Live-debugged on home-runner-1: confirmed the dnf install in Tooling does
NOT actually rewrite /etc/nsswitch.conf (no authselect trigger fires), so
the nss-resolve sed from 68b5376f/5f687a70 was never the actual fix.
gitea-runner-fleet on this box is a shared, resource-capped fleet (3
replicas, --cpus 5/--memory 7g each, on 16c/24G) serving punktfunk AND
several other orgs' repos concurrently. A push to main fans out ~8
punktfunk workflows at once on top of that other traffic, and the box's
Docker embedded DNS resolver (127.0.0.11) drops UDP lookups under the
combined load — exactly what retry.sh's own header already documented.
Manually dispatching this job while the box was idle succeeded instantly,
with or without the sed. retry.sh's 5-attempt (~100s) budget isn't always
enough to outlast a synchronized multi-org burst; bumped the three
flathub-facing calls (remote-add, install-deps-only, download-only) to 10
attempts (~9min). Verified end-to-end on a live re-run (run 8993) after
this change: full success including publish + OSTree deploy.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
When unom/infra deploy-all provisions a new unom-1, it dispatches this
deploy with the just-created box IP. Falls back to the DEPLOY_HOST secret
on push-triggered runs, so existing behavior is unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The earlier fix (5f687a70) dropped `resolve` from nsswitch.conf before the
Tooling step's dnf install. But that install pulls flatpak's recommended
xdg-desktop-portal -> pipewire/wireplumber, which drags in a systemd
package upgrade whose RPM trigger re-runs authselect and regenerates
nsswitch.conf — silently re-adding the `resolve` entry and clobbering the
fix before `flatpak remote-add` ever runs. Reapply the sed right after the
dnf install, immediately before the first flatpak network call.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
docker.yml (5-image matrix) and flatpak.yml (full flatpak-builder) couple the
docs + flatpak-server deploys to heavy builds. This adds a dispatchable,
build-free deploy-services.yml that just (re)places the compose files and pulls
the already-published images, so unom/infra's deploy-all can bring a fresh
unom-1 fully up in one dispatch. The flatpak ./site OSTree content is restored
from the unom-1 backup or regenerated by flatpak.yml's next publish.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
fedora:43 nsswitch routes host lookups through the resolve module
(systemd-resolved), absent in a CI container. git/curl/dnf fall through
to the dns module (Docker 127.0.0.11); flatpak/ostree's resolver trips
[!UNAVAIL=return] and dies with '[6] Could not resolve hostname'. Was
masked as flaky 'busy runner drops DNS' + retry.sh, but it's
deterministic on runners where the resolve module tips that way (started
failing when jobs landed on the newly-added home-runner-2). Drop the
resolve entry -> plain dns lookups.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The flatpak run dies most pushes at `flatpak remote-add` with an instant
"[6] Could not resolve hostname" (runs 8755/8702/8689), 25ms after dnf pulled
329 packages fine — and deploy-docs separately hits "dial tcp: i/o timeout"
to unom-1 (8716/8679). The busy runner drops UDP DNS + TCP dials under
parallel-job load; tools with built-in retries (dnf) ride it out, single-shot
fetches killed the whole 2h build slot.
* scripts/ci/retry.sh: linear-backoff wrapper (5 tries, attempt*10s); stdout
passes through untouched so $(…) capture works.
* Tooling: retry the flathub remote-add.
* Build split: a retried prefetch phase (--install-deps-only, then
--download-only for every crate in cargo-sources.json) warms the state dir;
the long compile then runs with no network left to flake on.
* Seed step: probe (retried) whether the server repo exists — ONLY first
publish may continue fresh. The old blanket `rsync || warn` swallowed
transient failures too, producing the single-branch summary that clobbers
the other channel (the exact bug the step's comment documents).
* Deploy: retry the idempotent ssh/rsync calls to unom-1.
* raw.githubusercontent curl: --retry 5 --retry-all-errors.
docker.yml's deploy-docs has the same disease but sits on drone-ssh; left
for a follow-up. Real root cause is runner-host networking (conntrack /
upstream DNS under burst load) — retries make the jobs indifferent to it.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Address findings from a repo security review:
- core: add a sliding-window anti-replay filter over the AEAD-authenticated
sequence in Session (poll_input/poll_frame), closing the input-replay gap the
data plane previously left to the LAN/VPN trust assumption. 4096-deep window,
unit-tested; the encrypted loopback suite confirms no false drops.
- clients: write the mTLS client private key 0600 and lock the config dir 0700
on Unix (it was world-readable at the umask default), re-locking existing
stores on load. pf-client-core::trust plus the probe's own identity writer.
Windows keeps the %APPDATA% ACL; Android/Apple already wrap the key.
- web: fix a post-login open redirect — resolve `next` via URL and require it to
stay same-origin, rejecting `/\evil.com` and tab/encoding variants the old
`!startsWith("//")` guard missed. Also fixes the dead safeNextPath helper.
- ci: SHA-256-pin the BtbN FFmpeg DLLs bundled into the signed Windows installer
(were fetched from the rolling `latest` tag unverified); fails closed on a
re-roll, matching the VB-CABLE gate.
- ci: fail-open fork-guard on the Windows/Apple host-mode PR build jobs that
share runner labels with the signing jobs. Definitive fix stays server-side
(Gitea outside-collaborator approval / isolated PR runners) — see the notes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The punktfunk-session Vulkan client (clients/linux-session, now clients/session)
builds and runs on Windows; the WinUI shell spawns it for every stream. Verified
live: 10-bit HEVC via Vulkan Video on both AMD (iGPU) and NVIDIA, 5120x1440 at
130 fps / 8 ms end-to-end on the RTX 4090.
