The installer's vendored driver binary (packaging/windows/pf-vdisplay/) was STALE — built from the OLD
oracle tree (packaging/windows/vdisplay-driver/, wdf-umdf, SudoVDA-compat GUID), so it was
ABI-mismatched with the host (which opens the owned proto GUID 70667664). Re-vendor it from the NEW
drivers/ tree so the rewrite's ACTUAL driver is what the installer ships.
Built RELEASE on the RTX box from the new tree + the new .inx: cargo build --release -p pf-vdisplay ->
FORCE_INTEGRITY clear -> stampinf (DriverVer 06/25/2026,9.5.0625.1614, > the old 06/22) -> Inf2Cat
/os:10_X64 -> signtool sign the .cat with punktfunk-ds-test (.cat sig Valid). Replaces the stale
.dll/.inf/.cat; the .cer is unchanged (same cert).
ON-GLASS VALIDATED (install-test): pnputil /add-driver /install the release package -> clean WUDFHost
reload -> Status=OK, init_adapter -> IddCxAdapterInitAsync -> 0x0 (FP16 accepted),
IddCxMonitorCreate(id=1) -> 0x0. The shipping installer now installs + loads the real wdk-sys
proto-GUID driver, FP16/HDR-capable, monitor-create working.
Remaining STEP 8 (recorded in memory, deferred): re-point the stale "built from vdisplay-driver/"
comments in stage-pf-vdisplay.ps1 / pack-host-installer.ps1 / packaging README; selector default ->
pf-vdisplay unconditional; CI build-sign-or-stale-vendored drift guard; then DELETE the oracle tree.
KEEP sudovda.rs (runtime fallback + the backend-neutral CCD helpers pf_vdisplay.rs reuses) and the
WGC-relay/DDA secure path (the secure-desktop lock/UAC gate is not yet proven on glass for IDD-push).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The new wdk-sys driver tree (packaging/windows/drivers/pf-vdisplay/) had no INF — it borrowed the
oracle's (packaging/windows/vdisplay-driver/.../pf_vdisplay.inx), which blocked deleting the oracle.
Port it verbatim: the proto-vs-SudoVDA control GUID is registered in CODE
(WdfDeviceCreateDeviceInterface), so the INF is GUID-agnostic and identical — HWID Root\pf_vdisplay,
UmdfExtensions=IddCx0102, the control-device security DACL, UpperFilters=IndirectKmd,
UmdfHostProcessSharing=ProcessSharingDisabled. Prerequisite for the STEP-8 re-vendor (build ->
stampinf -> Inf2Cat -> sign the .dll/.cat from the NEW tree into packaging/windows/pf-vdisplay/,
replacing the stale oracle-built binary) and for deleting the oracle tree.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Audit pass over the new pf-vdisplay driver's unsafe surface: 92 per-site // SAFETY comments added
across adapter.rs / monitor.rs / entry.rs / callbacks.rs / swap_chain_processor.rs /
frame_transport.rs / direct_3d_device.rs (control.rs already had full coverage). COMMENTS ONLY — zero
logic, signature, or control-flow change (verified via git diff: every added line is a // SAFETY
comment or blank).
The dominant gap was the pervasive `core::mem::zeroed()` FFI-struct builds (IDDCX_*/WDF_*/
DISPLAYCONFIG_* C PODs whose all-zero bit pattern is a valid uninitialized/Invalid state, with the
required .Size/fields set immediately after) — each now carries a one-line // SAFETY. Plus explicit
notes on the two stack/local-pointer-into-FFI hazards (adapter.rs `version` ptr into
IddCxAdapterInitAsync; monitor.rs `edid` Vec ptr into IddCxMonitorCreate — both read synchronously
before the local drops) and the frame_transport.rs raw-HANDLE / mapped-header derefs + cleanup paths.
The already-justified Send/Sync wrappers (SendAdapter, CtxTypeInfo/DevCtxInfo, MonitorObject,
Sendable, FramePublisher) were audited — each already carried a // SAFETY. No site needed a code
change.
First slice of STEP 8 (the SudoVDA drop). Comments-only ⇒ build-neutral; windows-drivers.yml verifies
on the next runner build. Remaining STEP 8: re-vendor the installer's driver binary from the new
drivers/ tree (the shipping packaging/windows/pf-vdisplay/ binary is still built from the OLD oracle
tree with the SudoVDA-compat GUID — ABI-mismatched with the host's proto GUID), add an .inx to the
new tree, re-point scripts/README from vdisplay-driver/ to drivers/, flip the selector default to
pf-vdisplay, then delete the old oracle tree. Keep sudovda.rs (the runtime fallback + the
backend-neutral CCD helpers pf_vdisplay.rs reuses) and the WGC-relay/DDA secure path (the
secure-desktop gate is not yet passed on glass).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The pf-vdisplay driver now advertises HDR/FP16 and the full glass-to-glass HDR path works
end-to-end — validated LIVE: the Mac client connected to the .173 host WITH HDR (display_hdr=true,
FP16 ring -> NVENC P010). The STEP-3 assumption that FP16 needs a higher UmdfExtensions was WRONG:
IddCx0102 + CAN_PROCESS_FP16 + the *2 DDIs works (the oracle proved it; confirmed on-glass
IddCxAdapterInitAsync -> 0x0 WITH the FP16 cap set). Driver-only change — the host FP16-ring ->
NVENC-P010 path and the HDR EDID were already in place.
- adapter.rs: caps.Flags = IDDCX_ADAPTER_FLAGS_CAN_PROCESS_FP16.
- entry.rs: register the 6 *2/HDR callbacks (ParseMonitorDescription2, MonitorQueryTargetModes2,
AdapterCommitModes2, AdapterQueryTargetInfo, MonitorSetDefaultHdrMetaData, MonitorSetGammaRamp)
ALONGSIDE the v1 set (matching the oracle — CAN_PROCESS_FP16 OBLIGATES the *2 DDIs or the
framework rejects the adapter at init; STEP 3 rejected FP16 only because they weren't registered).
- callbacks.rs: parse_monitor_description2 + monitor_query_modes2 now fill IDDCX_MONITOR_MODE2 /
IDDCX_TARGET_MODE2 with BitsPerComponent (8|10 bpc RGB); query_target_info already reports
IDDCX_TARGET_CAPS_HIGH_COLOR_SPACE; set_default_hdr_metadata + set_gamma_ramp accept (the gamma
one is mandatory under FP16).
- monitor.rs: wire_bits() (Rgb 8|10, no YCbCr) + target_mode2().
- EDID + INF UNCHANGED (the EDID already carries the CTA-861.3 BT.2020 + ST.2084/PQ block; the INF
stays UmdfExtensions=IddCx0102).
Built via the ultracode flow (STEP-7 map workflow -> agent-implement -> box build [driver green] ->
deploy -> on-glass HDR). OPERATIONAL NOTE: do NOT Disable/Enable the IddCx devnode to reload it —
that leaves the adapter STOPPED in the persisted WUDFHost process (ADAPTER OnceLock survives), so
monitor-create then fails with 0xc00002b6 (INDIRECT_DISPLAY_DEVICE_STOPPED). Kill the pf_vdisplay
WUDFHost process (or reboot) for a clean adapter re-init.
This completes the pf-vdisplay rewrite STEP 0-7, all on-glass validated (loads, adapter inits,
monitor appears, swap-chain drain, IDD-push frames at ~235fps, and HDR). Remaining: STEP 8 (unsafe-
reduction + delete the old vdisplay-driver tree + the vendored SudoVDA driver + unbundle from the
installer = the SudoVDA drop).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The driver now publishes each acquired swap-chain surface into the host-created shared ring (the
IDD-push path) — the full glass-to-glass transport is code-complete. Both sides use the canonical
pf_vdisplay_proto::frame layout (lockstep by compile-error, not "must match" comments). Driver compiles
+ LOADS on-glass (adapter inits, Status=OK; no regression — the publisher is dormant until a frame is
acquired); host cargo check green; adversarially reviewed (no blockers — token layout, keyed-mutex key 0,
names by target_id, and the format guard all match the host consumer).
- new driver frame_transport.rs: FramePublisher OPENS the host ring by target_id (OpenFileMapping header
+ magic Acquire readiness gate + OpenEvent + OpenSharedResourceByName RING_LEN keyed-mutex textures),
writes its render LUID + DRV_STATUS back into the header; publish() is NON-BLOCKING (round-robin 0ms
try-acquire -> CopyResource -> ReleaseSync -> FrameToken::pack store Release -> SetEvent; drops the
frame if every slot is busy or the surface format != the ring format). Manual handle/view cleanup on
every try_open early return; RAII Drop (slots -> unmap -> CloseHandle). Layout/consts/names/token all
from pf_vdisplay_proto::frame.
- swap_chain_processor.rs run_core: lazy rate-limited attach (every ~30 frames) + is_stale re-attach
(mid-session HDR ring recreate); publishes buffer.MetaData.pSurface via IDXGIResource::from_raw_borrowed
(preserves IddCx's refcount) BEFORE IddCxSwapChainFinishedProcessingFrame. run/run_core gain the render
LUID; callbacks.rs assign_swap_chain passes it.
- host idd_push.rs migrated onto pf_vdisplay_proto::frame (deleted the hand-rolled SharedHeader / MAGIC /
VERSION / RING_LEN / DRV_STATUS_* / name fns / token packing) — pure refactor, byte-identical, no
behavior or gating change. DebugBlock + DXGI_SHARED_RESOURCE_RW kept local (not in the proto).
