through -Sdeps.
#   nix build .#frq && ./result/bin/frq
# On a machine that is not NixOS the GL driver is the host's and the loader
# will not find it, so the window never opens ("GL display: argument does not
Find the host's GL driver when the store's is not the system's   16  # name a valid config"). The launcher handles that itself: off NixOS it hands
# the process to nixGL, which puts the host's driver ahead of the store's.
# Nothing extra to type, and a distrobox/container Arch is the same case as
# a bare one.
Build the client with Nix, not just a shell for it   20  description = "frq — a freeq client in jolt";
inputs = {
  nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
Take jolt and jolt-native as flakes, off my own fork   24    # `git+https` with `?submodules=1` rather than the github scheme: Jolt's
  # own flake declares `self.submodules`, which this Nix rejects when the
  # flake is fetched as `github:`. Its outputs are not what we take — the
  # runtime is built here, by joltFrom — but it is a flake all the same, so
  # its own inputs are locked with ours rather than left to float, and
  # `vendor/` comes along as the submodule the build needs.
  # The fork rather than jolt-lang/jolt: it is what jolt-android-src already
  # pins for the boot image, and a desktop runtime built from a different
  # tree than the APK's is the same drift the jolt-native comment warns
  # about. Unpinned here — the desktop follows the fork's main, while the
  # APK stays on the rev below.
Build the client with Nix, not just a shell for it   36    jolt-src = {
Take jolt and jolt-native as flakes, off my own fork   37      url = "git+https://gitlab.com/nandithebull/jolt?submodules=1";
    inputs.nixpkgs.follows = "nixpkgs";
Build the client with Nix, not just a shell for it   39
Take the Android objects from jolt-native's CI, always the latest   41    # The source half of jolt-native: the Jolt code under glimmer-backends/ that
  # binds the native objects, and the flake that builds the desktop ones. The
  # Android objects no longer come from here — jolt-native-android below
  # fetches those prebuilt — so this input is what `just run` builds against
  # and what an APK's Clojure side is read from, both at this rev.
Take the jolt-native ahead of the release, since a run can build it   46
  # Pinned all the same, and pinned to a rev: this input carries both halves of
Pin the native half to the release the rest of the tree names   48    # glimmer-vidya — libvidya, and the Jolt side that binds it — so an
  # unpinned `main` is a build whose native half is free to sit at a
  # different commit from the tree that talks to it. It did, and what the
  # drift cost was silence: the Jolt half sent a reaction pill's hover card
  # to a libvidya with no handler for one, and the pill said nothing.
Merge origin/main, and let one jolt-native answer for both backends   53    # It also carries the terminal backend — crates/jolt-tui, the same tree ABI
  # over a grid of cells, and jolt/glimmer-tui beside glimmer-vidya. That was
  # a second input at a second rev while it lived on a branch, which is the
  # drift this comment warns about wearing a different hat: one input, and
  # the window and the terminal are the same library either way.
Build the client with Nix, not just a shell for it   58    jolt-native = {
Take the Android objects from jolt-native's CI, always the latest   59      url = "git+https://gitlab.com/nandithebull/jolt-native?rev=d970307ccf1fe67e2e971f2837d2282d8ba79a62";
Take jolt and jolt-native as flakes, off my own fork   60      inputs.nixpkgs.follows = "nixpkgs";
Paint the same screens into a terminal   61
Take the Android objects from jolt-native's CI, always the latest   63    # The Android objects, prebuilt by jolt-native's CI rather than compiled
  # here: an APK needs libvidya and libjoltmoq for arm64, and building them
  # locally means an NDK, a Rust cross toolchain and the whole crane graph
  # for two files that upstream already built and published.
  # "latest" is the version its CI overwrites on every default-branch build,
  # so this input finds a new one on `nix flake update` -- but flake.lock
  # still records exactly which bytes an APK was built from, which is the
  # pin that matters. `just bump` decides when to move; this only decides
  # where to look. The archive is rooted at include/ and lib/arm64-v8a/, so
  # nothing here has to unpack it.
  jolt-native-android = {
    url = "https://gitlab.com/api/v4/projects/nandithebull%2Fjolt-native/packages/generic/jolt-native/latest/android-arm64-v8a.tar.gz";
    flake = false;
Build the APK from the flake, and from nothing   79    # Chez itself, because the APK needs a cross target nixpkgs does not
