Files

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

12 KiB

Raw Blame History

punktfunk client — Flatpak (Steam Deck / SteamOS, and any flatpak distro)

The native Linux client — the shell (crate punktfunk-client-linux, binary punktfunk-client) plus the Vulkan session binary it execs for streaming (crate punktfunk-client-session, binary punktfunk-session) — is published two ways by CI (.gitea/workflows/flatpak.yml), on every push to main (a rolling <next-minor>-ciN.g<sha> build, base derived from the latest stable tag by scripts/ci/pf-version.sh) and on v* tags (a clean X.Y.Z):

Hosted OSTree repo at https://flatpak.unom.io (recommended) — a GPG-signed Flatpak remote served by a static Caddy container on unom-1, so users install once and then flatpak update. Shared unom-wide repo (remote name unom), reusable by other unom apps under the same signing key. See "Install (recommended)" below.
Single-file .flatpak bundle in Gitea's generic package registry (unom org) — the no-remote fallback the Decky plugin consumes (stable latest/punktfunk-client.flatpak URL) and the offline/manual path. On tags it's also attached to the Gitea release.

The host is NOT a flatpak (it needs unsandboxed /dev/uinput + zero-copy NVENC — see ../README.md "Why not Flatpak"). Only the client is sandbox-friendly.

Why flatpak for the Steam Deck

SteamOS /usr is read-only and image-based, and the system is missing libadwaita and libSDL3 — so a bare punktfunk-client binary dropped into ~/.local/bin won't run. Flatpak is the Deck's native, update-survivable app path (the user already runs Moonlight and chiaki-ng as flatpaks), and the bundle carries libadwaita (from org.gnome.Platform//50) + a bundled SDL3. It carries no FFmpeg: since M10 the client decodes on the user's own GPU drivers (Vulkan Video, VAAPI) with openh264 + rav1d as the CPU floor, so the runtime's codecs-extra extension — and the encumbered-codec question it answered — no longer enter into it.

App id: io.unom.Punktfunk (matches the Apple bundle id family and the Decky plugin's flatpak fallback).

Install (recommended): the hosted repo

One command adds the signed unom remote and installs the client; it auto-adds Flathub for the GNOME runtime, and flatpak update tracks new builds from then on:

flatpak install --user https://flatpak.unom.io/io.unom.Punktfunk.flatpakref
flatpak run io.unom.Punktfunk

Equivalent two-step (add the whole remote, then install by app id):

flatpak remote-add --user --if-not-exists unom https://flatpak.unom.io/unom.flatpakrepo
flatpak install --user unom io.unom.Punktfunk

Updates — the whole point of the hosted repo:

flatpak update                    # or: flatpak update io.unom.Punktfunk

Install on the Deck via the bundle (no-remote fallback)

The generic registry is a plain HTTP file store, so just download the bundle and install it per-user (no root, survives SteamOS updates). This is what the Decky plugin uses; the hosted repo above is the better path for a human on the Deck:

# Pick a version: a tag like 1.2.3, or the newest main build's <next-minor>-ciN.gSHA.
VER=1.2.3
URL="https://git.unom.io/api/packages/unom/generic/punktfunk-client-flatpak/$VER/punktfunk-client-$VER.flatpak"

# Flathub must be enabled (it is on the Deck) so the GNOME runtime pulls in:
flatpak remote-add --user --if-not-exists flathub https://dl.flathub.org/repo/flathub.flatpakrepo

curl -fL -o /tmp/punktfunk-client.flatpak "$URL"
flatpak install --user --bundle /tmp/punktfunk-client.flatpak

Run it:

flatpak run io.unom.Punktfunk                 # GUI host list (mDNS)
flatpak run io.unom.Punktfunk --connect HOST:PORT

The Decky plugin launches exactly this (flatpak run io.unom.Punktfunk --connect …) once installed — see ../../clients/decky/README.md.