- pf-ffvk: Windows bindgen branch (FFMPEG_DIR + PF_FFVK_VULKAN_INCLUDE, no
pkg-config); provisioning fetches Vulkan-Headers (pinned v1.4.309).
- pf-client-core: builds on Windows — WASAPI audio (audio_wasapi.rs, cfg-swapped
via #[path], same surface as the PipeWire twin), VAAPI/dmabuf gated inline
(chain = vulkan -> software), trust reads the WinUI shell's %APPDATA% stores
(parity tests pin both serialized shapes), Settings gains adapter/hdr_enabled
(serde-defaulted; Linux stores unaffected).
- pf-presenter: builds on Windows — dmabuf module Linux-gated; SDL keyboard grab
while captured (Alt+Tab/Win reach the host); pick_device ranks discrete over
integrated (device 0 was the iGPU on hybrid boxes — the silent footgun) and
honors PUNKTFUNK_VK_ADAPTER (the Settings GPU pick, exported by the session).
- run loop: block in one SDL wait woken by input AND decoded frames (a per-
session forwarder pushes a FrameWake user event) instead of a 1 ms poll —
measured 111%% -> 5%% of a core (NVIDIA), 86%% -> 3.5%% (AMD), stats unchanged.
The pump's decode-fence wait became once-per-window sampling (no per-frame
pipeline stall; the stat now shows true backlog).
- pf-console-ui: builds on Windows (skia-safe msvc prebuilts); font lookup falls
through fontconfig aliases to concrete DirectWrite families (Consolas/Segoe UI)
— browse/coverflow works, verified against a live host.
- WinUI shell: session-always via new src/spawn.rs (GTK spawn.rs port —
CREATE_NO_WINDOW, stdout contract, kill handle); the Stream screen is a status
card (chips + stage lines from the child's stats). The legacy in-process
D3D11VA path stays behind Settings "Streaming engine" / PUNKTFUNK_BUILTIN_
STREAM=1 as the A/B baseline until Phase 8 deletes it. SessionParams.video_caps
makes the HDR toggle real.
- clients/linux-session renamed to clients/session (builds for both OSes).
- CI/MSIX: both workflows build/test both bins with widened path filters; the
MSIX ships punktfunk-session.exe. ARM64 session builds --no-default-features
(rust-skia has no aarch64-pc-windows-msvc prebuilts; flip when it does).
A/B on this box (5120x1440 HEVC vs home-worker-5): NVIDIA Vulkan 130 fps / 8 ms
e2e / 1.6 ms decode — clearly better than the built-in path. The AMD iGPU VCN
saturates at ~52 fps where its own D3D11VA does ~70 — Adrenalin Vulkan decode is
slower on APU silicon; discrete RDNA validation gates Phase 8.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The re-architecture split the Linux client into two binaries (shell + Vulkan
session streamer) and added the pf-ffvk crate, whose build.rs runs bindgen over
FFmpeg's libavutil/hwcontext_vulkan.h (#include <vulkan/vulkan.h>). Two release
regressions fell out of that on the desktop-package legs:
1. Missing build dep. The Arch build (archlinux:base-devel + its own pacman list)
fatal-errored at `vulkan/vulkan.h file not found`. Add the Vulkan dev headers
wherever the client compiles from a self-managed dep set: arch (PKGBUILD
makedepends + arch.yml), rpm (spec BuildRequires + rpm.yml + fedora image),
deb.yml (belt-and-suspenders; the rust-ci image already bakes libvulkan-dev),
and the screenshots job. The flatpak builds offline against the GNOME SDK, so
install Vulkan-Headers into /app as a pinned module and point pf-ffvk's bindgen
at it via PF_FFVK_VULKAN_INCLUDE.
2. Only the shell shipped. arch/deb/rpm built and installed just punktfunk-client;
the shell execs its sibling punktfunk-session for a connect, so desktop
streaming would break exactly as Decky's did. Build and install BOTH binaries
in all three, and declare the runtime Vulkan loader the session binary dlopens
(vulkan-icd-loader / vulkan-loader / libvulkan1 — not a DT_NEEDED, so
shlibdeps/auto-requires can't see it).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The re-architected Linux client is now two binaries: the punktfunk-client shell
execs its sibling punktfunk-session (ash/Vulkan presenter + Skia console UI) for
--connect/--browse. The Decky plugin's stream and browse paths launch the shell
with exactly those flags, but the flatpak built and installed only the shell, so
streaming and the gamepad library from the Deck failed at exec with
"punktfunk-session: No such file" (pair/wake/library still worked — the shell
handles them in-process).
Build and install both binaries. The session binary pulls in Skia (skia-safe),
whose build script downloads a prebuilt libskia — dead in the offline sandbox —
so point skia-bindings at a pinned, vendored archive via SKIA_BINARIES_URL=file://
(read directly, no curl); the tarball rides along as a sha256-pinned flatpak
source. Widen the flatpak CI path filters to the session binary's crates
(linux-session, pf-presenter, pf-console-ui, pf-client-core) and fix the moved
library.rs path in the Decky error-classifier comment.
All other plugin↔client contracts (flags, pairing/library output, config files,
env vars, exit codes, the 47990 mgmt port) already match — no changes needed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The APK was only reachable via the generic registry (or attached to a Gitea
Release on a vX.Y.Z tag) — a main-push canary or a PR/dispatch run surfaced no
downloadable APK on the run page itself. Add an upload-artifact step (v3, per
Gitea's GHES-identifying artifact backend, like apple.yml) that grabs whichever
APKs were built — the signed universal release APK on a main/tag push, else the
debug APK — so any run is a one-click sideload download.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Pin android-actions/setup-android, appleboy/scp-action, and
appleboy/ssh-action to commit SHAs (version kept in a trailing comment).