- driver windows crate gains Win32_System_Memory (MapViewOfFile/OpenFileMappingW/...); rustfmt'd the whole
driver workspace (incl. wdk-probe — fmt-only).
Built via the ultracode flow: STEP-6 map workflow -> agent-implement -> box build (driver + host both
green; caught nothing this time) -> adversarial-verify-agent (no blockers) -> FrameToken::pack hardening
-> deploy (loads). Glass-to-glass frame validation awaits a composited session (per the parity finding:
this headless box yields 0 frames for the proven SudoVDA path too). FOLLOW-UPs: port the optional
Global\pfvd-dbg DebugBlock triage channel to the new driver; STEP 7 HDR; STEP 8 drop SudoVDA.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
STEP 5 (d8a453f) added the windows + thiserror deps to pf-vdisplay/Cargo.toml but the
workspace lock was not updated (driver is windows-only, cant build on the Linux dev box).
Regenerated on the RTX box. Both crates were already resolved in the lock (pulled by
wdk-build), so this is purely the pf-vdisplay dependency edges.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The pf-vdisplay driver now consumes the OS swap-chain so a virtual monitor is a usable
display rather than a stalled one. Compiles + loads on-glass (no regression: adapter still
inits, Status=OK); adversarially reviewed — no blockers, the leak/deadlock invariants preserved.
- new swap_chain_processor.rs: a worker thread (MMCSS "Distribution") that binds the render D3D
device (IddCxSwapChainSetDevice, single-borrow 60x@50ms retry) then drains the swap-chain
(ReleaseAndAcquireBuffer2 -> FinishedProcessingFrame; E_PENDING waits 16ms on the surface
event). NO frame publisher yet (STEP 6). RAII terminate+join Drop; the load-bearing
top-of-loop terminate check (the oracle's reconnect-leak fix). Fixed a Rust-2021 disjoint-
capture bug: `.0` field access bypassed the Sendable Send wrapper -> rebind the whole wrappers.
- new direct_3d_device.rs: CreateDXGIFactory2 -> EnumAdapterByLuid(render LUID) -> D3D11CreateDevice;
a DEVICE_POOL of one Arc<Direct3DDevice> per render LUID (the NVIDIA-UMD-worker-thread leak fix).
- monitor.rs: MonitorObject gains swap_chain_processor; set/take helpers return it for the caller
to drop OUTSIDE the MONITOR_MODES lock (dropping joins the worker — must never happen under the
lock); remove_monitor/clear_all drop it before IddCxMonitorDeparture.
- callbacks.rs: assign_swap_chain spawns the processor (pooled device per RenderAdapterLuid;
WdfObjectDelete on D3D-init failure so the OS retries); unassign_swap_chain drops it. Fixed the
stale `panic = "abort"` doc (workspace is unwind; the extern "C" boundary aborts on unwind).
- Cargo.toml: windows 0.58 + thiserror (both already resolved in the driver lock). The 3 needed
swap-chain DDIs were already wrapped in wdk-iddcx; their HRESULT-shaped NTSTATUS is classified
by hand (hr>=0 success, 0x8000000A E_PENDING).
- Also rustfmt'd the whole driver workspace (it had never been driver-fmt'd).
Built via the ultracode flow: STEP-5 map workflow -> agent-implement -> box build (caught the
Send-capture bug) -> adversarial-verify-agent -> deploy (loads). Session-1 on-glass validation
(the drain loop servicing an ACTIVE monitor) is the next gate — assign_swap_chain only fires
under an interactive session. Note for STEP 6: target_id_for_object uses the MONITOR_MODES handle
lookup the oracle moved to a WDF context; revisit before target_id keys the shared frame ring.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
94e82df shipped the agent-written pf_vdisplay.rs unformatted (cargo fmt --all --check
gate) and omitted the Cargo.lock edges for the new windows-only deps (pf-vdisplay-proto +
bytemuck). cargo fmt --all is now clean; Cargo.lock records the host dep edges.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The host can now drive the new pf-vdisplay IddCx driver instead of SudoVDA. Compiles
clean on BOTH Windows (cargo check -p punktfunk-host green) and Linux (cfg(windows)-gated,
main CI unaffected); adversarially reviewed (no blockers, lockstep with the driver).
- new vdisplay/pf_vdisplay.rs: cloned from the proven sudovda.rs, repointed to
pf_vdisplay_proto — interface GUID 70667664 (not e5bcc234), IOCTL 0x900-0x905 (not the
gappy 0x800/0x888/0x8FF), AddRequest/AddReply/RemoveRequest/SetRenderAdapterRequest
(bytemuck Pod, not the GUID-keyed AddParams), a u64 session_id monitor key (not a minted
GUID), and a single IOCTL_GET_INFO handshake that HARD-asserts protocol_version (vs
SudoVDA two-IOCTL best-effort). Full MGR/linger/refcount/teardown lifecycle preserved.
- reuses sudovda.rs backend-neutral CCD/DXGI helpers (set_active_mode, isolate/restore_
displays_ccd, resolve_gdi_name, resolve_render_adapter_luid, MON_GEN/CURRENT_MON_GEN,
SavedConfig) — widened to pub(crate), not duplicated.
- vdisplay::open()/probe() select the backend: PUNKTFUNK_VDISPLAY=pf|sudovda forces one;
default auto-detects (prefer pf-vdisplay if its interface enumerates, else SudoVDA stays
the shipping fallback).
Notes: SET_RENDER_ADAPTER is tolerated as the driver returns NOT_IMPLEMENTED today (STEP 4
tail); the cross-MGR wait_for_monitor_released only paces sudovda's MGR (benign until
IDD-push lands on pf-vdisplay, STEP 6 — documented in-code). On-glass "monitor appears at
WxH@Hz" gate is next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The virtual-monitor lifecycle is now code-complete on the driver side (CI-green;
deployed — no load/adapter-init regression, Status=OK):
- new monitor.rs: the monitor/mode model (Mode/MonitorObject/MONITOR_MODES), ported from
upstream virtual-display-rs with guid:u128 -> session_id:u64. create_monitor builds an
EDID (serial=id) -> IddCxMonitorCreate -> IddCxMonitorArrival, stores the monitor, and
returns the OS target id + adapter LUID for AddReply. remove_monitor / clear_all depart
+ drop. display_info/target_mode build the DISPLAYCONFIG timing (the union videoStandard
u32 set directly — bindgen-API-agnostic, vs the oracle new_bitfield_1 transmute).
- callbacks.rs: parse_monitor_description (EDID-serial lookup -> count-then-fill
IDDCX_MONITOR_MODE) + monitor_query_modes (pointer-match -> IDDCX_TARGET_MODE) are real.
- control.rs: IOCTL_ADD -> create_monitor + AddReply, REMOVE -> remove_monitor, CLEAR_ALL
-> clear_all, via read_input/write_output_complete WDF buffer helpers. SET_RENDER_ADAPTER
still stubbed (hybrid-GPU pin, next) + the watchdog thread (next).
- DISPLAYCONFIG_* resolve at the wdk_sys root (pub use types::*), not iddcx.
Warnings are the STEP-7 *2/HDR stubs + created_at (read by the watchdog, next). The
on-glass "monitor appears at WxH@Hz" gate awaits the host switch to pf_vdisplay_proto.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
EvtIddCxDeviceIoControl now dispatches the pf-vdisplay-proto control plane (new
src/control.rs): IOCTL_GET_INFO writes InfoReply{protocol_version, watchdog_timeout_s}
(the host asserts the version + fails loudly on mismatch), IOCTL_PING bumps the watchdog
keepalive. ADD/REMOVE/SET_RENDER_ADAPTER/CLEAR_ALL are dispatched but stubbed
(STATUS_NOT_IMPLEMENTED) pending create_monitor + the real mode DDIs (next). Unknown
IOCTLs -> STATUS_NOT_FOUND. Builds CI-green; warnings are the *2/HDR stubs (STEP 7) +
the stored adapter handle (read by create_monitor, next).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Verified on-glass after cleanup: adapter still inits (IddCxAdapterInitAsync 0x0,
Status OK) and WdfDeviceCreateDeviceInterface 0x0.
- RESTORE WdfDeviceCreateDeviceInterface (regression from debugging): the proto control
plane sends IOCTLs via EvtIddCxDeviceIoControl, which needs the device interface for the
host to open. Upstream omits it only because it uses a socket; ours is IOCTL-based.
- Drop the framework_struct_size / version-table machinery + size.rs: size_of suffices
(these are IddCx 1.10 structs on a 1.10 framework, matching upstream). The version-table
reads were added chasing a size mismatch that was never the bug (GammaSupport was).
- Drop /OPT:NOICF (ICF folding was a non-issue) + fix the stale stub-pick comment (the
1.10 stub is needed for the dispatch table, not size.rs symbols).
- Debug-wait/PID-file/go-file gate already removed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The all-Rust wdk-sys IddCx driver now initializes its adapter on the RTX box:
IddCxAdapterInitAsync -> 0x0, EvtIddCxAdapterInitFinished fires, device Status=OK.
ROOT CAUSE (found via cdb wt-trace of iddcx!IddCxImplAdapterInitAsync + the upstream
virtual-display-rs source): IDDCX_ENDPOINT_DIAGNOSTIC_INFO.GammaSupport was left zeroed
= IDDCX_FEATURE_IMPLEMENTATION_UNINITIALIZED (0), which the framework adapter validator
(ddivalidation.cpp:797) rejects with STATUS_INVALID_PARAMETER. Must be NONE (1).