  # build: frq's Scheme is compiled to an arm64 boot image, and that wants
  # Chez's own `tarm64le` workarea — boot files, xpatch and libkernel.a.
  # The version is the one the hand-built tree under ~/.cache used, and the
  # submodules are not optional (zuo builds it, lz4 and zlib link into it).
Take jolt and jolt-native as flakes, off my own fork   84    # The same fork jolt-src takes, built here rather than fetched as a
  # release binary: upstream reads the socket address out of `struct
Build the APK from the flake, and from nothing   86    # addrinfo` at glibc's offset, which on Bionic is `ai_canonname`, so an APK
Take jolt and jolt-native as flakes, off my own fork   87    # built with upstream cannot open a TLS connection at all. Pinned to a rev
  # where jolt-src is not: the APK is a release artefact, so its runtime
  # moves when `just bump` says so rather than when the fork does.
Build the APK from the flake, and from nothing   90    jolt-android-src = {
    url = "git+https://gitlab.com/nandithebull/jolt?rev=2b80d68d1f7a31ba92b208b3957e5fb555617ada&submodules=1";
Take jolt and jolt-native as flakes, off my own fork   92      inputs.nixpkgs.follows = "nixpkgs";
Build the APK from the flake, and from nothing   93
  chez-src = {
    url = "git+https://github.com/cisco/ChezScheme?ref=refs/tags/v10.4.1&submodules=1";
    flake = false;
Build the client with Nix, not just a shell for it   100    # The sha deps.edn pins, on the fork with the reconciler fixes.
  glimmer = {
    url = "git+https://gitlab.com/nandithebull/glimmer?rev=399df371c790d690fb6e4560c3d4d7f838502857";
    flake = false;
Find the host's GL driver when the store's is not the system's   105
  # Only ever used off NixOS, to put the host GL driver on the loader path.
  nixgl = {
    url = "github:nix-community/nixGL";
    inputs.nixpkgs.follows = "nixpkgs";
Bundle the closure into one file the AppImage runtime mounts   111
  # Wraps a closure into a single self-extracting file. Only the `appimage`
  # output evaluates it.
  nix-appimage = {
    url = "github:ralismark/nix-appimage";
    inputs.nixpkgs.follows = "nixpkgs";
Build the client with Nix, not just a shell for it   118
Fetch libmoq_ffi again, and keep the codecs out of Rust   120  outputs = { self, nixpkgs, jolt-src, jolt-native, jolt-native-android, glimmer, chez-src, jolt-android-src, nixgl, nix-appimage }:
Build the client with Nix, not just a shell for it   121
    systems = [ "x86_64-linux" "aarch64-linux" ];
    forEachSystem = f:
      nixpkgs.lib.genAttrs systems (system: f nixpkgs.legacyPackages.${system});
Run this tree on a native half the builders made   125
    # Mesa, despite the name: it covers Intel and AMD alike. The NVIDIA
    # wrappers are the ones that need --impure (they read the host kernel
    # module's version), which is why this only ever reaches for Intel.
Stop carrying the 32-bit GL stack and half of git   129
    # Built from nixGL's default.nix rather than taken from its flake
    # outputs, for the one argument the flake hardcodes on: `enable32bits`,
    # which on x86_64 puts a second, i686 copy of mesa, its LLVM, and
    # intel-media-driver into the wrapper. frq is 64-bit on both halves —
    # the Rust cdylibs and the Chez runtime — so nothing here ever opens the
    # 32-bit driver, and carrying it is most of the dev shell's closure.
    nixGLFor = pkgs: (import nixgl {
      inherit pkgs;
      enable32bits = false;
    }).nixGLIntel;
Run this tree on a native half the builders made   140
    # egui reaches for these with dlopen rather than linking them, so being
    # in the cdylib's buildInputs is not enough — whatever starts frq has to
Follow the xorg renames, and stop wrapping nix in distrobox   143      # put them on the loader path itself. Without libx11 here, vidya reports
Run this tree on a native half the builders made   144      # "X11 unavailable", falls back to Wayland, and winit refuses to build a