These run in jobs holding the Android signing keystore, Play
service-account, and deploy SSH key, so a moved tag on a third-party
action could exfiltrate them.
- Add a bun-audit job to audit.yml over web/bun.lock — the console holds
the login gate, session sealing, and mgmt token, so its deps matter too
— and trigger the workflow on web/bun.lock changes alongside Cargo.lock.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
rust-toolchain.toml pinned the floating "stable" channel, so the CI image baked whatever
stable existed at image-build time. When the image is rebuilt onto a newer stable,
rustfmt's rules shift and `cargo fmt --all --check` fails on files nobody touched — the
recurring format-drift that keeps red-lighting CI.
Pin channel = "1.96.0" (== today's stable: rustc ac68faa20, rustfmt 1.9.0-stable), the
exact build CI already runs, so this is a no-op now but locks formatting for good: local
dev, the Linux CI image, and the Windows runner all use rustup and honor this file, so
they converge on one rustfmt. Formatting now only changes in a deliberate bump-this-pin-
and-reformat commit. ci.yml cache comment updated to match.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bring the Android client to parity with Apple's gamepad experience and finish
Wake-on-LAN.
- Console/gamepad home: host carousel, aurora chrome, mTLS game-library coverflow,
and an input-aware legend that switches between gamepad face buttons and a
TV-remote select-ring + arrows based on the last-used input.
- Wake-on-LAN: the fire-and-forget send is upgraded to wait-until-up
(WakeController/WakeOverlay: resend + mDNS poll, 90s timeout, cancel/retry,
fingerprint-matched so a host that cold-boots onto a new DHCP IP still connects),
plus host edit (touch dialog + console form) with an auto-filled MAC.
- Android TV: brand banner (android:banner), density-aware console scaling, D-pad/
remote nav (Up = Settings, Down or the pad Select button = host Options),
emergency stream-exit chord, and 120Hz console refresh.
- Touch UI: settings split into subpages with a tablet NavigationRail, axis-aware
tab animation (horizontal on phones, vertical on the tablet rail), animated
settings navigation, and a licenses screen with a back button + the real
workspace version (read from Cargo.toml).
- Vector Lock/controller icons (no emoji); bundled Geist font.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Gitea Arch registry signs its DB + packages, so 'SigLevel = Optional TrustAll' fails
non-interactively (pacman still needs the key to verify). Document the one-time
pacman-key import instead; install is then signature-validated under pacman's default
SigLevel (verified end-to-end: clean archlinux container -> repo sync -> install,
'Validated By: Signature').
Also cache /usr/local/cargo/git in arch.yml: the workspace pulls clients/windows'
git-pinned windows-reactor/windows deps to resolve, cloning windows-rs (huge) every run
otherwise — same registry+git cache deb.yml uses.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New arch.yml builds the split PKGBUILD (host/client/web, PF_WITH_WEB=1) in an
archlinux:base-devel container on every push and publishes to the pacman repos
'punktfunk' (tags) / 'punktfunk-canary' (main, X.Y.Z-0.<run#> — pkgrel allows
only digits+dots, so the run number carries the ordering). Consumers add one
pacman.conf section; no more build-it-yourself as the only Arch path.
PKGBUILD: pkgver/pkgrel env-driven (PF_PKGVER/PF_PKGREL), source=() when
PF_SRCDIR is set (a canary version has no tag to clone), stale NVENC-only
header fixed, and options=('!lto' '!debug') — makepkg's lto option injects
-flto=auto into CFLAGS, aws-lc-sys compiles its C with it, and rust's lld
cannot read GCC LTO bitcode: 'undefined symbol: aws_lc_*' at link (reproduced
minimally on Arch + rust 1.90). Full build + clean-container install
smoke-tested locally (binaries run, payload + scriptlets intact).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Layering is a last resort per the Bazzite docs (slows every OS update, can
block upgrades until removed); a sysext never enters an rpm-ostree
transaction, survives OS updates, and installs/updates with no reboot —
the mechanism Fedora Atomic ships via fedora-sysexts.
- build-sysext.sh wraps the built host+web RPMs into punktfunk-<V-R>-x86-64.raw:
/etc payload relocated to /usr/share/punktfunk/etc (a sysext carries only
/usr), the punktfunk-sysext helper embedded, ID=fedora + VERSION_ID pinned
(merges on Bazzite via ID_LIKE; REFUSED after a major rebase instead of
running soname-broken binaries — both behaviors validated live on Bazzite 43).
SELinux labels are baked in as squashfs pseudo-xattrs from matchpathcon:
unlabeled files run fine for user units but system daemons are DENIED
(udev couldn't read the gamepad rule under enforcing) — validated on-glass.
Refuses duplicate input package names (a stale noarch punktfunk-web next to
the x86_64 one built a chimera image with the dead node launcher once).
- punktfunk-sysext.sh: install/update/status/remove against per-Fedora-major
feeds (…/generic/punktfunk-sysext/f43[-canary]), SHA-256-verified, applies
the udev/sysctl scriptlet work + /etc copies, prints the layering-migration
hint. Live-validated on the .41 Bazzite box incl. service restart + web console.
- publish-sysext-feed.sh + rpm.yml: build + publish the image per matrix leg
(fedver 43/44), canary feeds pruned to 6, stable release assets attached.