Also required (matched to the proven-working upstream virtual-display-rs, installed +
verified Status=OK on the same box):
- caps Flags = NONE (SDR). CAN_PROCESS_FP16 needs a newer contract than UmdfExtensions=
IddCx0102 grants; deferred to STEP 7 (HDR).
- SDR config: only the 7 required callbacks (+ DeviceIoControl for the proto control
plane). The *2/gamma/HDR-metadata/query-target-info callbacks are FP16-obligated and
rejected without FP16 caps; they return in STEP 7.
- device WDF context type on WdfDeviceCreate; adapter WDF context type on the init attrs.
Debugging note: cdb is reliable via live-attach (go-file gate to avoid the
IsDebuggerPresent race) but cdb -z static hangs on the VM; iddcx WPP needs the control
GUID (TMF GUIDs are not it). Diagnostics trimmed; log.rs dbglog kept for STEP 4+.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
DECISIVE: installed the pre-built UPSTREAM virtual-display-rs (Rust wdf-umdf IddCx)
driver on the SAME box -> Status=OK. So a Rust IddCx driver inits an adapter here,
self-signed, right now. My wdk-sys driver still fails ONLY at IddCxAdapterInitAsync
(0xc000000d) despite matching virtual-display-rs on EVERY inspectable dimension:
- same iddcx 1.10 headers+stub
- IDDCX_ADAPTER_CAPS + IDD_CX_CLIENT_CONFIG byte-perfect (offsets match C header)
- runtime pointers all valid/non-null (names .rdata, version stack, dev handle)
- identical IddFunctions[idx]+IddDriverGlobals dispatch; indices 0/1/2
- matched the minimal link (tested vendored wdk-build WITHOUT OneCoreUAP/
NODEFAULTLIB/OPT/INTEGRITYCHECK -> still fails; export pollution ruled out)
- device context, no device interface (control via EvtIddCxDeviceIoControl), init order
The IddCx ClassExtension ETW provider emits no decodable reason (WPP/kernel-debugger
only). The remaining difference is the wdk-sys IddCx binding itself, invisible to
inspection. This commit keeps the upstream-matching structure (device context, no
interface) + the on-glass instrumentation; vendored wdk-build reverted to pristine.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On-glass diagnosis narrowed decisively. PROVEN it is the driver, NOT the box:
enabling the installed SudoVDA devnode -> Status=OK (the box inits a self-signed
IddCx adapter right now). SudoVDA uses the IDENTICAL UmdfExtensions=IddCx0102 and is
built against IddCx 1.10 (DriverVer 1.10.9.289) — exactly our config.
Matched SudoVDA/the oracle on every inspectable dimension, none fixed the
IddCxAdapterInitAsync INVALID_PARAMETER: caps byte-perfect (offsets+sizes vs C +
framework table), minimal SDR adapter fails identically, dispatch byte-identical to
the oracle (IddFunctions[idx] + IddDriverGlobals), IddMinimumVersionRequired=4 (same
as oracle), version pointers, ObjectAttributes, init order, and now an adapter WDF
context type (this commit). The remaining difference is the Rust binary itself vs
SudoVDA C++. Next: capture IddCx ETW/WPP rejection reason (or kernel debugger), or
build the oracle (wdf-umdf Rust) on-glass to isolate Rust-wide vs wdk-sys-specific.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
wstr! used `const W; W.as_ptr()` which points to a temporary dropped at the end of
the statement (dangling) — fixed to `static W` (stable address). On-glass it did NOT
change the IddCxAdapterInitAsync INVALID_PARAMETER, and a minimal SDR adapter
(Flags=NONE + required callbacks only) fails identically, so the caps content +
callbacks are NOT the blocker (offsets are byte-perfect vs C; sizes match the
framework table; dispatch + device are correct). Config restored to FP16 + full HDR
callbacks. Remaining suspects: IDARG_IN_ADAPTER_INIT layout, the missing DeviceContext
(oracle always sets one), or a box/framework regression.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
IddCxAdapterInitAsync still INVALID_PARAMETER. Logged offset_of! for every
IDDCX_ADAPTER_CAPS + IDDCX_ENDPOINT_DIAGNOSTIC_INFO field on the box: ALL match the
expected C x64 layout exactly (caps Flags=4 MaxRate=8 MaxMon=16 Diag=24 Static=80;
diag Trans=4 Friendly=8 Model=16 Manuf=24 HwVer=32 FwVer=40 Gamma=48). So the wdk-sys
bindgen lays the struct out correctly — NOT a layout bug. The caps are byte-identical
to C + match the framework size table + the oracle, yet rejected. Next: runtime
compare vs the oracle (does it init an adapter on this box now?) + WDK-docs deep-dive.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Major on-glass progress on the RTX box. The all-Rust wdk-sys IddCx driver now LOADS
under Secure Boot and runs the ENTIRE init chain: DriverEntry -> WdfDriverCreate ->
driver_add -> IddCxDeviceInitConfig(0x0) -> WdfDeviceCreate -> CreateDeviceInterface
-> IddCxDeviceInitialize -> D0Entry -> init_adapter. Findings:
- Signing was a RED HERRING (the driver loads); std works in WUDFHost (DualSense uses
it too).
- THE unblock: link the iddcx **1.10** IddCxStub (build.rs now picks the highest
version-aware), not 1.0 — the 1.0 stub lacks the version-table symbols AND its
dispatch table mismatched the 1.10 framework, which made IddCxDeviceInitConfig
return INVALID_PARAMETER. With 1.10 the whole chain runs.
- Added a file/OutputDebugString logger (log.rs, matches the DualSense driver) — the
driver was silent; this is how the chain was traced.
- size.rs: framework_struct_size() reads the frameworks authoritative struct sizes
from IddStructures[] (the config keeps size_of=208, validated working).
- adapter.rs: version ptrs + ObjectAttributes(InheritFromParent) + FP16 + framework
caps/diag/version sizes — matches the oracle.
KNOWN WIP: IddCxAdapterInitAsync still returns INVALID_PARAMETER though caps match
the framework size table (88/56/24) + the oracle exactly — likely a subtle wdk-sys
bindgen field-layout detail in IDDCX_ADAPTER_CAPS/IDDCX_ENDPOINT_DIAGNOSTIC_INFO.
CI gate (compile+link) stays green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
adapter.rs: init_adapter(device) builds IDDCX_ADAPTER_CAPS (CAN_PROCESS_FP16,
MaxMonitorsSupported=16, endpoint diagnostics with wstr! PCWSTR names) +
IDARG_IN_ADAPTER_INIT and calls IddCxAdapterInitAsync; EvtDeviceD0Entry triggers it
(idempotent), EvtIddCxAdapterInitFinished stashes the adapter in a OnceLock for
later DDIs. zeroed()+named-field construction dodges the Default-derive +
field-order questions. Compiles + links clean on the box (pf_vdisplay.dll 268KB).
CI gate = compile+link; the on-glass load/enumerate gate needs the box + an INF +
SwDeviceCreate (next).
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>
DriverEntry -> driver_add builds the full IDD_CX_CLIENT_CONFIG (14 IddCx callbacks +
PnP EvtDeviceD0Entry, all stubs with correct PFN signatures) sized via the ported
IDD_STRUCTURE_SIZE! (size.rs), runs IddCxDeviceInitConfig -> WdfDeviceCreate ->
WdfDeviceCreateDeviceInterface(the owned pf-vdisplay GUID, not SudoVDA) ->
IddCxDeviceInitialize. callbacks.rs has all 14 + device_d0_entry; query_target_info
implements HIGH_COLOR_SPACE. edid.rs salvaged verbatim from the oracle. proto gains
interface_guid_fields() (u128 -> Windows GUID fields). Links IddCxStub (the CI gate);
adapter/monitor/swapchain/IDD-push fill the stubs in STEP 3-6.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Graduate the proven iddcx_rt.rs dispatch into wdk-iddcx + add the full DDI set the
pf-vdisplay driver needs: DeviceInitConfig/Initialize, AdapterInitAsync,
MonitorCreate/Arrival/Departure, AdapterSetRenderAdapter (void-returning DDI — its
PFN returns ()), SwapChainSetDevice/ReleaseAndAcquireBuffer2/FinishedProcessingFrame.
