unom/punktfunk
6
0
Fork 1
Code Issues 3 Pull Requests 1 Actions 1 Packages Projects Releases 46 Wiki Activity
Files
dee97e893cf4f2a4403a880acf7a267598fedd36
punktfunk/packaging/flatpak/io.unom.Punktfunk.yml
T
enricobuehler 5c05246098 feat: M10 — FFmpeg is gone from the client 
cargo tree -p punktfunk-client-session finds no ffmpeg. The host still does,
which is the whole point: pf-encode keeps libavcodec unconditionally and no
host workflow, packaging script or licence file was touched.

Deleted: crates/pf-ffvk, video_vulkan.rs, video_vaapi.rs, video_libav.rs, the
libavcodec half of video_d3d11.rs, the av_log machinery, ffmpeg::codec::Id as
the decoder's vocabulary (the quic CODEC_* wire constants now serve, which is
why the evidence table was keyed on them), DecodedImage::VkFrame and ::Dmabuf,
the presenter's AVVkFrame lane, and the ffmpeg-fallback feature with
everything behind it. DrmFrameGuard collapses from an enum to a newtype, which
removes an unsafe impl Send. Roughly 25,000 lines.

Then the CI, packaging, licensing and docs work the plan's §6 lists: the
Windows workflows lose FFMPEG_DIR, PF_FFVK_VULKAN_INCLUDE and their PATH
prepend; the MSIX loses its DLL wildcard; the client .deb stops emitting libav
sonames on its own because depends come from dpkg-shlibdeps; arch, flatpak and
nix drop the dependency; and the README's "FFmpeg 7 or 8" contract narrows to
the host.

Three defects reached users' machines in the first cut, and none was in the
deletion itself.

All three desktop Settings UIs offer vulkan, vaapi and d3d11va as stored
decoder values, so those strings sit in shipped settings files today. Refusing
them by name — which is the correct rule for a stale pin — would have bricked
every upgraded client whose owner ever touched that dropdown. They now migrate
onto the native rung for the same hardware family, at decoder construction AND
at each dialog's lookup, because a legacy value that matches no preset
displays as "Automatic" and silently rewrites the user's preference on the
next save.

M9's evidence filter was deleted on the argument that with no libavcodec twin
below, barring an unproven rung removes hardware decode rather than moving
down one rung. That is true on Windows and false on Linux for Intel and every
unknown vendor id, where prefer_vulkan_first is false and the order is
native-vaapi → native-vk: a rung that has decoded nothing anywhere sitting
above one that is 250/250 on three drivers. Every Intel Linux desktop would
have moved from libavcodec VAAPI, shipping for years, onto pf-vaadec by
default — and a rung that constructs and then produces wrong pixels leaves
only by the error-streak demotion, which this codebase already documents as
not tripping on the B580's strobing. The filter is restored as a narrow, pure,
testable rule: an unproven rung yields to a proven one, and to nothing else.
Windows deliberately passes no rung below, because that vendor family is the
one with a measured wrong-pixel report against Vulkan decode, and trading no
evidence for evidence of corruption is the wrong direction.

And the notices still said FFmpeg was bundled. The root file is what both
desktop clients include_str! and what the MSIX ships, three lines under the
new card saying no FFmpeg is bundled; Apple's Acknowledgements said it too, on
iOS, tvOS and macOS. The generator now emits four per-client files scoped by
transitive closure — 0 FFmpeg mentions in each, verified — while the root file
keeps it for the host. That also ends the standing false attribution of
ffmpeg-next, GTK4, windows-rs and the NVENC SDK to an iPhone.

Windows has no reachable box, so it was compiled instead: a cross clippy at
-D warnings on x86_64 and aarch64-pc-windows-msvc with the C toolchain stubbed
so build scripts run without linking. That gate immediately caught an
include_str! path one directory too deep, which nothing else could have.

Gates: container clippy -D warnings, 160 tests, workspace check, both Windows
targets clean, client ffmpeg count 0 and host 2. The four decode crates are
untouched, so the hardware rungs' 250/250 stands.