- update-punktfunk.sh warns when the sysext shadows a layered install.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The notarized Developer ID .dmg was SIGKILLed at launch ("Launchd job spawn
failed", POSIX errno 163) before main() ran: the sandboxed macOS app declares
the MANAGED keychain-access-groups entitlement, which AMFI only honors when an
embedded provisioning profile authorizes it. The DMG embedded none — App Sandbox
and the network/device keys are self-asserted for Developer ID, but a keychain
access group is not — so every launch was killed at spawn. Validly signed and
notarized (Gatekeeper accepted it), which is why this looked like a mystery. ⌘R
and the App Store build hid it: Xcode embeds a development / App Store profile;
the raw-codesign DMG path did not, so "⌘R == DMG" never held for this entitlement.
Embed a "Punktfunk macOS Developer ID" profile (Keychain Sharing) into
Contents/embedded.provisionprofile before codesign so its entitlements authorize
the access group, exactly like the App Store build's profile does. If the profile
isn't installed on the runner, warn and strip keychain-access-groups instead so
the app still launches via ClientIdentityStore's legacy file-keychain fallback —
a missing/expired profile can never reship the errno-163 brick again.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The surfaced install command
flatpak install --user https://flatpak.unom.io/io.unom.Punktfunk.flatpakref
failed with "No such ref 'app/io.unom.Punktfunk/x86_64/stable'". The stable
commit's objects are on the server, but the repo *summary* (what flatpak reads
to resolve refs) listed only canary.
Root cause: each CI run builds a fresh SINGLE-branch local OSTree repo,
build-update-repo regenerates the summary from that one branch, and rsync
uploads it without --delete. Objects for both channels accumulate, but the
summary is overwritten every run and only names that run's branch. Canary runs
on every main push, stable only on tags — so a tag published stable, then the
next canary push clobbered the summary back to canary-only.
Fix: seed the local repo from the live server (rsync repo/ DOWN) before the
build, so it carries every published branch; the build only adds this run's
commit and the regenerated+signed summary keeps both channels. Single shared
repo kept (no URL/Caddyfile change; existing installs fixed transparently).
Adds a refs log after build-update-repo as a clobber tripwire. Also adopts
scripts/ci/pf-version.sh for the canary base (see previous commit).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Every release workflow hardcoded a canary base version (0.5.0 in
Apple/Android/rpm/flatpak/deb, 0.3 in windows-msix/windows-host/decky) that
had to be hand-bumped on each stable release and wasn't. With stable at
v0.6.0, every canary was a version *behind* stable — e.g. the Apple canary
showed up on TestFlight as 0.5.0 while 0.6.0 was already published.
Add scripts/ci/pf-version.{sh,ps1} (bash + pwsh twin) as the single source of
truth: stable = the vX.Y.Z tag; canary = latest stable tag with minor+1,
patch 0 (v0.6.0 -> 0.7.0), so canary is always exactly one minor ahead of the
newest release with zero maintenance. Falls back to the workspace Cargo.toml
version when no tag is fetchable. All workflows now eval/call it and format
their own channel suffix off $PF_BASE; only the canary branch changed, stable
branches and per-channel suffixes are untouched. channels.md drops the old
manual "bump the canary base" release step.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
New crates/punktfunk-tray — a small per-user companion showing the host service
state at a glance (running / stopped / starting / degraded / failed + the live
session in the tooltip) with one-click actions: open web console, approve a
pending pairing request, start/stop/restart, open logs. No more digging through
logs to learn whether the service came back after a reboot or an update.
Status is service-manager-FIRST (SCM / systemd user unit — a port squatter can
never fake Running), then the new loopback-only unauthenticated
GET /api/v1/local/summary (counts/booleans only; the mgmt token and cert.pem
are SYSTEM/Admins-DACL'd on Windows, so a non-elevated tray cannot bearer-auth).
Windows: windows_subsystem binary (a console exe in the Run key would flash a
terminal at sign-in), Shell_NotifyIcon + hidden window, per-session single
instance, TaskbarCreated re-add, --quit for the uninstaller; service actions
elevate per click via ShellExecuteW "runas" onto the new
`punktfunk-host service restart` (stop → wait Stopped → start).
Linux: ksni/StatusNotifierItem over zbus, systemctl --user actions (no polkit),
/etc/xdg/autostart entry whose --autostart self-gates to actual host users.
Icons: scripts/gen-tray-icons.py (pure stdlib) renders the brand lens + status
dot into committed .ico/hicolor assets; deb/rpm/arch ship binary+autostart+icons.
Live-validated: Linux on the headless KDE session (SNI registration, state
transitions, menu-driven start, dbusmenu layout); Windows on the RTX box
(session-1 launch with no NIM_ADD failure, single instance, --quit, restart
round-trip, summary loopback-200/LAN-401).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Frame ring (pf-vdisplay) and both gamepad SHM channels move off named Global\
objects (openable by any sibling LocalService) to UNNAMED sections/events whose
handles the host DuplicateHandles into the driver's verified WUDFHost with least
access — frame delivery over the SYSTEM+admins-only IOCTL_SET_FRAME_CHANNEL,
pads over a 32-byte named bootstrap mailbox (pid + handle value only, DoS-bounded;
HID minidrivers have no control device). Driver-validated pad_index kills
cross-pad redirects; v1↔v2 mixes fail closed with diagnosis logs on both sides.
Sibling-LocalService denial proven empirically (design/idd-push-security.md,
design/gamepad-channel-sealing.md).
Driver-side raw ops now live behind pf-umdf-util (checked shm accessors, the
forbid(unsafe_code) ChannelClient state machine, WDF request tokens) — the pad
drivers' logic is 100% safe Rust; whole drivers workspace clippy-gated in CI.
driver install --gamepad now sweeps SWD\punktfunk phantom devnodes: a re-created
SwDevice REVIVES the old devnode with its previously-bound driver (never
re-ranks), so an upgrade otherwise leaves the old driver serving — or, across
the v1→v2 fence, a dead pad (found live on the RTX box).