One dispatch macro pins each (_IDDFUNCENUM index, PFN_* type) pair exactly once
(the only place table dispatch can be UB). Box-compiles green; IddCxStub link gets
validated when pf-vdisplay (cdylib) consumes it in STEP 2.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M1 step 2 begins. Add the wdk-iddcx (lib, re-exports wdk_sys::iddcx) + pf-vdisplay
(cdylib) workspace members. pf-vdisplay STEP 0 = DriverEntry + WdfDeviceCreate
skeleton + a #[used] _std_link_gate forcing std::thread + OwnedHandle to link, so
the build proves the std surface resolves under the wdk-build UMDF link settings
(kernel32 is /NODEFAULTLIB - std must come via OneCoreUAP). If std fails to link
here, the SwapChainProcessor worker-thread design needs a CreateThread shim before
any callback work (port-plan critique gap #9).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Record the full driver port plan from the iddcx-driver-port-map workflow: the 11
DDIs to wrap, the 15 IDD_CX_CLIENT_CONFIG callbacks, the DeviceContext-owned state
model (single Monitor identity + monitor EvtCleanupCallback RAII), the
pf-vdisplay-proto frame transport, and the 8-step CI/box-gated checklist. Fold in
the adversarial critique: secure-desktop is a BLOCKING gate (do not retire the WGC
relay until proven), define the recreate/concurrency/Reconfigure failure branches,
host<->driver protocol_version lockstep. De-risk status: the full IddCx symbol
surface + .Size machinery is CI-proven present (ae803b2).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port-plan critique #1: convert "the (?i).*iddcx.* allowlist may miss a symbol the
full driver needs" from a box-only surprise into a CI compile gate. New
wdk-probe/src/iddcx_surface_assert.rs size_of-asserts every *2/HDR struct
(IDDCX_TARGET_MODE2/PATH2/METADATA2, IDARG_*RELEASEANDACQUIREBUFFER2 — these embed
DISPLAYCONFIG_*/LUID, which RESOLVE from crate::types: no allowlist gap),
None-asserts all 14 inbound PFN_IDD_CX_* callbacks, and confirms the .Size
machinery (IddStructures/IddStructureCount/IddClientVersionHigherThanFramework/
_IDDSTRUCTENUM::INDEX_*) + the FP16/HIGH_COLOR_SPACE flags. Box-built green; the
wdk-sys binding is proven complete for the ENTIRE driver, not just init. Also
silence the bindgen naming lints in the iddcx module.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First USE of the iddcx binding: a minimal table-dispatch (src/iddcx_rt.rs) over
wdk_sys::iddcx — IddFunctions[_IDDFUNCENUM::<Name>TableIndex] cast to PFN_*,
IddDriverGlobals as implicit arg 1 (the WDF model; ModuleConsts i32 index, not the
oracle NewType .0). The probe EvtDeviceAdd now calls IddCxDeviceInitConfig →
WdfDeviceCreate → IddCxDeviceInitialize → IddCxAdapterInitAsync, exports
IddMinimumVersionRequired=4, and build.rs links IddCxStub (globbed from the SDK
Lib dir that ships iddcx). CI gate = compile + link IddCxStub.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
CI-green @ 6d8c7a5 (run 5548): IddCx bindgens + compiles in wdk-sys with WDF
type-identity. Record the exact generate_iddcx recipe (c++ parse, IDD_STUB,
allowlist_recursively(false), DXGI/OPM/D3D local emit, UINT alias,
translate_enum_integer_types) and that the wdf-umdf fallback is unneeded.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Last UINT errors were all `pub type Type = UINT;` inside bindgen enum modules
(pub mod _DXGI_X {..}) — the top-level UINT alias cannot reach nested modules. C++
parsing made bindgen keep the UINT typedef as the enum underlying repr (C mode
emits a primitive). translate_enum_integer_types(true) emits native u32 reprs, so
the enum modules are self-contained; struct-field UINT stays covered by the
src/iddcx.rs alias.
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>
Last iddcx type gaps: (1) DXGI enum newtypes are `pub use self::_DXGI_X::Type as
DXGI_X` — the `_DXGI_X` module needs allowlisting too (broaden DXGI_.* to
_?DXGI_.*, matching the OPM fix); (2) UINT bindgen raw_line landed in a scope the
bindings cannot see — define `pub type UINT` directly in src/iddcx.rs next to
`use crate::types::*` instead.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
DXGI resolved. Remaining iddcx type gaps: OPM typedefs need their _OPM_* struct
tags too (recursively(false) drops them), D3DCOLORVALUE (an OPM field), and UINT
(unsigned int — absent from crate::types, and allowlist_type does not emit bare
primitive aliases). Broaden to _?OPM_.* + _?D3DCOLORVALUE and raw_line the UINT
alias.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The iddcx bindgen now SUCCEEDS (C++ fix). Generated module had 38 unresolved-type
errors — a bounded set wdk-sys does not bindgen: UINT, DXGI_FORMAT,
DXGI_COLOR_SPACE_TYPE, IDXGIDevice/Resource, 6 OPM_* types. No WDF type is
missing, so the crate::types sharing (type-identity) holds. Allowlist those
families so they emit locally in iddcx.rs (non-conflicting — absent from
crate::types), keeping allowlist_recursively(false).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Direct clang test on the box proved IddCx.h parses with 0 errors as C++ but fails
as C (wdk_default has no --language=c++) — the IDARG_* typedef names hit "must use
struct tag" in C mode. Fix generate_iddcx: --language=c++ + keep -DIDD_STUB +
allowlist_recursively(false) + full codegen, so it emits ONLY IddCx items
(structs, the IddFunctions table enums, DDI fn-ptr typedefs) and references
WDF/Win/DXGI types from wdk-sys via `use crate::types::*` (no re-emission, no
blocklist). Reverted the ENABLED_API_SUBSETS Iddcx entry (it wrongly pulled
IddCx into the C-mode constants/types passes).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The iddcx bindgen failed with IddCxFuncEnum.h "IDDCX_VERSION_MAJOR is not defined"
+ a cascade of "must use struct tag" on IDARG_* types — NOT the feared #515
header conflict (IddCx parsed fine alongside Base+Wdf). IddCx.h needs STUB mode
(function-table dispatch) for the version macros to resolve; add -DIDD_STUB to
generate_iddcx, matching the wdf-umdf oracle. Deliberately NOT WDF_STUB (wdk-sys
parses wdf non-stubbed; desyncing only here would break WDF type-identity).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Vendor the published, self-contained windows-drivers-rs 0.5.1 crates
(wdk-build, wdk-sys) under vendor/ and add a first-class ApiSubset::Iddcx that
bindgens iddcx/1.10/IddCx.h in an extra pass reusing bindgen::Builder::wdk_default
(allowlist_file (?i).*iddcx.* — emits only IddCx items; WDF/DXGI types resolve to
the shared base/wdf bindings, type-identity by construction). Mirrors the existing
gpio/hid/spb subsets exactly: wdk-build gets the enum variant + iddcx_headers()
(UMDF-only), wdk-sys gets generate_iddcx + the iddcx feature + pub mod iddcx.
[patch.crates-io] redirects all wdk-sys/wdk-build (incl. wdk 0.4.1 transitive) to
the patched copies. wdk-probe enables the iddcx feature.
MAKE-OR-BREAK: does IddCx.h bindgen in wdk-sys config without a header conflict
(issue #515) + does the generated module compile (type-identity)? CI answers it.
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>
The LLVM NSIS .exe /S silent install HANGS in the headless SYSTEM CI session
(stuck >15min after download, blocking the single runner). Switch to the portable
clang+llvm-21.1.2-x86_64-pc-windows-msvc.tar.xz (curl + Win11 tar -xf, strip 1) —
deterministic, no installer. And make driver-build run the provision script itself
(idempotent) so it self-provisions LLVM and never races a separate provision run.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Use the provisioned C:\\llvm-21 libclang for the driver build so wdk-sys bindgen
builds clean (the runner default LLVM is a ToT/22-dev with the E0080 layout-test
overflow bug). Queues behind the in-progress LLVM provision on the single runner.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
wdk-sys bindgen layout tests overflow (E0080 on threadlocaleinfostruct etc.) with
the runner default LLVM (a ToT/22-dev build). windows-drivers-rs maintainers
confirm released LLVM 21.1.2 builds clean (discussion #591). Install it to
C:\\llvm-21 (dedicated path; client LLVM untouched); the driver-build job will set
LIBCLANG_PATH there. Idempotent.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
wdk-build errored StaticCrtNotEnabled + the generated wdk-sys layout asserts
overflowed (E0080) — UMDF needs the static CRT. Add the canonical
windows-drivers-rs .cargo/config.toml: explicit target = x86_64-pc-windows-msvc
(separates host proc-macros, which stay dynamic-CRT, from the driver) +
target-feature=+crt-static scoped to that target. DLL now under the triple subdir.
The WDK bindgen itself now runs (it generated out/types.rs) — this is the last
build-config layer before the /INTEGRITYCHECK verdict.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
wdk-sys build script: "missing field driver-model" deserializing
workspace_metadata[wdk] — a workspace build reads the model from the WORKSPACE
metadata, not the package. Set [workspace.metadata.wdk.driver-model] = UMDF 2.31
(all our drivers are UMDF 2.x incl. pf-vdisplay IddCx). Past the Cargo.lock fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
wdk-build find_top_level_cargo_manifest() walks UP from OUT_DIR to the first
ancestor with a Cargo.lock; the relocated CARGO_TARGET_DIR=C:\\t\\drvws hid the
workspace lock (ancestors C:\\t, C:\\ have none) -> the "Cargo.lock should exist"
panic. Drop the override; the driver deps have no deep CMake crates so the
in-tree target stays under MAX_PATH.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
wdk-build requires a Cargo.lock next to the top-level Cargo.toml (it panics
otherwise — "a Cargo.lock file should exist..."). Generated on Linux
(resolution is platform-independent; only the build needs the WDK). Everything
else compiled on the runner — pf-vdisplay-proto, bindgen, wdk-build/sys/macros.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Stand up packaging/windows/drivers/ — the unified driver workspace on crates.io
windows-drivers-rs (wdk 0.4.1 / wdk-sys + wdk-build 0.5.1), retiring the dev-box
../../crates/wdk* path-deps. First member: wdk-probe, the smallest UMDF2 driver
(DriverEntry -> WdfDriverCreate -> EvtDeviceAdd -> WdfDeviceCreate) that
force-links the shared pf-vdisplay-proto ABI crate. It validates on the runner:
wdk-sys bindgen + WDF stub link against the WDK + LLVM, the cross-workspace
no_std proto path-dep, and the produced DLL's PE FORCE_INTEGRITY bit.