    # second event loop after the failed first one.
    # Out here rather than beside the package that first needed them: the
    # dev shell starts frq too, on this tree's source rather than the store's
    # copy of it, and a second copy of this list is a second chance for the
    # two ways of running to disagree about what the window needs.
    runtimeLibsFor = pkgs: with pkgs; [
      libGL
      libxkbcommon
      wayland
Follow the xorg renames, and stop wrapping nix in distrobox   155        libx11
      libxcursor
      libxi
      libxrandr
Run this tree on a native half the builders made   159        vulkan-loader
Build the client with Nix, not just a shell for it   161
    packages = forEachSystem (pkgs:
      let
        inherit (pkgs) lib;
Run this tree on a native half the builders made   167          nixGL = nixGLFor pkgs;
Find the host's GL driver when the store's is not the system's   168
Take the native objects from their own flake rather than rebuilding them   169          # libvidya (the retained-tree ABI glimmer-vidya binds, on egui),
        # libjolttui (the same tree over a grid of cells) and libjoltmoq (the
        # AV media plane) — one workspace, three cdylibs, taken from
        # jolt-native's own flake rather than rebuilt here.
        # This used to be a rustPlatform.buildRustPackage over the same
        # source, which meant restating upstream's build: the seven git deps
        # hashed by hand in `cargoLock.outputHashes` and re-hashed whenever
        # its Cargo.lock moved, the linuxHeaders path v4l2r's bindgen wants,
        # and a postPatch dropping the .cargo/config.toml that pointed the
        # build at DotSlash. Upstream's flake says all of that itself now,
        # and says it once. It also builds cpal with the `pipewire` feature,
        # which the restatement did not — so device names in a call are
        # PipeWire's rather than raw ALSA PCMs.
Switch frq.av onto the jolt media plane, and stop shipping libjoltmoq   183          # libvidya and libjolttui, NOT libjoltmoq. Its `default` is all
        # three joined, and the third is the Rust media plane frq no
        # longer loads — `frq.av.plane` replaced it.
        # This makes the closure smaller and the APK smaller. It does NOT
        # make the build shorter, and it is worth being exact about why:
        # jolt-native compiles its external crates ONCE, in a
        # `buildDepsOnly` derivation shared by all three objects, so
        # asking for two of them still builds every dependency the third
        # has — the 440 crates that are jolt-moq's alone. Getting those
        # out of the build is a change in jolt-native, not here: either
        # jolt-moq leaves that workspace, or its deps artifact stops
        # being workspace-wide.
        native =
          let np = jolt-native.packages.${pkgs.stdenv.hostPlatform.system};
          in pkgs.symlinkJoin {
            name = "jolt-native-ui";
            paths = [ np.libvidya np.libjolttui ];
Build the client with Nix, not just a shell for it   202
Fetch libmoq_ffi again, and keep the codecs out of Rust   203          # libmoq_ffi — MoQ over QUIC behind UniFFI's C ABI, FETCHED rather
        # than built. This is the object `frq.moq.raw` is generated from.
Bind libmoq_ffi from the object's own metadata, not from its header   205
Fetch libmoq_ffi again, and keep the codecs out of Rust   206          # Fetched because building it is the thing this whole exercise is
        # about: moq-ffi pulls moq-native, iroh, quinn, rustls and aws-lc-sys
        # behind it, which is 440 crates that nothing else in this tree
        # needs. Upstream already publishes the object for both Linux
        # architectures, so we take those bytes.
Bind libmoq_ffi from the object's own metadata, not from its header   211
Fetch libmoq_ffi again, and keep the codecs out of Rust   212          # Pinned to a release and to a hash, and the hashes below are