On-glass validated on the RTX 4090 box: frame path 7007 frames p50 2.06 ms
cross-machine; DualSense + XUSB "sealed pad channel mapped"/proto=2 attach via
both the test harness and a real streaming session; phantom-sweep repro.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A moved release tag re-fires the publish workflows, and the Gitea
registries reject duplicate uploads with 409 (deb pool, rpm group, and
the generic packages' versioned URLs; the channel aliases already
pre-deleted). Delete any prior copy of the exact version before
uploading (404 on first publish tolerated), so a republished tag
overwrites instead of wedging — v0.5.0's retag left stale no-port-change
artifacts published and every re-run red.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Modern branded wizard: WizardStyle=modern dynamic windows11 (Inno >= 6.6,
plain-modern fallback for older compilers; CI provisioning upgrades a
pre-6.6 Inno). Brand-mark wizard side panels + header tiles (100-200% DPI)
and a multi-size punktfunk.ico (SetupIconFile + Apps & Features), generated
AND committed by branding/gen-branding.ps1 from the canonical brand geometry.
Gotcha encoded in the script: ISCC rejects all-PNG icons, so entries <= 64px
are classic DIBs (PNG only at 128/256), and the ICO is load-verified.
- VB-CABLE actually ships now: windows-host.yml never set VBCABLE_DIR, so every
published installer silently omitted the virtual mic (broken mic passthrough
in the field). CI provisions the pinned, SHA-256-verified official Pack45
(provision-windows-punktfunk-extras.ps1) and the pack now FAILS on a
supplied-but-invalid dir instead of shipping mic-less again. Attribution per
VB-Audio's bundling grant surfaced in the visible wizard task text (vendor,
vb-cable.com, donationware) on top of the licenses notice.
- GameStream (Moonlight) compat is a wizard task (checked by default) ->
service install --gamestream=on|off writes PUNKTFUNK_HOST_CMD=
serve[ --gamestream] into host.env. Only the two canonical values are ever
rewritten - a hand-customized command line survives upgrades. Silent
installs: /MERGETASKS="!gamestream".
- Driver uninstall (field report: our virtual-device drivers survived
uninstall): new `driver uninstall [--gamepad]` removes the pf-vdisplay
device node(s) + the pf-vdisplay/pf-dualsense/pf-xusb driver-store packages,
wired into [UninstallRun] after service uninstall. Locale-safe by
construction: devices matched on unlocalized VALUES (never pnputil's
localized labels), packages found by INF content scan - validated against a
German-locale box ("Instanz-ID:" parse; 7/7 punktfunk INFs matched, no
foreign hits). VB-CABLE is deliberately left installed (shared third-party
component with its own uninstaller).
Installer compile, cargo check/clippy/fmt, and the ASCII locale gate are green;
the wizard look + uninstall flow still need one on-glass pass on a disposable
box (this box runs the live host).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
FFMPEG_DIR alone satisfies the linker, but the test binary needs the actual DLLs
on PATH at runtime. The daemon's own env (project-env.ps1, written by the
punktfunk-extras provisioning step) only takes effect on daemon *restart*, so a
freshly cloned/registered runner's first-ever job runs before that file has ever
been written, let alone picked up - confirmed live as STATUS_DLL_NOT_FOUND on the
new home-windows-runner-1's first real CI run. Setting PATH via GITHUB_PATH here
makes the workflow self-sufficient regardless of daemon restart timing.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The Windows CI runner (home-windows-runner-1, vmid 210) is now provisioned/owned by
unom/infra and can be rebuilt or joined by additional windows-amd64-labeled runners at
any time - a manually-dispatched provisioning workflow has no way to target a specific
runner instance, so it could land on an already-provisioned box instead of the one that
needed it. Replace windows-drivers-provision.yml / windows-punktfunk-provision.yml with
scripts/ci/ensure-windows-toolchain.ps1, a shared idempotent pre-flight (WDK/cargo-wdk,
FFmpeg, Inno Setup, ARM64 rustup target) that every Windows workflow now runs at job
start - a fast no-op once already provisioned, so any runner self-heals on first real use.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Both Windows publish steps threw on any non-zero curl exit, so re-running a vX.Y.Z
tag (e.g. after a force-push) failed at the versioned generic-registry path —
that path is immutable and 409s a re-upload of an already-published version. The
channel alias right below already delete-then-reuploads to dodge this; mirror that
intent for the versioned path by reading the HTTP status and treating 409 as a
no-op. The MSIX/installer still build, sign, and attach to the release fine — this
only unbreaks the redundant re-publish on a tag re-run.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The [workspace.package] version (inherited by every crate via version.workspace)
lagged at 0.3.0 through the 0.4.0 release — bump it to 0.4.1, the release being
cut, and refresh the 8 workspace entries in Cargo.lock to match (CI builds
--locked).
Also advance the CI canary-base fallbacks (deb/rpm/flatpak/android/release
workflows + build-rpm.sh) from 0.3.0 to 0.5.0 so main/canary builds sort one
minor ahead of the latest stable line, per the documented channel convention.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Web console
- Pairing/Library/Stats refactored into self-contained subsections that each own
their own queries + mutations; a shared slot-based layout (view.tsx) is filled by
the live page (containers) and Storybook (pure cards + fixtures) so the layout can't
drift.
- All paired devices in one list on Pairing with a protocol column (punktfunk/1 +
Moonlight), routing each unpair to the right endpoint; the redundant Clients page is
removed.
- Library: overview grid split from the add/edit form into separate files.