windows-drivers.yml gains a driver-build job: cargo build -p wdk-probe (pinning
Version_Number=10.0.26100.0) + a PE inspection that prints whether /INTEGRITYCHECK
is set — the M0 self-signed-load question.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The first provision run installed the WDK (iddcx headers + stampinf appeared) +
cargo-wdk, but the verification threw on two wrong checks: UMDF wdf.h lives at
Include\wdf\umdf\<ver>\ (not under the SDK-version dir), and inf2cat is x86-only
(the search filtered \x64\). Rewrite verification to enumerate the real layout
(wdf\umdf versions, km dir, iddcx versions, tool paths) and fail only on the
build-essential pieces (wdf.h + km + iddcx + cargo-wdk). Skip-check now keys off
iddcx presence (the reliable "WDK installed" signal), so a re-run skips the install.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The windows-amd64 runner has the base Windows SDK + MSVC + LLVM + Rust but NOT
the WDK (probed: km=False, no um/iddcx, no inf2cat/stampinf/devgen) or cargo-wdk,
so the all-Rust UMDF drivers can't build there yet. Adds an idempotent
provisioning script (scripts/ci/provision-windows-wdk.ps1: download wdksetup 26100
-> /q /norestart, cargo install --locked cargo-wdk, then verify km/wdf + iddcx
headers + inf2cat/stampinf + cargo-wdk) and a workflow_dispatch/push workflow that
runs it on the persistent runner (one-time; install persists).
cargo-wdk (not cargo-make) is windows-drivers-rs's current build+package tool
(cargo build -> stampinf/inf2cat/signtool). Driver builds must pin
Version_Number=10.0.26100.0 (the runner also has 10.0.28000.0, which lacks km/crt).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Stage-1 CI for the Windows-host rewrite: a probe job on the self-hosted
windows-amd64 runner that reports the driver toolchain (WDK Include km/ +
iddcx versions, inf2cat/stampinf/devgen/signtool, EWDK, LLVM/clang version,
cargo-make, installed Rust targets) so we know what's provisioned BEFORE
writing driver code, and builds+tests+lints pf-vdisplay-proto on MSVC to prove
the owned ABI crate compiles cross-OS and the CI wiring works. No RTX GPU needed
for any of this (only live NVENC encode needs one — that defers to the RTX box).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First foundation of the Windows-host rewrite (docs/windows-host-rewrite.md): a
self-contained, no_std + bytemuck crate that defines the host<->driver binary
contract ONCE — the control-plane IOCTLs (add/remove/set-render-adapter/ping/
get-info/clear-all) and the IDD-push frame transport (SharedHeader, the
(gen<<40|seq<<8|slot) FrameToken, the Global\pfvd-* name scheme, driver-status
codes). Previously these were hand-duplicated byte-for-byte across
idd_push.rs/frame_transport.rs and sudovda.rs/control.rs with only "must match"
comments; here const size-asserts + bytemuck round-trips make any drift a COMPILE
error.
Clean break from SudoVDA: a freshly-minted interface GUID (not e5bcc234), a
contiguous 0x900 op space (not the gappy 0x800/0x888/0x8FF), a u64 session id (not
the 16-byte GUID + pid-mangling), a single u32 protocol version. Self-contained
(no workspace inheritance, no Windows deps) so the out-of-workspace driver build
graph can path-dep it identically. 7 tests green on Linux; clippy + fmt clean.
Also lands the full rewrite plan in docs/windows-host-rewrite.md (decisions:
greenfield; IDD-push primary incl. secure desktop, WGC+DDA demoted to fallbacks;
unify drivers on windows-drivers-rs + solve /INTEGRITYCHECK; keep GameStream,
default secure).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
HDR (display-driven, matching the WGC path):
- CTA-861.3 HDR EDID (BT.2020 primaries + HDR Static Metadata block) so Windows
offers "Use HDR" on the virtual display. The host FOLLOWS the display's live
advanced-color state, recreating the shared ring at the matching format
(FP16 in HDR / BGRA in SDR) on a toggle — no freeze.
- Always emit Main10/BT.2020-PQ Rgb10a2 while the display is HDR; the client
auto-detects PQ from the HEVC VUI (clients under-report VIDEO_CAP_10BIT).
Generic HDR10 mastering SEI on every IDR.
- Generation-tagged `latest` (gen<<40|seq<<8|slot) + driver `is_stale` re-attach
kill the toggle-time garbage frame and any stale-ring read.
Perf:
- Pipeline the encode loop (Capturer::pipeline_depth; IDD-push = 2): submit N+1
before polling N so the convert/copy on the 3D engine overlaps the NVENC encode
of N on the ASIC. PUNKTFUNK_IDD_DEPTH overrides (1 = synchronous).
- Rotating host output ring (OUT_RING) so the in-flight encode and the next
convert never touch the same texture.
- HDR converts directly from the keyed-mutex slot's SRV into the output ring
(drops the redundant slot->fp16 scratch copy); SDR copies the BGRA slot in.
The slot mutex is held only across the convert/copy, not the encode.
RING_LEN 3->6 for publish headroom.
- Capture-health diagnostic: new_fps vs repeat_fps under PUNKTFUNK_PERF (a low
new_fps at a high send rate means the source isn't compositing, not an encode
stall).
Validated live on the RTX box: 5120x1440@240 HDR streams; driver composes
~180 new fps, encode 240 fps @ ~4.3 ms p50.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Switch the Inno Setup installer's virtual-display driver from the vendored SudoVDA
C++ binary to our own all-Rust pf-vdisplay (validated streaming at 5120x1440@240).
- packaging/windows/pf-vdisplay/: vendored SIGNED driver (pf_vdisplay.dll/inf/cat +
punktfunk-driver.cer, the same cert the gamepad drivers ship), built from
vdisplay-driver/ via deploy-dev.ps1.
- install-pf-vdisplay.ps1 / stage-pf-vdisplay.ps1: mirror the SudoVDA scripts -
trust cert -> gated ROOT\pf_vdisplay node via nefconc (NEVER devgen) -> pnputil
/add-driver /install. Idempotent, best-effort (never aborts the install).
- punktfunk-host.iss + pack-host-installer.ps1: install the pf-vdisplay bundle
under the existing installdriver task.
- Removed the vendored SudoVDA driver + install-sudovda.ps1 + stage-sudovda.ps1.
- README + windows-host.yml: SudoVDA -> pf-vdisplay.
The host's vdisplay/sudovda.rs backend is unchanged - it drives whichever driver
provides the {e5bcc234} interface, now pf-vdisplay. Live installer build/test on
the runner is the remaining step.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The "5-6 stale monitors that never tear down" failure (also seen with SudoVDA):
an orphan from a crashed/killed previous host lingers because the driver watchdog
is kept reset by a still-pinging new session, so it never fires for the orphan.
- Driver (pf-vdisplay control.rs): new IOCTL_CLEAR_ALL (0x804) -> tear down every
monitor. A pf-vdisplay extension; SudoVDA returns invalid for it (ignored), so
the host can issue it unconditionally.
- Host (vdisplay/sudovda.rs): send IOCTL_CLEAR_ALL once on startup (best-effort)
to reap orphans before creating ours; and surface a failing keepalive PING (the
old `let _ =` swallowed it, masking a lost control handle).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
P1 done: a pure-Rust UMDF2 IddCx driver, drop-in compatible with the host's
existing vdisplay/sudovda.rs control plane (the {e5bcc234} interface + the
SudoVDA IOCTL ABI), so the host drives it unchanged. Validated streaming on
glass at 5120x1440@240 — steady 240 fps, ~2.4 ms encode, clean teardown, full
parity with SudoVDA.
- Vendored wdf-umdf-sys / wdf-umdf bindgen crates (MIT, from virtual-display-rs)
+ the SDK-version build.rs fix that resolves the IddCxStub lib path by the WDK
version actually containing um\x64\iddcx, not the max base SDK.
- pf-vdisplay crate: entry/callbacks/context/control/monitor/edid/
swap_chain_processor. Our OWN 128-byte EDID (manufacturer PNK, product
punktfunk — no SudoVDA bytes), a real swap-chain drain (faithful vdd port,
required so DWM keeps compositing), the SudoVDA-compatible IOCTL control plane
(ADD/REMOVE/PING/GET_WATCHDOG/GET_VERSION/SET_RENDER_ADAPTER) + a watchdog that
tears down orphaned monitors when the host stops pinging.
- deploy-dev.ps1: stage + sign + stampinf (date.time DriverVer) + Inf2Cat +
install, codifying the "bump DriverVer or pnputil keeps the old binary" gotcha.
- docs/windows-virtual-display-rust-port.md: investigation, the on-glass
validation, and the two traps that cost time (Session-0 measurement +
accumulated device-state needing a reboot).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Discovery: replace the flaky per-OEM NsdManager with the same mdns-sd browse
the Linux/Windows clients use, in the Rust core over JNI and polled by Kotlin
(discovery.rs + nativeDiscovery{Start,Poll,Stop}); Kotlin keeps only the Wi-Fi
MulticastLock + permission UX. IPv4-only (the core can't dial a bare/scoped v6
literal); daemon + fold-thread cleanup on every failure path; field
sanitization so a rogue advert can't corrupt the picker snapshot. Discovery
now starts regardless of NEARBY_WIFI_DEVICES (raw multicast only needs the
MulticastLock) — a denial no longer kills it forever. ParseTxtTest replaced by
ParseRecordTest.
Hosts: hide already-saved hosts from the "Discovered" section (match by
fingerprint, else address:port — mirrors the Apple client); add an optional
Name field to the Add-host sheet and a Rename action on saved cards.