⚠ Owed and unrun: no GPU has executed any of this milestone. M8's on-glass
software check, M7's D3D11 and VAAPI AV1 hardware legs, and M9's field bake
all still want hardware, and the bake window and criteria remain the user's.
2026-08-07 08:58:47 +02:00

257 lines
16 KiB
YAML
Raw Blame History

# Flatpak manifest for the native punktfunk Linux client: the shell (crate
# punktfunk-client-linux, binary `punktfunk-client`) PLUS the Vulkan session binary (crate
# punktfunk-client-session, binary `punktfunk-session`) that the shell execs for
# `--connect`/`--browse` — both must ship or streaming from the Deck breaks (the Decky
# wrapper launches the shell with those flags and the shell resolves its sibling
# `punktfunk-session`, see clients/linux/src/spawn.rs). Built into a single-file `.flatpak`
# bundle and published to Gitea's generic package registry (see .gitea/workflows/flatpak.yml
# + packaging/flatpak/README.md).
#
# Why flatpak for the CLIENT (the host stays an RPM/deb — see packaging/README.md "Why not
# Flatpak"): on SteamOS the Steam Deck's /usr is read-only and image-based, so a bare
# `punktfunk-client` binary in ~/.local/bin can't bring its own libadwaita / SDL3 (both
# MISSING from the SteamOS system) — but flatpak is the Deck's native, update-survivable app
# path (the user already runs Moonlight + chiaki-ng as flatpaks). Unlike the host, the client
# is sandbox-friendly: it only needs the GPU render node, the host PipeWire socket, the
# network, Wayland, hidraw for DualSense, and its config dir — all expressible as finish-args.
#
# Runtime: GNOME 50 ships GTK 4.20 and libadwaita 1.8 — both far exceed the crate floors
# (gtk4 0.11 "v4_16", libadwaita 0.9 "v1_5"). GNOME 50 is built on freedesktop-sdk 25.08, so
# `org.freedesktop.Sdk.Extension.rust-stable` resolves to //25.08 (rustc 1.96 — the GTK4 dep
# chain, e.g. pango-sys 0.22, needs >= 1.92, which the older GNOME-48/24.08 rust-stable at 1.89
# could NOT satisfy). GNOME 50 is also a *supported* runtime (GNOME 48 went EOL in March 2026).
# libopus and the PipeWire client lib are in the freedesktop base; SDL3 is NOT, so it is built
# from source as a bundled module.
#
# HEVC decode needs NOTHING from the runtime since M10 (design/client-native-decode.md §6):
# the client links no FFmpeg at all. It decodes on the user's own GPU drivers — Vulkan Video
# through the loader in the GL runtime, or VAAPI through libva, which is dlopen'd and never
# linked — with openh264 + rav1d (both vendored, both BSD-2) as the CPU floor. Nothing here
# depends on the base runtime's stripped libavcodec, and nothing depends on
# `org.freedesktop.Platform.codecs-extra` shadowing it either; that whole arrangement (and the
# encumbered-codec question it existed to answer) is simply not this app's problem any more.
#
# ⚠ The one thing that follows: the CPU floor has no HEVC — no permissively-licensed HEVC
# decoder exists — so an HEVC session on a box whose GPU cannot decode HEVC reconnects on a
# codec this client can finish (pf-client-core's `last_rung_verdict`) instead of falling back
# to software. That is a refusal by design, not a packaging gap.
app-id: io.unom.Punktfunk
runtime: org.gnome.Platform
runtime-version: '50'
sdk: org.gnome.Sdk
# Build-time SDK extensions:
# - rust-stable: cargo/rustc 1.96 + the bundled mold linker (/usr/lib/sdk/rust-stable/bin).
# - llvm20: provides libclang (/usr/lib/sdk/llvm20/lib), which bindgen needs — sdl3-sys and
# pyrowave-sys generate their FFI bindings via bindgen at build time. The base SDK ships no
# clang/libclang, so without this the build panics ("Unable to find libclang").
# Both are added to PATH / LIBCLANG_PATH in build-options below.
sdk-extensions:
- org.freedesktop.Sdk.Extension.rust-stable
- org.freedesktop.Sdk.Extension.llvm20
command: punktfunk-client
cleanup:
- /include
- /lib/pkgconfig
- /lib/cmake
- /share/aclocal
- /man
- /share/man
- '*.a'
- '*.la'
finish-args:
# --- display ---
- --socket=wayland # GTK4 native Wayland window (the client is Wayland-first)