- Login screen links out to the docs.
Docs
- "Console login password" section on every host page (apt/RPM/Bazzite/SteamOS/Windows)
plus a new "Forgot your Password?" troubleshooting page, linked from the login screen.
- Console served as HTTP/1.1 over TLS (drop the unusable HTTP/3 advertising) across the
Bun entry, launchers, systemd units, and packaging.
Tooling
- Biome now respects .gitignore (stops linting generated code), config migrated to
2.5.1; all lint issues fixed cleanly.
Also includes this branch's in-progress host, Apple client, packaging, and CI changes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The in-app OSS license screens (a05c08b) added a `resources:` array to the
PunktfunkKit SwiftPM target, which makes SwiftPM emit a resource-bundle target
(PunktfunkKit_PunktfunkKit). A resource bundle is a product type that cannot
carry a provisioning profile, so the explicit PROVISIONING_PROFILE_SPECIFIER
each release.yml archive step set — global on macOS, sdk-scoped on iOS/tvOS —
now lands on it and fails the archive ("does not support provisioning profiles")
on all three platforms. (Before that commit there was no resource bundle, so the
profile was harmless.)
Switch all three archive steps to CODE_SIGN_STYLE=Automatic (development):
Automatic signing assigns a profile only to the app target and leaves the
resource bundle (and the macOS-host SwiftPM macro plugins) alone, and bakes the
sandbox entitlements in. No -allowProvisioningUpdates, so it stays offline and
never cloud-signs (the App-Manager ASC key can't). DISTRIBUTION signing is
unchanged — still manual, in the -exportArchive step (which maps the profile to
io.unom.punktfunk only). Drops the now-unneeded manual signing xcconfigs.
Requires the runner to have a development provisioning profile for
io.unom.punktfunk on each platform (now installed for macOS/iOS/tvOS).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The MIT OR Apache-2.0 SOURCE license is clean (audit found no copied copyleft); the
gaps were all binary-distribution (Layer-2). This makes the shipped artifacts honest:
- Windows host + client: bundled FFmpeg BtbN gpl-shared -> lgpl-shared (AMF/QSV/decode
unaffected; the GPL-only x264/x265 were never used), and ship the FFmpeg LGPL notice
+ license text in the installer + MSIX (licenses/).
- THIRD-PARTY-NOTICES.txt generated + bundled into installer/MSIX/deb/rpm. Offline
generator (scripts/gen-third-party-notices.{py,sh}) + cargo-about config (about.toml/
.hbs) with a permissive-only accepted-license allow-list as a copyleft regression gate.
- Reword the win32u GPU-preference hook comments to reflect independent reimplementation
(no Apollo/Sunshine GPL-3.0 source copied).
- README dual-license + inbound=outbound contributor clause + non-affiliation trademark
disclaimer; new CONTRIBUTING.md.
- LICENSE files into the standalone driver + vk-layer workspaces; deb copyright holder
aligned to "unom and the punktfunk contributors".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
With cmake now found, Homebrew's CMake 4 refuses the vendored libopus's
`cmake_minimum_required(VERSION <3.5)` ("Compatibility with CMake < 3.5 has been
removed"). Export CMAKE_POLICY_VERSION_MINIMUM=3.5 (the same knob the Windows
build uses) so the cmake crate's child cmake configures the audiopus_sys libopus.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The self-hosted macOS runner runs steps with `bash --noprofile --norc`, so
Homebrew's bin dir is not on PATH — the previous `brew install cmake` died with
`brew: command not found` (exit 127). Find brew at its known prefix, install cmake
only if missing, and export the brew bin dir to $GITHUB_PATH so the subsequent
xcframework build (audiopus_sys → vendored libopus) actually finds `cmake`.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The 5.1/7.1 surround commit (7612238) added in-core Opus, which broke two CI jobs
that the merge didn't touch:
* Windows MSIX client: clients/windows/src/main.rs's headless `SessionParams`
initializer was missing the new `audio_channels` field (the GUI path sets it
from settings). Default the CLI/test path to stereo (2), matching trust.rs.
* Apple xcframework (apple.yml + release.yml): in-core Opus decode pulls
`audiopus_sys`, which builds a vendored *static* libopus via CMake when
pkg-config finds no system Opus — keeping the xcframework self-contained (no
runtime libopus.dylib on end-user Macs/devices). The self-hosted macOS runner
lacked `cmake`; install it self-healing before every xcframework build.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Play-listing/marketing screenshots of the Compose client rendered on the host JVM
by Roborazzi (Robolectric Native Graphics) — no emulator, GPU, KVM, host, or JNI
core. Five scenes render the REAL composables with embedded mock state under a
forced brand palette (Material You has no wallpaper to seed from on the JVM):
hosts grid, settings, TOFU + PIN dialogs, and the live stats HUD. Validated 5/5
locally.
- New JVM unit-test source set (app/src/test) + Roborazzi/Robolectric test deps;
@Config(sdk=36) is mandatory (no android-all jar for compileSdk 37) and the
animation clock is paused so a text-bearing scene reaches idle.
- kit: `-PskipRustBuild` skips the cargo-ndk native build so the JVM-only test job
needs no Rust/NDK; normal APK/AAR builds are unchanged.
- Widen BrandDark / StatsOverlay to internal so the tests can use them.
- Standalone best-effort tag-gated workflow; PNGs upload as a 30-day artifact.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Host-free UI screenshots of the GTK4/libadwaita client under a virtual X display
(clients/linux/tools/screenshots.sh) — Xvfb + software GL (llvmpipe) + a root-window
grab, one app launch per scene. PUNKTFUNK_SHOT_SCENE routes build_ui to render one
mock-populated REAL view (hosts grid / settings dialog / TOFU + PIN dialogs) and
print PF_SHOT_READY once it has settled; the saved-hosts grid is driven by a seeded
client-known-hosts.json. NON_UNIQUE in shot mode so back-to-back launches don't
collide. The stream scene is deferred — its page needs a live NativeClient.