Input: touch -> absolute mouse "direct pointing" like the Apple client — the
host cursor follows the finger (new nativeSendPointerAbs -> MouseMoveAbs). Tap
= left click, two-finger tap = right click, two-finger drag = scroll,
tap-then-drag = left-drag, three-finger tap = HUD toggle.
Settings: revert the dropdowns to the stock ExposedDropdownMenuBox look (a
controller-focus UI will come separately); even out the Add-host field gaps.
Docs updated (CLAUDE.md, client READMEs, docs-site status).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Windows installer ballooned to 154 MB and installed forever because the node-server
bundle externalized the WHOLE @unom/ui dependency tree (payload, lexical, date-fns,
prismjs…) to .output/server/node_modules — 47,567 files / 730 MB copied into Program
Files. Set Nitro `noExternals: true` so every dependency is bundled + tree-shaken into the
server output: .output drops to ~75 files / 10 MB, and the bare external imports
(srvx, seroval…) bun couldn't resolve at runtime are gone — so the console runs on bun
(no node, no node_modules), which is the issue we previously worked around with node.
Windows installer now ships bun.exe + the ~75-file .output (was node.exe + a node_modules
forest) and runs `bun .output\server\index.mjs`:
- windows-host.yml: fetch a pinned portable bun (build tool AND shipped runtime); drop the
node fetch + the .output/server install; smoke-boot under the bundled bun.
- pack-host-installer.ps1 / punktfunk-host.iss: -NodeExe -> -BunExe; stage {app}\bun\bun.exe.
- web-run.cmd / build-web.ps1: run/restart on bun; docs updated.
Net win everywhere: the Linux .deb shrinks (node still runs the self-contained output), and
the docker web image — which already ran `bun run .output/server/index.mjs` with only
.output copied — is fixed (the externals had no node_modules to resolve at runtime).
Validated locally: noExternals build = 75 files / 10 MB; node AND bun both serve /login
(200) + static assets (200) + gate /api (401).
(A true single binary via `bun build --compile` is blocked for now: Nitro serves public
assets from an import.meta-relative path `--compile` doesn't embed (/$bunfs/public); the
75-file payload is the clean result.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Diagnosed from the first run: only the iPad shots were produced. The runner
lacks an "iPhone 16 Pro Max" device, is headless (no window server -> the macOS
window capture's app window never appears), and the Tier-3 tvOS build-std slice
failed.
- screenshots.sh: shoot_sim now creates a throwaway Simulator (matching device
type + newest available runtime) when the runner has no matching device, so
the iPhone 6.9" shots are reproducible instead of skipped.
- apple.yml: scope the CI job to the two REQUIRED iOS sizes (iPhone 6.9" +
iPad 13"), captured via `simctl io screenshot` (no Screen Recording grant
needed). Drop macOS (headless runner has no window server) and tvOS (build-std
slice) from CI — generate those locally with `tools/screenshots.sh macos tvos`.
Faster, deterministic xcframework build (BUILD_IOS=1 only).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Gitea's artifact storage identifies as GHES, which @actions/artifact v2+
(upload-artifact@v4) refuses outright. v3 uses the older artifact API Gitea
supports; the downloaded artifact is still a zip. (The capture itself already
worked — 5 macOS scenes were produced; only the v4 upload failed.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
ISCC aborted compiling the installer at the web-console [Code] section: a comment
`{ ... {tmp} is auto-cleaned. }` — Pascal `{ }` comments don't nest, so the `}` in
`{tmp}` closed the comment early and `is auto-cleaned. }` parsed as code ("Identifier
expected"). Reword to drop the brace. (All other {app}/{tmp} uses are `;` line-comments
or code strings, which are fine.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A DEBUG-only "shot mode" renders one mock-populated screen full-bleed
(PUNKTFUNK_SHOT_SCENE=<name> -> ScreenshotHostView instead of ContentView),
so the OS can screenshot the REAL, fully-rendered UI. tools/screenshots.sh
drives it: screencapture for the mac window, `simctl io booted screenshot`
for the iOS/iPad/tvOS Simulators, at exactly the App Store Connect sizes.
ImageRenderer was tried first and rejected: it can't rasterize this app's
chrome (NavigationStack, Form/TabView, Liquid-Glass/NSVisualEffect all render
black or the "can't render" placeholder). Capturing the live window/Simulator
avoids that. Only the stream hero is synthetic (StreamView needs a live
connection) - a synthwave frame + the real glass HUD, overridable via
PUNKTFUNK_SHOT_HERO.
CI: a new `screenshots` job in apple.yml builds the iOS (+ tvOS best-effort)
xcframework slices, runs the harness per platform best-effort, and attaches
the result as a single zip artifact (punktfunk-appstore-screenshots). It is
isolated from the build/test job and skipped on PRs, so a capture gap (missing
Simulator runtime, or no Screen Recording grant for the mac window capture)
never reds the core signal.
Generated PNGs (clients/apple/screenshots/) are gitignored.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The first windows-host run with the bundled console failed at "bun not found": the
self-hosted runner executes as SYSTEM, so the dev user's bun (and its ~/.npmrc with the
@unom registry token) aren't on PATH. Make the web-build step self-sufficient:
- Install bun via bun.sh/install.ps1 when it isn't already present (checking PATH +
the SYSTEM/Public profile locations first), like deb.yml bootstraps it.
- Write the private @unom registry mapping + auth token (REGISTRY_TOKEN) into the SYSTEM
home .npmrc so `bun install` can fetch the @unom packages — kept out of the project
tree and the shipped .output bundle (.output\server\.npmrc stays mapping-only).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Windows host installer shipped only the host exe + SudoVDA driver + FFmpeg, so a
fresh install had no web management console — required for basically every user (status,
paired devices, the PIN pairing flow). The console was only ever set up by hand on the
dev box (build-web.ps1 + a hand-made PunktfunkWeb task whose web-run.cmd wasn't even
committed). Bundle it into the same installer, mirroring the proven Linux punktfunk-web
deploy.
- windows-host.yml builds the Nitro node-server console (bun, deb.yml's shape) + fetches
a pinned portable Node, smoke-boots it under node (/login == 200) to gate the build, and
hands web/.output + node.exe to the pack script.
- pack-host-installer.ps1 gains -WebDir/-NodeExe and stages the .output tree, node, and
the two new scripts into the non-WOW64-redirected build area.
- punktfunk-host.iss lays the payload into {app}\web\.output + {app}\node\node.exe, adds
a wizard page for the console login password pre-filled with a crypto-random default
(shown on the finish page; kept on upgrade), and runs web-setup.ps1.
- web-setup.ps1 writes the ACL'd %ProgramData%\punktfunk\web-password (Administrators +
SYSTEM), registers the PunktfunkWeb scheduled task (boot, SYSTEM, restart-on-failure ->
web-run.cmd -> node on :3000), opens inbound TCP 3000, and starts it. web-run.cmd
sources the host's mgmt-token + the password and runs the bundled node.
- The console proxies the host's loopback mgmt API with the host's own
%ProgramData%\punktfunk\mgmt-token (no host-code change). Uninstall removes the task +
firewall rule.
Validated locally: bun build -> node-server bundle, node boot serves /login (200) and
gates /api (401). The Windows-only bits (ISCC compile, scheduled task, password page,
firewall) validate on the Windows runner CI + on-glass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Near-term 7.1 channel bed; moonshot object-based spatial audio via
Wine/Proton (where dynamic objects are currently discarded) with
client-side head-tracked spatialization.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Windows virtual gamepads now have zero external dependencies - ViGEmBus is removed.
- DualShock 4: Windows UMDF backend (inject/dualshock4_windows.rs + dualshock4_proto.rs),
reusing the DualSense SwDeviceCreate game-detection identity fix. The one UMDF driver serves
the DS5 or DS4 identity/descriptor/features/strings per a device_type byte the host stamps into
shared memory. Driver also gains IOCTL_HID_GET_STRING and a 41-byte calibration feature.
- Xbox 360: a new UMDF2 XUSB companion driver (packaging/windows/xusb-driver/) that registers
GUID_DEVINTERFACE_XUSB and answers the buffered XInput IOCTLs from a shared section, so classic
XInputGetState/SetState work with no kernel bus driver. inject/gamepad_windows.rs is rewritten
to drive it and the vigem-client dependency is removed. Xbox One folds to the 360 XInput path.
- Installer: vendor + pnputil-install the three UMDF drivers (packaging/windows/gamepad-drivers/
+ install-gamepad-drivers.ps1, wired into pack-host-installer.ps1 + punktfunk-host.iss).
- Multi-pad: the host stamps each pad index into the device Location (pszDeviceLocation); the
driver reads it via WdfDeviceAllocAndQueryProperty to map its own *-shm-<index>, with
UmdfHostProcessSharing=ProcessSharingDisabled giving each pad its own host (per-pad statics).
Validated live on the Windows host: Cyberpunk native DualSense detection, DS4 identity + descriptor,
XInputGetState + rumble round-trip, two pads -> two distinct XInput slots, and a full installer build.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
dualshock4.rs left `cargo fmt --all --check` red on main (it landed with the
Windows-host DualSense work): a standalone comment placed directly after a line
ending in a trailing comment gets absorbed and re-aligned to the trailing-comment
column. A blank line before the comment block keeps rustfmt happy — and the
comment readable.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The canary/stable split (0205c7b) gated the tvOS archive/upload — and its
xcframework slice — to vX.Y.Z tags, while moving iOS/macOS onto canary main
pushes. No tag has been cut since (both existing tags predate the split), so
tvOS stopped reaching TestFlight entirely while iOS/macOS kept shipping on canary.