- --socket=fallback-x11 # Xwayland fallback when no Wayland socket is exposed
- --share=ipc # required alongside X11 for shared-memory surfaces
# --- GPU + all input devices ---
# --device=all (not just --device=dri): covers the GPU render node (VAAPI HEVC decode + GL),
# evdev joysticks, AND the hidraw CHAR devices SDL3's HIDAPI needs for DualSense touchpad/
# motion/adaptive-triggers/lightbar. flatpak cannot bind individual /dev/hidrawN via
# --filesystem (they are char devices — "unsupported type 0o20000"), and there is no granular
# --device=hidraw; --device=all is what game/emulator flatpaks (RetroArch, Dolphin) use. We
# self-host via the Gitea generic registry — NOT Flathub — so its --device=all review rule
# does not apply.
- --device=all
- --filesystem=/run/udev:ro # SDL/HIDAPI enumerates devices via udev
# --- audio: the client speaks the NATIVE PipeWire protocol (audio.rs `pw connect`), NOT the
# PulseAudio shim — so it needs the real `pipewire-0` socket in the sandbox. With only
# --socket=pulseaudio the sandbox has just `pulse/native`, no `pipewire-0`, and playback +
# mic both die with "pw connect (is PipeWire running in this session?)" (observed live on the
# Deck in Gaming Mode). We bind the native socket via --filesystem=xdg-run/pipewire-0 (NOT
# --socket=pipewire: this flatpak-builder toolchain rejects it as an "Unknown socket type",
# and the Deck's flatpak 1.16 override CLI does too — the filesystem bind is the portable
# form, validated on-Deck to make pipewire-0 appear + the client register its audio node).
# --socket=pulseaudio stays as a fallback for any pulse-only path. ---
- --filesystem=xdg-run/pipewire-0
- --socket=pulseaudio
# --- network: QUIC control + UDP data plane + mDNS discovery (_punktfunk._udp) ---
- --share=network
# --- persistent client identity / pairing store (shared with punktfunk-probe) ---
- --filesystem=~/.config/punktfunk:create # client-{cert,key}.pem, known-hosts, settings
# --- HDR under gamescope (Steam Deck Game Mode) ---
# A flatpak's Vulkan loader can't see the host's gamescope WSI layer, so the SDL3 surface never
# offers the HDR10 (ST.2084) colorspace and the presenter silently tone-maps PQ->SDR — the
# Game-Mode HDR indicator stays dark (verified on a Deck OLED: the sandbox loader found NO
# frog/gamescope layer). The layer ships as the runtime extension
# `org.freedesktop.Platform.VulkanLayer.gamescope`, which org.gnome.Platform//50 auto-mounts at
# /usr/lib/extensions/vulkan/gamescope once installed (a one-time, per-Deck step; keep it in the
# Decky plugin's setup / docs):
# flatpak install --user -y flathub org.freedesktop.Platform.VulkanLayer.gamescope//25.08
# THREE things are needed, not two (verified live on a Deck OLED — the env vars alone left
# hdr10_format=None). (1) VK_ADD_IMPLICIT_LAYER_PATH puts the layer's implicit-layer JSON on the
# Vulkan loader's search path (the runtime point mounts the files but not onto the path). (2)
# ENABLE_GAMESCOPE_WSI flips the layer's own `enable_environment` gate. (3) The layer, once
# loaded, must open a *Wayland* connection to gamescope's private socket ($GAMESCOPE_WAYLAND_DISPLAY
# = gamescope-0) to negotiate the HDR10 colorspace via the gamescope_swapchain protocol — but the
# Deck runs games as X11 clients (DISPLAY=:1, no WAYLAND_DISPLAY exported), so --socket=wayland
# binds nothing and that socket never enters the sandbox. Without it the layer loads, maps, and
# silently can't reach the compositor → no HDR10 offered → PQ tone-mapped to SDR, badge dark.
# Binding xdg-run/gamescope-0 is the missing half (chiaki-ng does the same). With all three the
# surface offers HDR10 and the presenter's existing HDR10 swapchain path engages — no client code
# change. Harmless off-Deck: the layer no-ops when there's no gamescope socket to bind.
- --env=VK_ADD_IMPLICIT_LAYER_PATH=/usr/lib/extensions/vulkan/gamescope/share/vulkan/implicit_layer.d
- --env=ENABLE_GAMESCOPE_WSI=1
- --filesystem=xdg-run/gamescope-0 # gamescope's private Wayland socket (HDR negotiation)
build-options:
append-path: /usr/lib/sdk/rust-stable/bin:/usr/lib/sdk/llvm20/bin
# The rust build resolves everything via pkg-config: gtk4/libadwaita/pipewire/opus AND a
# the SDK's own .pc files from the multiarch /usr dir (libopus for audiopus_sys, the
# PipeWire client lib, GTK4/libadwaita), plus the bundled SDL3's .pc from /app.
env:
PKG_CONFIG_PATH: /app/lib/pkgconfig:/usr/lib/x86_64-linux-gnu/pkgconfig:/usr/lib/pkgconfig
# bindgen (sdl3-sys, pyrowave-sys) loads libclang from the llvm20 extension.
LIBCLANG_PATH: /usr/lib/sdk/llvm20/lib
# mold (shipped in rust-stable) speeds the ~450-crate link on the Deck APU.
RUSTFLAGS: -C link-arg=-fuse-ld=mold
# The multiarch directory in PKG_CONFIG_PATH is per-architecture, and flatpak-builder does NOT
# shell-expand `env` values — so ${FLATPAK_ARCH} would be taken literally. An `arch` override is
# the supported way to vary it; everything else above is inherited. Only the runtime's own
# /usr/lib/<triple> changes, so aarch64 differs from x86_64 in exactly this one string.
arch:
aarch64:
env:
PKG_CONFIG_PATH: /app/lib/pkgconfig:/usr/lib/aarch64-linux-gnu/pkgconfig:/usr/lib/pkgconfig
modules:
# ---------------------------------------------------------------------------------------
# SDL3 — NOT provided as a linkable libSDL3.so.0 by org.gnome.Platform/freedesktop-sdk
# 25.08, and there is no SDL3 recipe in flathub/shared-modules. Build it from source.
# Pinned to 3.4.10 to match the crate exactly: sdl3-sys is `0.6.6+SDL-3.4.10`, i.e. its
# bindings target SDL 3.4.10 — building an older SDL risks missing symbols at link time.
# HIDAPI is enabled (DualSense touchpad/motion/triggers/lightbar over hidraw).
# ---------------------------------------------------------------------------------------
- name: sdl3
buildsystem: cmake-ninja
config-opts:
- -DCMAKE_BUILD_TYPE=Release
- -DSDL_SHARED=ON
- -DSDL_STATIC=OFF
- -DSDL_HIDAPI=ON # DualSense full fidelity over hidraw
- -DSDL_TEST_LIBRARY=OFF
- -DSDL_EXAMPLES=OFF
sources:
- type: archive
url: https://github.com/libsdl-org/SDL/releases/download/release-3.4.10/SDL3-3.4.10.tar.gz
# `sha256sum SDL3-3.4.10.tar.gz` (verified 2026-06-15). Bump url + sha together.
sha256: 12b34280415ec8418c864408b93d008a20a6530687ee613d60bfbd20411f2785
x-checker-data:
type: anitya
project-id: 4974
stable-only: true
url-template: https://github.com/libsdl-org/SDL/releases/download/release-$version/SDL3-$version.tar.gz
cleanup:
- /bin
- /include
- /lib/cmake
- /lib/pkgconfig
# ---------------------------------------------------------------------------------------
# The client. cargo-sources.json is the GENERATED offline crate cache:
# python3 flatpak-cargo-generator.py Cargo.lock -o packaging/flatpak/cargo-sources.json
# (run from the repo root; the CI step does exactly this). With it present the build is fully
# offline (CARGO_NET_OFFLINE). For quick LOCAL iteration WITHOUT regenerating it, drop the
# cargo-sources.json source and pass --build-args=--share=network to flatpak-builder
# (non-reproducible; cargo fetches from crates.io during the build).
# ---------------------------------------------------------------------------------------
- name: punktfunk-client
buildsystem: simple
build-options:
env:
CARGO_HOME: /run/build/punktfunk-client/cargo
CARGO_NET_OFFLINE: 'true'
# The session binary's Skia (pf-console-ui → skia-safe) fetches prebuilt libskia in
# its build script — dead in the offline sandbox. skia-bindings accepts a file://
# override (read directly, no curl), so we point it at the archive that
# flatpak-builder pre-downloaded as a pinned source below. No {tag}/{key}
# placeholders needed: a template without them is used verbatim.
SKIA_BINARIES_URL: file:///run/build/punktfunk-client/skia-binaries.tar.gz
build-commands:
# Drop every reference to the windows-rs GIT dependency before building. That git
# source is deliberately NOT vendored into cargo-sources.json — see
# packaging/flatpak/prune-windows-lock.py, which strips it so flatpak-builder doesn't
# full-clone the multi-GB windows-rs repo. But `cargo --offline` needs every DECLARED
# dependency's source just to build the unit graph (cfg(windows)-gated or not), so any
# manifest still pinning it fails with "can't checkout ... you are in the offline
# mode". Two declare it: the Windows client (removed from the workspace members) and
# pf-client-core's D3D11VA backend (entry pruned in place — nothing cfg(windows)
# compiles here). Both edits touch only the sandbox copies. (No --locked: the lock no
# longer matches the reduced graph; --offline still pins every crate to the vendored
# cargo-sources.json, so the build stays reproducible.)
- sed -i '\#"clients/windows",#d' Cargo.toml
- python3 packaging/flatpak/prune-windows-toml.py crates/pf-client-core/Cargo.toml
# One cargo invocation for both binaries: the shell and the Vulkan session binary it
# execs for --connect/--browse. Building them together keeps skia-bindings' feature
# unification identical to a workspace build (the pinned skia-binaries archive's
# `pdf-textlayout-vulkan` key must match the resolved feature set).
- cargo --offline build --release -p punktfunk-client-linux -p punktfunk-client-session
-p punktfunk-cli
- install -Dm0755 target/release/punktfunk-client ${FLATPAK_DEST}/bin/punktfunk-client
- install -Dm0755 target/release/punktfunk-session ${FLATPAK_DEST}/bin/punktfunk-session
# The headless CLI (design/client-architecture-split.md §4) — reachable as
# `flatpak run --command=punktfunk io.unom.Punktfunk …`.
- install -Dm0755 target/release/punktfunk ${FLATPAK_DEST}/bin/punktfunk
# Desktop entry (renamed to the app id; Exec is the in-sandbox binary).
- install -Dm0644 packaging/flatpak/io.unom.Punktfunk.desktop
${FLATPAK_DEST}/share/applications/io.unom.Punktfunk.desktop
# AppStream metainfo (required for a well-formed flatpak / Software listings).
- install -Dm0644 packaging/flatpak/io.unom.Punktfunk.metainfo.xml
${FLATPAK_DEST}/share/metainfo/io.unom.Punktfunk.metainfo.xml
# Scalable icon named for the app id (GNOME runtime renders SVG via librsvg).
# One icon source for every Linux packaging (it used to live under packaging/flatpak,
# which is why the deb/rpm/arch entries shipped a stock `video-display` instead).
- install -Dm0644 packaging/linux/icons/hicolor/scalable/apps/io.unom.Punktfunk.svg
${FLATPAK_DEST}/share/icons/hicolor/scalable/apps/io.unom.Punktfunk.svg
sources:
# The repo checkout. For a Flathub/published build, replace with a pinned git source:
# - type: git
# url: https://git.unom.io/unom/punktfunk
# tag: vX.Y.Z
# commit: <sha>
# For ON-DECK / CI builds we build the checked-out working tree in place:
- type: dir
path: ../..
# Generated offline crate cache (see the comment block above). Remove for --share=network.
- cargo-sources.json
# Prebuilt Skia for skia-bindings 0.87.0 (SKIA_BINARIES_URL above). The key encodes the
# crate's pinned skia commit + target + resolved features — when bumping skia-safe, build
# the workspace once locally and take the new name from
# target/release/build/skia-bindings-*/out/.cache/, then update url + sha256 together.
# One entry per architecture, selected by `only-arches`; both land on the SAME
# dest-filename, so SKIA_BINARIES_URL above stays a single literal path. The upstream
# release publishes the aarch64 archive under the same skia commit hash and the same
# resolved-feature key, so the two stay in lockstep on a skia-safe bump — take both new
# names and sha256s together.
- type: file
only-arches: [x86_64]
url: https://github.com/rust-skia/skia-binaries/releases/download/0.87.0/skia-binaries-e551f334ad5cbdf43abf-x86_64-unknown-linux-gnu-pdf-textlayout-vulkan.tar.gz
sha256: b46e7061e6b9df792025acaf9b8b90180224c7cec63f4c3ce09af7ddddb8abfa
dest-filename: skia-binaries.tar.gz
- type: file
only-arches: [aarch64]
url: https://github.com/rust-skia/skia-binaries/releases/download/0.87.0/skia-binaries-e551f334ad5cbdf43abf-aarch64-unknown-linux-gnu-pdf-textlayout-vulkan.tar.gz
sha256: 6994fb993064d7d4fff00bf7c79544ffc881bb5441b6768b31cc088540f4bef7
dest-filename: skia-binaries.tar.gz
Reference in New Issue View Git Blame Copy Permalink