Gated to stable release tags in a standalone best-effort workflow that builds the
client in the rust-ci image and captures under Xvfb; PNGs upload as a 30-day
artifact, not committed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Marketing/store screenshots of the console, captured from the built Storybook
with headless Chromium (web/tools/screenshots.mjs) — every Pages/* + Shell/*
story rendered at 1440x900@2x. The page stories render from fixtures, so no live
mgmt API, login, or GPU is needed (the web analogue of apple.yml's screenshots
job). Gated to stable release tags in a standalone best-effort workflow; PNGs
upload as a 30-day artifact, not committed.
- Add Stats + Pairing stories (the two pages that lacked them) with stats/pairing
fixtures typed against the generated models.
- Extract a pure PairingView (index.tsx -> view.tsx), matching the
Dashboard/Clients/Stats split, so the page renders host-free from mock state
instead of racing its polling queries. Container wiring is behaviour-identical.
- Playwright driver + a chromium-capable tag-gated job.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Plugin self-update (no Decky store): CI publishes a per-channel manifest.json
({version, immutable per-version artifact, sha256}) beside the zip and bakes
update.json {channel, manifest} into the plugin. main.py `check_update` reads the
installed version from package.json (the value Decky reports — not plugin.json),
fetches the channel manifest, and the frontend shows an "Update to vX" button that
drives Decky Loader's own install RPC (root downloads + SHA-256-verifies + hot-reloads).
CI now stamps a plain-numeric semver (0.3.<run> canary / X.Y.Z stable) into
package.json — a -ciN suffix would mis-order under compare-versions.
Linux client: `--fullscreen` (plus SteamDeck/gamescope env fallback) enters GTK
fullscreen on stream start so Gaming-Mode chrome is hidden; native-mode resolution
falls back to the display's first monitor when the window isn't mapped yet (was
dropping to the 1080p floor — wrong on the Deck's 1280×800); add a confirmed
"Remove saved host" action (KnownHosts::remove_by_fp).
Docs: new docs/steam-deck.md (Decky install/pair/stream/self-update/troubleshooting),
wired into meta.json nav, and cross-linked from clients/install-client/channels. This
is the page docs.punktfunk.unom.io/docs/steam-deck — the website's download link
pointed at it before it existed; committing it makes that link resolve.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Now that the drivers build from source in CI, remove the dead checked-in binaries and
the toolchain cruft they left behind:
- Delete packaging/windows/{pf-vdisplay,gamepad-drivers}/ (the prebuilt .dll/.inf/.cat/.cer).
pack-host-installer.ps1 builds + signs all three drivers from the drivers/ workspace and
nothing reads the vendored dirs anymore; stage-pf-vdisplay.ps1's -VendorDir is now a
mandatory build-output path, not a vendored default.
- Drop the LLVM-21 pin. The vendored bindgen 0.71->0.72 bump (the shipping pack already
builds green on the runner-default clang 22) retired the bindgen-0.71 layout-test overflow
that needed LLVM 21.1.2, so windows-drivers.yml + provision-windows-wdk.ps1 no longer
install/point at C:\llvm-21 (~898 MB off a fresh provision) - both driver builds now use one
toolchain (clang 22 + bindgen 0.72).
- pack -SkipBuild on the gamepad build (build-pf-vdisplay.ps1 already builds the whole
workspace), build-web.ps1 reaps a stale node too, deploy-dev.ps1 nefconc path + comments.
- Reword the vendored-driver references (build scripts, .iss, READMEs, the vite web-bundle
comment) to the build-from-source reality.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The pf-vdisplay virtual-display driver shipped as a checked-in PREBUILT binary
that went stale - two field failures on a fresh install (live-repro'd on a
German-locale Dell laptop):
* Bug A (every box): a repo-wide rename edited the vendored pf_vdisplay.inf
but never re-signed pf_vdisplay.cat, so the catalog stopped covering the INF
-> `pnputil /add-driver` fails SPAPI_E_FILE_HASH_NOT_IN_CATALOG -> driver
never installs -> every session dies "pf-vdisplay driver interface not
found".
* the prebuilt binary also predated IOCTL_SET_RENDER_ADAPTER (added to the
driver source after the vendor freeze) that the host needs to pin the IDD
render GPU on hybrid/Optimus boxes.
Fix: build the driver FROM SOURCE every release (build-pf-vdisplay.ps1, wired
into pack-host-installer.ps1) so .dll/.inf/.cat are always in lockstep and
current driver features ship. The runner's clang 22 made the driver's pinned
bindgen 0.71 emit opaque structs (157 layout-assert errors), so bump the
vendored wdk-sys/wdk-build bindgen 0.71 -> 0.72 (+ lock). The build self-signs
the driver per build (installer trusts the bundled .cer); a stable
DRIVER_CERT_PFX_B64 secret can override.
* Bug B (non-English boxes): the installer runs install-pf-vdisplay.ps1 etc.
via powershell.exe (5.1), which reads a BOM-less script in the ANSI codepage
- an em-dash's trailing 0x94 byte becomes a curly quote on German
Windows-1252 and the script aborts "unterminated string", so the driver
never installed (the gamepad script survived only because it was already
ASCII). Scrub every installer-run .ps1/.cmd to ASCII + add a CI gate that
fails on any non-ASCII so it can't regress.