Build the tvOS tier-3 slice unconditionally again (BUILD_TVOS=1; the nightly
-Zbuild-std std is cached on the self-hosted runner) and drop the tag gate on the
tvOS step so its if: matches the iOS / macOS App Store steps exactly — tvOS now
uploads on canary main pushes + stable tags + dispatch, same as the others.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GameController's CHHapticEngine never reaches the DualSense's motors on macOS — its
adaptive triggers and lightbar work, but rumble stays silent (a documented platform
gap). Drive the motors directly via the DualSense HID output report instead, the way
SDL and the Linux hid-playstation driver do — the same report that already rumbles
the pad on a Linux host. Confirmed live on macOS.
- DualSenseHID (macOS): opens the Sony DualSense via IOHIDManager and writes the USB
(0x02, 48 bytes) and Bluetooth (0x31, 78 bytes + CRC32) output reports through
IOHIDDeviceSetReport. Allowed under the App Sandbox by the existing device.usb +
device.bluetooth entitlements; coexists with GameController (non-seized open).
Flags mirror the kernel driver (COMPATIBLE_VIBRATION | HAPTICS_SELECT +
COMPATIBLE_VIBRATION2); valid_flag1 = 0 so a rumble report leaves the
GameController-managed lightbar / triggers / player LEDs untouched.
- RumbleRenderer routes a DualSense to the HID backend and keeps CoreHaptics for
every other pad, fixing both live sessions and the test panel (shared renderer).
- CoreHaptics path reworked too: bake the target intensity + an explicit sharpness
into the continuous event (the dynamic-parameter scaling is silent on controller
engines) and tear down outside the inout access to fix a latent exclusivity hazard.
Adds a DEBUG-only Settings -> Controllers -> "Test Controller" panel (ControllerTestView
+ ControllerTester) that shows live input and fires rumble / adaptive triggers /
lightbar / player LEDs straight at the pad, with a readout of the active rumble backend
("DualSense HID - USB/Bluetooth"). Used to validate the fix.
Tests: DualSenseHIDTests pins the USB/BT report layout and the BT CRC32 (canonical
0xCBF43926 check vector). Debug + release build clean; gamepad suite green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reads PUNKTFUNK_NVENC_LIB_DIR/LIBCLANG_PATH/CMAKE_POLICY_VERSION_MINIMUM directly from
Machine scope into the process, so the build is correct even when the SSH/parent shell
predates setup-build-env.ps1 (env is inherited at spawn).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
scripts/windows/: setup-build-env.ps1 persists the NVENC build env (Machine scope:
PUNKTFUNK_NVENC_LIB_DIR, LIBCLANG_PATH, CMAKE_POLICY_VERSION_MINIMUM -- no FFMPEG_DIR, the
nvenc build doesn't link libavcodec). deploy-host.ps1 rebuilds --release --features nvenc and
restarts the PunktfunkHost service with .bak rollback on build/start failure. build-web.ps1
rebuilds the Nitro web console (bun build, node runtime) and restarts the PunktfunkWeb task.
README documents the flow -- a redeploy is now a single script call.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The virtual DualSense is a correct, complete DS5 at the HID level (SDL3 reports PS5) and
input works, but a game's native DualSense path (Cyberpunk) doesn't detect the
software-enumerated (SWD) device that SDL/HIDAPI accept. Captures the diagnosis, the on-box
layout + tools (SDL oracle, dualsense-windows-test, driver rebuild recipe), and the on-glass
next experiments (WGI/RawInput/GameInput enumeration) so the work continues from any machine
without agent memory.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
create_swdevice now succeeds. The two requirements (each E_INVALIDARG otherwise): the
enumerator name must have no underscore (use "punktfunk"), and the completion callback is
mandatory (the docs mark pCallback [in], not optional -- NULL is rejected). Back on the
typed windows-rs SwDeviceCreate (a raw-FFI diagnosis confirmed it's the OS, not the
binding), parameterized by pad index (instance pf_pad_<index>), waiting on the callback.
Per-session device: created on connect, SwDeviceClose'd on drop -- no leftovers, no phantom.
Live-verified on the RTX box: device materializes, the UMDF driver binds, SDL3 identifies it
as a PS5 ("DualSense Wireless Controller"), input flows; removed on disconnect. The
dualsense-windows-test CLI now cycles input + prints any 0x02 feedback for diagnosis.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
cargo audit fails on the rsa "Marvin Attack" advisory, which has NO fixed release
(the constant-time rewrite is still unreleased upstream) and rsa is required for
GameStream/Moonlight pairing. The attack targets RSA *decryption* (PKCS#1 v1.5
padding oracle); the host uses rsa ONLY for PKCS#1 v1.5 signing/verifying
(gamestream/cert.rs + pairing.rs), never for decryption, so the vulnerable path is
not exercised. Add the documented .cargo/audit.toml ignore with the justification.
The 3 unmaintained warnings (audiopus_sys / paste / rustls-pemfile) are left visible
on purpose — `cargo audit` does not fail on them, and they carry a maintenance signal.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Windows host was NVIDIA-only (NVENC) with an openh264 software fallback. Add
AMD AMF and Intel QSV via libavcodec — the Windows analogue of the Linux VAAPI
backend — so one installer serves all three GPU vendors.
- encode/ffmpeg_win.rs: new WinVendor{Amf,Qsv} encoder. System-memory NV12/P010
readback (default, robust) + opt-in zero-copy D3D11 (PUNKTFUNK_ZEROCOPY: shares
the capturer's ID3D11Device; AMF takes AV_PIX_FMT_D3D11, QSV derives a QSV frames
ctx and maps) with a system fallback for the format-group mismatch the capturer's
video-processor fallback can produce. HDR Main10 (P010 + BT.2020/PQ VUI; an
Rgb10a2->P010 swscale covers the shader fallback).
- encode.rs: Codec::amf_name/qsv_name; open_video + windows_resolved_backend()
resolve PUNKTFUNK_ENCODER=auto|nvenc|amf|qsv|sw via a DXGI adapter VendorId probe.
- capture/dxgi.rs: gpu_mode mirrors the resolved backend (D3D11 NV12/P010 for AMF/QSV).
- gamestream/serverinfo.rs: GPU-aware codec advertisement (windows_codec_support;
AV1 gated to RDNA3+/Arc, like the VAAPI path).
- Cargo.toml: amf-qsv feature (optional ffmpeg-next in the windows target block).
- CI/installer: windows-host.yml sets FFMPEG_DIR + builds --features nvenc,amf-qsv;
the Inno installer bundles the FFmpeg DLLs; host.env default nvenc -> auto.
CI-green target; AMF/QSV not yet on-glass validated (no AMD/Intel Windows box in the
lab) — NVENC stays live-validated. An adversarial-review pass caught + fixed real
FFI bugs (AV_PIX_FMT_P010 is a macro -> P010LE; windows-rs 0.62 GetImmediateContext/
GetDesc1 return Result; AV_HWFRAME_MAP_* is a bindgen enum with no BitOr).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
DualSenseWindowsManager now SwDeviceCreate's the pf_dualsense devnode per session
(SwDeviceClose on drop), matching the Linux UHID pad's lifecycle. It's best-effort:
SwDeviceCreate currently hits an unresolved E_INVALIDARG when a completion callback is
passed (an underscore in the enumerator name was a second cause, fixed by using
"punktfunk"), so on failure the host keeps the section + data plane and falls back to
an out-of-band devnode (installer/devgen) — see docs/windows-dualsense-scoping.md.
Add a `dualsense-windows-test` host CLI that drives the manager (create devnode + push
a frame + hold), used to validate the path. Live on the RTX box: the manager creates
the section + pushes report 0x01 and a devnode serves it to a HID read (b1=0xC0,
b8=0x28) — the host-side data plane works end to end.
cargo check + clippy -D warnings clean on x86_64-pc-windows-msvc.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The host<->driver channel is the shared-memory section (hidclass blocks the device
stack and UMDF has no control device), so the first-attempt in-driver IOCTL channel
never fired. Remove it: the custom device interface, IOCTL_PFDS_SET_INPUT/GET_OUTPUT,
the output queue, and the on_set_input/complete_one_read/deliver_output helpers. The
driver keeps the HID handshake, the 8ms read timer fed from the shared section, and
on_output_report publishing the game's 0x02 to the section. Rebuilt + reloaded + the
channel still verifies both directions live on the RTX box.
Also list `pf_dualsense` as a second hardware id (alongside `root\pf_dualsense`) so the
host's SwDeviceCreate'd software device binds the same driver as a devgen one.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Wire the Windows UMDF DualSense driver into the host as a real pad backend, so a
client that requests a DualSense gets a genuine one on a Windows host (instead of
folding to Xbox 360).
- Extract the transport-independent DualSense contract (DsState + from_gamepad,
serialize_state, parse_ds_output, DUALSENSE_RDESC, feature blobs, DS_* consts)
out of the Linux-only UHID backend into inject/dualsense_proto.rs, shared by both
platforms; dualsense.rs is now just the /dev/uhid plumbing.
- Add inject/dualsense_windows.rs: DualSenseWindowsManager mirroring the Linux
DualSenseManager (same new/handle/apply_rich/pump/heartbeat surface) over a
DsWinPad that creates the Global\pfds-shm-<idx> section (CreateFileMappingW +
SDDL D:(A;;GA;;;WD) so WUDFHost can open it), writes serialize_state -> input
slot, polls output_seq -> parse_ds_output -> rumble/hidout callbacks.