        # upstream's own published .sha256 files rather than ones observed
        # here — a `nix-prefetch` of whatever the URL serves today would
        # record that it downloaded something, not that it downloaded the
        # right thing.
        # WHAT THIS BUILD IS NOT: moq-ffi's `audio` and `video` features are
        # on by default upstream and are OFF in these artifacts, so there is
        # no publish_audio/publish_video and no moqaudio*/moqvideo* here —
        # 206 functions where the Apple artifact has 230. That is why the
        # bindings are generated from the object (`just gen-moq`) and not
        # from the C header the release ships, which describes the Apple one.
        moqFfi =
            version = "0.3.17";
            target = {
              "x86_64-linux" = "x86_64-unknown-linux-gnu";
              "aarch64-linux" = "aarch64-unknown-linux-gnu";
            }.${pkgs.stdenv.hostPlatform.system};
            hash = {
              "x86_64-linux" = "sha256-dzQXpV4JgdtD+g33WX51FFAQdfCUXkNsx1xPbobPfUI=";
              "aarch64-linux" = "sha256-PdzRwbJFqOZWRgI0HHX2XUH+Ljh4V3jvQ9asfvCuIPA=";
            }.${pkgs.stdenv.hostPlatform.system};
          pkgs.stdenv.mkDerivation {
            pname = "libmoq-ffi";
            inherit version;
            src = pkgs.fetchurl {
              url = "https://github.com/kixelated/moq/releases/download/moq-ffi-v${version}/moq-ffi-${version}-${target}-libmoq_ffi.so";
              inherit hash;
            dontUnpack = true;
            # It carries no RUNPATH and needs libgcc_s, libm and libc — the
            # host's on an ordinary distro, and nothing at all on NixOS
            # unless they are bound here.
            nativeBuildInputs = [ pkgs.autoPatchelfHook ];
            buildInputs = [ pkgs.stdenv.cc.cc.lib ];
            installPhase = ''
              mkdir -p $out/lib
              cp $src $out/lib/libmoq_ffi.so
              chmod +w $out/lib/libmoq_ffi.so
Bind libmoq_ffi from the object's own metadata, not from its header   255
        # One directory for the loader to look in. jolt resolves every
        # :jolt/native name against JOLT_NATIVE_LIB, and the objects now come
        # from two places — jolt-native's flake, and the moq-ffi release — so
        # they are joined rather than the path being made a list, which the
        # loader does not take.
Bind the C libraries a call actually needs   261          # The C codecs, from nixpkgs. libmoq_ffi carries the transport and
        # nothing else — moq-ffi's `audio` and `video` features would have
        # brought Opus and H.264 with them, at the price of compiling a
        # 1062-crate workspace — so the codecs are linked here instead,
        # where they have always lived.
        # Named in :jolt/native, so the loader resolves them the same way it
        # resolves libvidya: by name, out of one directory.
        # A flat C face for openh264, because openh264 has none. Its
        # `ISVCEncoder` is `const ISVCEncoderVtbl*` — every method is a
        # function pointer in a vtable — and jolt.ffi cannot call one: Chez
        # fixes a foreign procedure's types when it compiles it, and the
        # target must be a literal C symbol name. So the vtable is walked in
        # c/frq_h264.c and jolt binds the five plain symbols it exports.
        # One translation unit against a library nixpkgs already has. It is
        # a calling convention adapter, not a second media plane, and the
        # distinction from the moq-ffi build it replaces is the whole point:
        # this compiles one .c file, not a 1062-crate workspace.
        frqH264 = pkgs.stdenv.mkDerivation {
          pname = "frq-h264";
          version = "0.1";
          src = ./c;
          nativeBuildInputs = [ pkgs.pkg-config ];
          buildInputs = [ pkgs.openh264 ];
          buildPhase = ''
            $CC -O2 -fPIC -shared frq_h264.c -o libfrqh264.so \
              $(pkg-config --cflags --libs openh264)
          installPhase = ''
            mkdir -p $out/lib && cp libfrqh264.so $out/lib/
        # openh264 is here for frqH264's DT_NEEDED; alsa-lib for capture
        # and playback. V4L2 needs nothing: it is ioctls against libc and
        # the kernel, so there is no library to name.
        codecs = [ pkgs.libopus pkgs.openh264 frqH264 pkgs.alsa-lib ];
Let the plane take real devices, not only test thunks   300          # ALSA's PipeWire plugin, which is how `default` resolves to
        # anything on a machine running PipeWire — and every machine frq
        # targets does. Without it alsa-lib fails to dlopen
        # libasound_module_pcm_pipewire.so and the only devices that open
        # are raw hardware ones, which PipeWire is already holding.
        # An environment variable rather than a library in the join:
        # alsa-lib looks plugins up by directory, not by soname.
        alsaPluginDir = "${pkgs.pipewire}/lib/alsa-lib";
Bind libmoq_ffi from the object's own metadata, not from its header   310          nativeAll = pkgs.symlinkJoin {
          name = "frq-native";
Bind the C libraries a call actually needs   312            paths = [ native moqFfi ] ++ codecs;
Bind libmoq_ffi from the object's own metadata, not from its header   313
Build the client with Nix, not just a shell for it   315          # Jolt itself: Clojure on Chez, built the way its own flake builds it.