Updating the bundle install

If you installed from the bundle (not the hosted repo), it has no remote to track, so updates are "download the newer bundle and reinstall":

flatpak install --user --bundle /tmp/punktfunk-client.flatpak   # same command, newer file

Installs from https://flatpak.unom.io instead just take flatpak update (see "Install (recommended)" above).

Build locally / the CI fallback

CI builds this in a --privileged Fedora container, because flatpak-builder runs bubblewrap, which needs user namespaces the default Docker executor denies. If the Gitea runner can't grant --privileged (the job fails at flatpak-builder with "Creating new namespace failed: Operation not permitted"), build it out-of-band and upload by hand. The easiest place is on the Deck itself (it can run org.flatpak.Builder user-scope, no root):

# On the Deck (or any flatpak box), one-time:
flatpak install --user -y flathub org.flatpak.Builder

# build-flatpak.sh auto-detects org.flatpak.Builder, generates cargo-sources.json (or reuses an
# existing one — see below), builds, and exports dist/punktfunk-client-<version>.flatpak:
bash packaging/flatpak/build-flatpak.sh

# Upload to the generic registry (PAT with write:package):
curl -fsS --user "enricobuehler:$REGISTRY_TOKEN" \
  --upload-file dist/punktfunk-client-*.flatpak \
  "https://git.unom.io/api/packages/unom/generic/punktfunk-client-flatpak/0.0.1-manual/punktfunk-client.flatpak"

cargo-sources.json generation needs python3 + aiohttp + tomlkit, which the Deck lacks. Generate it on a dev box (build-flatpak.sh does it, or run the upstream flatpak-cargo-generator.py Cargo.lock -o packaging/flatpak/cargo-sources.json), rsync it next to the manifest, and build-flatpak.sh reuses it (it only regenerates when the file is absent or FORCE_GEN=1).

The Mac build host cannot build a Linux flatpak (no flatpak-builder for macOS), and home-worker-2 has no flatpak and no passwordless sudo to install it — so the Deck or the privileged CI container are the only two viable build sites.

aarch64

The manifest builds for aarch64 as well as x86_64. Two things are architecture-specific, and both are now expressed properly rather than hardcoded:

PKG_CONFIG_PATH contains the runtime's multiarch directory. flatpak-builder does not shell-expand env values, so ${FLATPAK_ARCH} would be taken literally — a build-options.arch override supplies the aarch64 string instead, inheriting everything else.
The prebuilt Skia archive is per-target and pinned by sha256. There are now two type: file sources discriminated by only-arches, both landing on the same dest-filename, so SKIA_BINARIES_URL stays one literal path. Upstream publishes the aarch64 archive under the same skia commit hash and the same resolved-feature key (pdf-textlayout-vulkan), so on a skia-safe bump update both URLs and both hashes together.

ARCH=aarch64 bash packaging/flatpak/build-flatpak.sh
# -> dist/punktfunk-client-<version>-aarch64.flatpak

ARCH defaults to this machine's, and the bundle name now carries the architecture so an x86_64 and an aarch64 build can coexist in dist/. This is not a cross-compile: flatpak-builder runs the build in a sandbox for the target arch, so building aarch64 anywhere but an arm64 machine needs qemu binfmt and is very slow. Not yet verified end to end — the manifest is correct by construction and the Skia hash was checked against the published archive, but no aarch64 flatpak has been built.

Manifest

io.unom.Punktfunk.yml. Runtime org.gnome.Platform//50 (GTK 4.20 + libadwaita 1.8 ≥ the crate floors of v4_16 / v1_5), built on freedesktop-sdk 25.08, with two build-time SDK extensions: org.freedesktop.Sdk.Extension.rust-stable (→ //25.08, rustc 1.96 — the GTK4 dep chain, e.g. pango-sys 0.22, needs ≥ 1.92, which the EOL GNOME-48 / 24.08 rust-stable at 1.89 could not provide) and org.freedesktop.Sdk.Extension.llvm20 (libclang, needed by bindgen in sdl3-sys / pyrowave-sys). No libavcodec at any layer — the client links no FFmpeg since M10, so neither the SDK's stripped build nor the runtime's codecs-extra shadow of it is involved; HEVC decodes on the GPU's own driver, and there is deliberately no software HEVC rung (see the manifest header). A bundled SDL3 3.4.10 module (pinned to match sdl3-sys 0.6.6+SDL-3.4.10), and finish-args for Wayland + --device=all (GPU/VAAPI render node + evdev + the hidraw char-devices SDL3 needs for DualSense)