* Bug C (upgrades): nothing stopped the OLD web console before re-registering
its task, so a stale server kept :3000 (the new one restart-looped on
EADDRINUSE) and served a broken old bundle (500 on /login). Stop + reap it
(runtime-agnostic, by the :3000 listener owner) in web-setup.ps1 and in the
.iss before the file copy + on uninstall.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The file moves (docs/ → design/, docs/api/openapi.json → api/openapi.json) landed
in 8897286, but the matching reference updates did not — so mgmt.rs's drift-test
`include_str!("../../../docs/api/openapi.json")` pointed at a path that no longer
exists and the host failed to build. This restores it and updates every reference:
- mgmt.rs include_str! → ../../../api/openapi.json (fixes the build)
- web/orval.config.ts codegen target, web/Dockerfile, .dockerignore
- deb/rpm/Arch packaging install paths
- CLAUDE.md, the .gitea CI workflows, code doc-comments, design-doc cross-links
docs-site route URLs (/docs/...) untouched.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
NVIDIA/AMD Vulkan ICDs refuse to *advertise* an HDR color space for a surface on an
IddCx indirect/virtual display, so Vulkan games (Doom: The Dark Ages, id Tech, Indiana
Jones, …) report "device does not support HDR" — even though Windows HDR, DWM compose,
and the client PQ stream all work, and the ICD happily *accepts + presents* a forced HDR
swapchain there. The whole gap is enumeration; the community (Apollo/Sunshine/VDD) wrote
this off as kernel-side / unfixable.
Add VK_LAYER_PUNKTFUNK_hdr_inject (packaging/windows/pf-vkhdr-layer/): a standalone
cdylib Vulkan implicit layer that appends {A2B10G10R10, HDR10_ST2084} + {RGBA16F, scRGB}
to vkGetPhysicalDeviceSurfaceFormats[2]KHR (no need to hook vkCreateSwapchainKHR — the
ICD doesn't validate the color space there). Self-gated on the surface monitor's actual
advanced-color state (DisplayConfig GET_ADVANCED_COLOR_INFO), so it is a complete no-op
on SDR sessions and real monitors (dedup). Always-on (registry-discovered) so it works
regardless of how a game is launched — env-scoping silently fails for already-running
Steam. Escape hatches: DISABLE_PF_VKHDR, PF_VKHDR_EXCLUDE, and a built-in kernel-anti-
cheat denylist.
The installer builds/signs/stages it and registers it under
HKLM64\SOFTWARE\Khronos\Vulkan\ImplicitLayers (opt-out "Install the HDR Vulkan layer"
task); windows-host CI fmt+clippy-gates it (msvc-only FFI).
Live-validated on the RTX box: Doom: The Dark Ages enables HDR over the pf-vdisplay
virtual display.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The shared host<->driver ABI crate already contains more than the virtual
display: the IDD-push frame ring + control plane AND the gamepad shared-memory
layouts (XusbShm / PadShm). "pf-vdisplay-proto" was a misnomer — the name now
represents all the drivers it serves.
Mechanical rename, no behavior change:
- git mv crates/pf-vdisplay-proto -> crates/pf-driver-proto (package name +
path-deps in the host crate and the driver workspace).
- pf_vdisplay_proto -> pf_driver_proto across host + driver Rust, both Cargo.lock
files, the workspace members, the CI path triggers (windows-drivers.yml), and
the docs/INF comments. The runtime Global\pfvd-* shared-object names are a
SEPARATE contract and are deliberately untouched (host<->driver name matching).
- The pf-vdisplay DRIVER crate + its INF service name (Root\pf_vdisplay,
UmdfService=pf_vdisplay, pf_vdisplay.dll) are unchanged — only the full
`pf_vdisplay_proto` token was replaced, never the `pf_vdisplay` driver name.
Linux-verified: cargo test -p pf-driver-proto (const size-asserts compile) +
cargo clippy -p punktfunk-host -D warnings clean; Cargo.lock regenerated. The
driver-workspace side (path-dep + imports + its Cargo.lock) is Windows-CI-gated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The versioned IDD_STRUCTURE_SIZE path referenced IddClientVersionHigherThanFramework/
IddStructureCount/IddStructures — LNK2019 unresolved, because the WDK links the iddcx
1.0 IddCxStub which lacks those (they are >=1.4). We target 1.10 against a current
framework (higher==false) where size_of is exactly the versioned result, so use it
directly (the surface-assert refs linked only because they were DCE-eliminated).
pf-vdisplay now COMPILES + LINKS IddCxStub on the box (263,680B). Point
windows-drivers.yml at the whole workspace + clear FORCE_INTEGRITY on pf_vdisplay.dll;
drop the obsolete UINT diagnostic dump.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
UINT fails to resolve despite a top-level `pub type UINT` in the same scope as the
working `use crate::types::*` — error count byte-identical before/after the fix.
Add an if:always() step dumping the generated module structure + UINT-use context
to pinpoint the scope mismatch (RTX box rebooted to Proxmox, so CI is the only
validator).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
wdk-build links UMDF drivers with /INTEGRITYCHECK unconditionally (no opt-out),
so the self-signed DLL would be refused by Code Integrity (3004/3089). Add a
deterministic, idempotent, reusable packaging step
(packaging/windows/clear-force-integrity.ps1) that clears the PE
IMAGE_DLLCHARACTERISTICS_FORCE_INTEGRITY bit (0x0080 @ e_lfanew+0x5e) and verifies
— the gamepad recipe, no longer hand-run. driver-build now inspects the bit
(before) then clears+verifies it. Real drivers will: build -> clear -> sign .dll
-> Inf2Cat -> sign .cat.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
enricobuehler