Build the APK from the flake, and from nothing   316          # A function, because there are two of them — upstream for the
        # desktop, and the Bionic-addrinfo fork for the boot image the APK
        # carries. Nothing else about the build differs.
        joltFrom = src: pkgs.stdenv.mkDerivation {
Build the client with Nix, not just a shell for it   320            pname = "jolt";
          version = "dev";
Build the APK from the flake, and from nothing   322            inherit src;
Build the client with Nix, not just a shell for it   323
          strictDeps = true;
          nativeBuildInputs = with pkgs; [ chez makeWrapper pkg-config xxd ];
          buildInputs = with pkgs; [ lz4 zlib ncurses openssl libuuid ];
          JOLT_VERSION = "dev";
          dontConfigure = true;
          buildPhase = ''
            runHook preBuild
            scheme --script host/chez/build-jolt.ss release target/release/jolt
            runHook postBuild
          installPhase = ''
            runHook preInstall
            mkdir -p "$out/bin"
            install -m755 target/release/jolt "$out/bin/jolt"
            runHook postInstall
          # jolt.deps shells out to git and unzip, and jolt.mvn-http dlopens
Stop carrying the 32-bit GL stack and half of git   345            # OpenSSL through the JOLT_OPENSSL_LIBDIR seam. gitMinimal rather
          # than git: all jolt.deps asks for is clone/fetch/rev-parse, and
          # the full package carries Perl and Python for the subcommands
          # written in them — a quarter of a gigabyte for git-send-email.
Read the clock in the reader's zone when /etc/localtime is a file   349
          # TZDIR so a zone *name* resolves wherever this runs: frq.clock
          # hands one to tzset, and glibc then looks for the tzfile under
          # /usr/share/zoneinfo unless told otherwise — which a NixOS host
          # does not have. The store's own tzdata is there on both kinds of
          # machine. --set-default, so a TZDIR the user set still wins.
Build the client with Nix, not just a shell for it   355            postFixup = ''
            wrapProgram "$out/bin/jolt" \
Stop carrying the 32-bit GL stack and half of git   357                --prefix PATH : "${pkgs.lib.makeBinPath [ pkgs.gitMinimal pkgs.unzip ]}" \
Build the client with Nix, not just a shell for it   358                --set-default JOLT_OPENSSL_LIBDIR "${pkgs.lib.makeLibraryPath [ pkgs.openssl ]}" \
Read the clock in the reader's zone when /etc/localtime is a file   359                --set-default TZDIR "${pkgs.tzdata}/share/zoneinfo" \
Build the client with Nix, not just a shell for it   360                --set-default SSL_CERT_FILE "${pkgs.cacert}/etc/ssl/certs/ca-bundle.crt"
Build the APK from the flake, and from nothing   364          joltRuntime = joltFrom jolt-src;
        joltAndroid = joltFrom jolt-android-src;
Build the client with Nix, not just a shell for it   367          # glimmer-vidya lives inside the jolt-native checkout, and its own
        # deps.edn asks for glimmer by git — the top-level override below
        # answers for both.
Take the Android objects from jolt-native's CI, always the latest   370          glimmerVidya = "${jolt-native}/glimmer-backends/glimmer-vidya";
        glimmerTui = "${jolt-native}/glimmer-backends/glimmer-tui";
Build the client with Nix, not just a shell for it   372
Run this tree on a native half the builders made   373          runtimeLibs = runtimeLibsFor pkgs;
Build the client with Nix, not just a shell for it   374
        # The project as jolt sees it: source, deps.edn, nothing else.
        frqSource = pkgs.runCommand "frq-source" { } ''
          mkdir -p "$out"
          cp -r ${self}/src ${self}/deps.edn "$out/"
        # Jolt resolves deps.edn from the working directory, so the launcher
        # runs from the store copy. Its .jolt/cpcache write lands on a
        # read-only directory and jolt treats that as a quiet cache miss, so
        # the only cost is re-resolving the (already local) graph per start.
        frqScript = pkgs.writeShellScript "frq" ''
Bind libmoq_ffi from the object's own metadata, not from its header   386            export LD_LIBRARY_PATH="${nativeAll}/lib:${lib.makeLibraryPath runtimeLibs}''${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
Let the plane take real devices, not only test thunks   387            export ALSA_PLUGIN_DIR="${alsaPluginDir}"
Build the client with Nix, not just a shell for it   388            cd ${frqSource}