--socket=pulseaudio (PipeWire-pulse: playback + mic) + --share=network. Alongside it: io.unom.Punktfunk.desktop, io.unom.Punktfunk.metainfo.xml, io.unom.Punktfunk.svg (all installed by the manifest). No vulkan-headers module: it existed for pf-ffvk's bindgen over FFmpeg's hwcontext_vulkan.h, and ash generates its own bindings and dlopens the loader. cargo-sources.json (the offline crate cache) is a pure function of Cargo.lock; CI regenerates it each build and it is gitignored — generate it on any box with network + python3/aiohttp/tomlkit (build-flatpak.sh does this automatically) and, for a build host that lacks those (the Deck), rsync the generated file in alongside the manifest.

Offline Skia: the session binary's Skia console UI (pf-console-ui → skia-safe) normally downloads prebuilt libskia binaries at build time, which is dead in the offline sandbox — so the manifest pins a skia-binaries-….tar.gz source and points the build at it with SKIA_BINARIES_URL: file://…. When bumping the skia-safe/skia-bindings crate version, update that pinned tarball (URL + sha256) to the matching skia-binaries release or the build breaks offline.

Hosting the repo (unom-1) + one-time setup

The OSTree repo flatpak-builder produces is GPG-signed in CI and rsynced to unom-1, where a tiny static Caddy container (server/compose.production.yml + server/Caddyfile, port 3230) serves the ./site tree (repo/ + unom.flatpakrepo + io.unom.Punktfunk.flatpakref + index.html). The edge Caddy on home-reverse-proxy-1 fronts it at https://flatpak.unom.io. The CI deploy step no-ops until the secret + infra exist, so it won't redden builds mid-setup.

Signing key: dedicated RSA-4096 key unom Flatpak Repo <flatpak@unom.io>. Public key committed at unom-flatpak.gpg (its base64 goes into the .flatpakrepo/.flatpakref GPGKey=); private key (ASCII-armored, then base64) lives only in the CI secret.

One-time setup (mirrors any new unom DMZ service — see the deploy-infra notes):

Secret FLATPAK_GPG_PRIVATE_KEY on this repo = base64 of the armored private key (gpg --armor --export-secret-keys <fpr> | base64 -w0). DEPLOY_* + REGISTRY_TOKEN already exist.
Edge Caddy on home-reverse-proxy-1 (/home/caddy/caddy/Caddyfile, apply by hand + ./reload.sh): flatpak.unom.io { reverse_proxy 192.168.50.50:3230 }
Port allowlist: add 3230 to caddy_target_ports in unom/infra (proxmox/unom-1) + terraform apply.
DNS: ensure flatpak.unom.io resolves to the edge proxy.

Re-signing/rotation: regenerate the key, replace unom-flatpak.gpg + the secret; every client must re-add the remote (the GPGKey changed), so rotate rarely.

Alternatives considered

Hosted OSTree repo (chosen): the only option that gives flatpak update. We self-host the static tree on unom-1 behind Caddy (Gitea has no flatpak/ostree registry); the build already produces the repo, so the marginal cost is GPG signing + an rsync + a 10-line static container.
Generic registry bundle (kept as fallback): one curl to publish, one flatpak install --bundle to consume; mirrors the deb/rpm curl-upload pattern. No auto-update — this is what the Decky plugin pulls (stable latest/punktfunk-client.flatpak), plus the offline/manual path.
Release attachment: also done on tags, good for a human-facing download page.
Flathub (deferred): best discoverability + zero hosting, but a separate submission/review process and less control; revisit once the client is past scaffold quality.

12 KiB Raw Blame History