Find the host's GL driver when the store's is not the system's   389
          # On NixOS the store's Mesa is the system's and the window opens.
          # Anywhere else the real driver is the host's, so defer to nixGL —
          # it prepends the host driver, which has to win over ours.
          runner=""
          [ -e /run/current-system ] || runner="${nixGL}/bin/nixGLIntel"
          exec ''${runner} ${joltRuntime}/bin/jolt \
Build the client with Nix, not just a shell for it   397              -Sdeps '{:deps {jolt-lang/glimmer {:local/root "${glimmer}"} nandi/glimmer-vidya {:local/root "${glimmerVidya}"}}}' \
            -M:frq "$@"
Paint the same screens into a terminal   401          # The same source, the other backend. No GL, no nixGL and no X11 —
        # a terminal is the one surface that needs nothing from the host but
        # a terminal, which is the reason this output exists.
        tuiScript = pkgs.writeShellScript "frq-tui" ''
Bind libmoq_ffi from the object's own metadata, not from its header   405            export LD_LIBRARY_PATH="${nativeAll}/lib''${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
Let the plane take real devices, not only test thunks   406            export ALSA_PLUGIN_DIR="${alsaPluginDir}"
Paint the same screens into a terminal   407            cd ${frqSource}
          exec ${joltRuntime}/bin/jolt \
            -Sdeps '{:deps {jolt-lang/glimmer {:local/root "${glimmer}"} nandi/glimmer-vidya {:local/root "${glimmerVidya}"} nandi/glimmer-tui {:local/root "${glimmerTui}"}}}' \
            -m frq.tui "$@"
        tui = pkgs.runCommand "frq-tui-0.1.0"
            meta = {
              description = "frq's screens in a terminal";
              mainProgram = "frq-tui";
              platforms = systems;
            mkdir -p "$out/bin"
            ln -s ${tuiScript} "$out/bin/frq-tui"
Build the client with Nix, not just a shell for it   427          frq = pkgs.runCommand "frq-0.1.0"
            meta = {
              description = "A freeq client in jolt";
              mainProgram = "frq";
              platforms = systems;
            mkdir -p "$out/bin"
            ln -s ${frqScript} "$out/bin/frq"
Build the APK from the flake, and from nothing   439          # --- Android ------------------------------------------------------
        # The SDK and the NDK are Google's, which means unfree and a licence
        # to accept — so this is its own import of nixpkgs rather than the
        # `legacyPackages` everything above uses. Confined to the Android
        # outputs: `nix build` of frq itself never evaluates it.
Build the APK as derivations only, and retire the buck2 graph   445          # The NDK here is r29, which is the one the pinned libvidya was
        # built with.
Build the APK from the flake, and from nothing   447          androidPkgs = import nixpkgs {
          inherit (pkgs.stdenv.hostPlatform) system;
          config = {
            allowUnfree = true;
            android_sdk.accept_license = true;
        androidComposition = androidPkgs.androidenv.composeAndroidPackages {
          buildToolsVersions = [ "36.0.0" ];
          platformVersions = [ "36" ];
          includeNDK = true;
        android = import ./nix/android.nix {
Take the Android objects from jolt-native's CI, always the latest   462            inherit pkgs self chez-src jolt-native jolt-native-android glimmer joltAndroid;
Build the APK from the flake, and from nothing   463            inherit (pkgs) lib;
          androidSdk = androidComposition.androidsdk;
          ndk = androidComposition.ndk-bundle;
Build the client with Nix, not just a shell for it   467
Bind the C libraries a call actually needs   469          inherit native moqFfi frqH264 nativeAll frq;
Let the plane take real devices, not only test thunks   470          inherit (pkgs) pipewire;
Merge origin/main, and let one jolt-native answer for both backends   471          inherit tui;
Build the client with Nix, not just a shell for it   472          jolt = joltRuntime;
        default = frq;
Bundle the closure into one file the AppImage runtime mounts   474
        # frq and everything it loads, squashed into one runnable file for
        # hosts without Nix. The whole closure rides along — Mesa included,
        # which is not waste: off NixOS the launcher goes through nixGL, and
        # nixGL needs a store Mesa to put the host's driver in front of.
        appimage =
          nix-appimage.bundlers.${pkgs.stdenv.hostPlatform.system}.default frq;
Build the APK from the flake, and from nothing   481
      # An APK is built by a linux-x86_64 NDK and a linux-x86_64 jolt, and
      # Google ships no other; on aarch64 the Android outputs are simply
      # absent rather than present and broken.
      // lib.optionalAttrs (pkgs.stdenv.hostPlatform.system == "x86_64-linux") {
        inherit (android) apk chezAndroid joltBoot libjoltapp;
        apk-unsigned = android.apk-unsigned;
Build the client with Nix, not just a shell for it   488        });
Run this tree on a native half the builders made   490      # Where `just run` runs, and — because entering it realises what it
    # names — what builds the half of frq that is not this working tree.
    # The two halves, and the split is the whole point of the shell. The frq
    # source is the files on disk, uncommitted edits and all. Everything
    # under it — jolt, glimmer, glimmer-vidya, both native objects — is the
    # flake's, at the revs flake.lock names, so a run says what it ran
    # against and both halves of glimmer-vidya move together. That is the
    # drift the `jolt-native` input's comment is about, and a pin frq can
    # answer for is worth more here than the convenience of a checkout.
Take the native objects from their own flake rather than rebuilding them   501      # `native` is jolt-native's own flake output. It was a buck2 graph when
    # this comment was first written and a cargo build restated here when it
    # was second: buck2 fetches its rustc, zig and every third-party crate as
    # it goes and writes buck-out into the tree it builds, so a sandbox with
    # no network and a read-only store was the one place it could not run.
    # Upstream builds with nix now, so the thing its CI runs and the thing
    # this shell hands a builder are the same derivation.
Run this tree on a native half the builders made   508
    # Nothing here says "nixbuild", though: it is a plain derivation, and
Inline the babashka scripts into the justfile   510      # where it gets built is the machine's business. The `run` recipe asks for
Run this tree on a native half the builders made   511      # the shell with --max-jobs 0, which is what sends it to the `builders`
    # entry rather than compiling egui on a laptop.
    devShells = forEachSystem (pkgs:
      let
        inherit (pkgs) lib;
Bind libmoq_ffi from the object's own metadata, not from its header   516          inherit (self.packages.${pkgs.stdenv.hostPlatform.system}) jolt native nativeAll;
Run this tree on a native half the builders made   517
        default = pkgs.mkShellNoCC {
          name = "frq";
          # jolt, because the runtime frq is run by should be the flake's
          # too. nixGL for the same reason the launcher reaches for it — see
Inline the babashka scripts into the justfile   524            # frqScript. just so the recipe runner comes from here too rather
          # than the host — `nix develop` and then `just run` is the whole of
          # what a machine with nix needs.
          packages = [ jolt pkgs.just (nixGLFor pkgs) ];
          # Read by the recipes rather than baked into a wrapper: the frq
Run this tree on a native half the builders made   530            # source `just run` runs is the working tree, so the launcher has
Inline the babashka scripts into the justfile   531            # to live in that tree and the shell has to hand it its answers.
          # Naming these is also what makes the shell build them.
Bind libmoq_ffi from the object's own metadata, not from its header   533            JOLT_NATIVE_LIB = "${nativeAll}/lib";
Let the plane take real devices, not only test thunks   534            # Spelled out rather than shared with the packages block, which
          # is a different `let`. See `alsaPluginDir` there for why.
          ALSA_PLUGIN_DIR = "${pkgs.pipewire}/lib/alsa-lib";
Run this tree on a native half the builders made   537            GLIMMER_SRC = glimmer;
Take the Android objects from jolt-native's CI, always the latest   538            GLIMMER_VIDYA_SRC = "${jolt-native}/glimmer-backends/glimmer-vidya";
          GLIMMER_TUI_SRC = "${jolt-native}/glimmer-backends/glimmer-tui";
Run this tree on a native half the builders made   540            FRQ_LIB_PATH = lib.makeLibraryPath (runtimeLibsFor pkgs);
          NIXGL = "${nixGLFor pkgs}/bin/nixGLIntel";
      });
Build the client with Nix, not just a shell for it   545      apps = forEachSystem (pkgs: {
      default = {
        type = "app";
        program = "${self.packages.${pkgs.stdenv.hostPlatform.system}.frq}/bin/frq";
Paint the same screens into a terminal   550        tui = {
        type = "app";
        program = "${self.packages.${pkgs.stdenv.hostPlatform.system}.tui}/bin/frq-tui";
Build the client with Nix, not just a shell for it   554