- Un-gate the seam: PadBackend::DualSenseWindows arm; pick_gamepad gains a
windows flag (DualSense honored on linux||windows; DS4/Xbox One stay Linux-only).
Verified: Linux cargo test gamepad_resolution_precedence + clippy clean; Windows
cargo check + clippy -D warnings clean (on the RTX box). Device lifecycle still
uses an out-of-band devnode (devgen/installer); SwDeviceCreate per session is next.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A self-authored UMDF2 HID minidriver (packaging/windows/dualsense-driver) that
presents a virtual Sony DualSense (VID 054C/PID 0CE6) on Windows — adaptive
triggers / lightbar / rumble that ViGEm structurally cannot deliver.
Validated live on an RTX box (Win11 25H2, Secure Boot ON): the self-signed driver
loads, Steam recognizes it as a genuine DualSense, and a game's 0x02 output report
reaches the driver. The host<->driver channel is a named shared-memory section
(Global\pfds-shm-<idx>) the host creates and the driver maps from its timer: input
report 0x01 host->driver, output report 0x02 driver->host — input and output proven
both directions live. This bypasses hidclass, which gates both a custom device
interface and custom IOCTLs on the HID node, and UMDF has no control device.
Built in Rust on microsoft/windows-drivers-rs. The load wall was the PE
FORCE_INTEGRITY bit that wdk-build sets via /INTEGRITYCHECK (forces a CI-trusted
page-hash signature a self-signed cert cannot satisfy) — cleared post-build. See
packaging/windows/dualsense-driver/README.md for the build/sign/install recipe.
Deferred: SwDeviceCreate per-session device lifecycle; removing the inert in-driver
IOCTL-channel code; full on-glass session test.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Scalar's /api reference injects a *global* `body { background-color:
var(--scalar-background-1) }` (via its linked stylesheet + a runtime
<style id=scalar-style>) that TanStack doesn't remove on a client-side route
change. After navigating /api -> /docs without a reload, that rule kept
painting the docs body: Scalar's stock gray (#0f0f0f) while .dark-mode lingered
on <body>, or transparent once the class was gone. A hard reload was fine
because the stylesheet was never loaded there.
Fix: give --scalar-background-1 a global fallback = --color-fd-background so any
non-API page paints its own surface while Scalar's sheet lingers; /api itself
overrides it via the higher-specificity body.{dark,light}-mode rule. Also strip
the leftover #scalar-style/#scalar-refs nodes and body mode-class when /api
unmounts so the DOM matches a fresh load. Verified light + dark via headless
CDP: post-nav docs body now equals a fresh reload (#141019 / #f0ebff).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Scalar puts .light-mode/.dark-mode on document.body and renders customCss
*before* its built-in theme preset in the same <style> tag, so a bare
.dark-mode override loses at equal specificity and the stock #0f0f0f gray
showed through. Scope the palette to body.{dark,light}-mode (0,1,1) so it beats
both the linked base sheet and the in-component preset, and add a full
light-lavender palette to match the docs light surface.
Drive Scalar's darkMode from the resolved Fumadocs theme (next-themes) instead
of hard-locking it on, so toggling the docs theme switch flips the API
reference too; the React wrapper's updateConfiguration effect live-swaps the
body mode class.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A push to main publishes canary builds to canary channels (fast iteration,
unchanged); a single vX.Y.Z tag releases every platform at one version to the
stable channels and attaches all artifacts (.deb/.rpm/.msix/.apk/.aab/.dmg +
flatpak/decky/host-installer) to one Gitea Release. Collapses the
host-v*/win-v*/host-win-v* tag namespaces into v* — the channel split makes the
version-shadow bug structurally impossible (canary and stable are separate repos,
never a shared version line).
- scripts/ci/gitea-release.{sh,ps1}: one idempotent release helper
(create-or-fetch + delete-before-upload), replacing 3 copy-pasted inline blocks
and fixing their latent 409-on-reupload bug; prerelease flag auto-derived from
the tag (an -rc tag won't shadow "Latest")
- channels: apt canary/stable distributions; rpm *-canary/base groups; flatpak
canary/stable OSTree branches + a 2nd .Canary.flatpakref; generic-registry
canary/ vs latest/ aliases; Play internal/alpha; Apple TestFlight vs notarized DMG
- android versionName threaded through gradle (versionCode stays run_number);
Apple canary = TestFlight-only (no DMG/tvOS); canary base bumped to 0.3.0
- docs: new docs-site channels.md (subscribe table + cut-a-release runbook +
box migration), refreshed ci.md workflow table + packaging READMEs
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extends virtual-controller support beyond Xbox 360 + DualSense. Goal: a
physical Xbox One or PS4 pad on the client gets a near-native matching virtual
pad on the host, auto-resolved from the controller type.
Protocol/core:
- GamepadPref gains XboxOne (wire 3) + DualShock4 (wire 4); to_u8/from_u8/
from_name/as_str + C ABI PUNKTFUNK_GAMEPAD_XBOXONE/_DUALSHOCK4 constants
(compile-time guard ties them to the enum). Single-byte wire form is
unchanged, so it's forward-compatible (older peers degrade to Auto).
Host (Linux):
- New UHID DualShock 4 backend (inject/dualshock4.rs) bound by hid-playstation:
lightbar, touchpad, motion, rumble — DualSense minus adaptive triggers /
player LEDs / mute. Reuses the DualSense pure state + button mapping; only the
report byte layout, the real-DS4 HID descriptor, the GET_REPORT handshake
(0x12 MAC mandatory; 0x02 calibration; 0xa3 firmware) and the touchpad
resolution (1920x942) differ. Touchpad/motion ride the existing 0xCC plane,
lightbar the 0xCD Led plane (deduped); rumble the universal 0xCA plane.
- Xbox One/Series is the uinput Xbox-360 backend parameterized with the One S
USB identity (045e:02ea) for matching glyphs — XInput-identical otherwise.
- PadBackend dispatch + resolver handle both; off Linux the UHID pads and
One/Series fold into Xbox 360. Windows-host DS4 (ViGEm) deferred.
Clients (auto-resolve physical pad -> virtual type, plus manual settings):
- Linux/Windows (SDL3): SDL_GAMEPAD_TYPE_PS4 -> DualShock 4, _XBOXONE ->
Xbox One; PadInfo carries the resolved pref; DS4 touchpad/motion capture +
lightbar already type-agnostic. Linux settings combo + label updated.
- Apple (GameController): GCDualShockGamepad/GCXboxGamepad detection, DS4
touchpad capture, settings picker entries.
- Android (Kotlin): InputDevice VID/PID auto-detect (matching the other
clients) + settings entries.
- probe: --gamepad help/aliases.
Also hardens the Android JNI boundary: wrap the teardown + poll-thread shims in
catch_unwind so a panic degrades to a logged no-op instead of aborting the app.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Dashboard session card: the header stacks the title above the action buttons
on narrow screens (flex-col -> sm:flex-row) and the button group wraps
(flex-wrap), so "Request IDR" / "Stop session" no longer overflow the card.
- Mobile bottom nav: give each label a fixed two-line-tall centered box so a
1- or 2-line label (labels vary by locale) keeps every tab icon at the same
height.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@scalar/api-reference-react@0.9.47's entry imports createApiReference but does
NOT import its own style.css (nor inject it at runtime), so /api rendered with
no Scalar CSS at all. Import the sheet as a route-scoped <link> (?url +
head.links, same pattern as the root app.css) so it loads for SSR + the
client-side Vue mount. The brand customCss still themes on top.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- console: remove @unom/ui's specular "material" gloss (drop UnomProviders +
the material.css import) so components render flat like the marketing site;
the violet brand + Geist stay.
- mobile bottom tab bar: center the labels (w-full text-center, leading-tight)
and even out the per-tab layout.
- docs /api: roll the punktfunk dark-violet palette across the whole Scalar
reference (surfaces/text/sidebar/links/buttons/method colours via the full
--scalar-* token set), locked to dark (hideDarkModeToggle).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Move the management console (web/) off shadcn/ui to the shared @unom/ui
design system the marketing site + docs are built on, on the punktfunk
violet brand over dark chrome:
- Add @unom/ui/@unom/style/motion/radix-ui/zod + Geist; web/.npmrc maps the
@unom scope (packages are public-read, so CI needs no npm auth).
- styles.css: one dark-violet palette (#141019/#1c1530, brand #6c5bf3 ->
#a79ff8) exposed under BOTH the shadcn token names the routes use and
@unom/ui's contract, so routes + components both resolve; pulls in
@unom/ui's material gloss + easings.
- components/ui/* now back onto @unom/ui (AnimatedButton/InputText/Label/
AnimatedCard); brand-mark/wordmark/logo replace the generic Radio icon in
the shell + login.
- MaterialProvider (specular gloss) at the root. No UI sounds, like the site.
docs-site: new /api route renders the host management REST API as an
interactive Scalar reference (reads public/openapi.json, a snapshot of
docs/api/openapi.json), branded violet and linked from the top nav, the
docs sidebar, the landing page, and host-cli.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Rename steam-deck-host.md → steamos-host.md (nav + install table updated).
- Lead with the rationale: SteamOS host support targets the upcoming Steam
Machine; the Steam Deck is the SteamOS device validated against today.
- Soften the WiFi note: ~250 Mbps was our testing on one device/network,
not a universal ceiling — other SteamOS hardware/drivers/bands may do more.
- Generalize Deck-specific language to SteamOS devices throughout.
- Document --no-gamestream (secure native-only) + GameStream-compat caveat.
- decky README: drop stale `serve --native` (now just `serve`).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Completes the GameStream-opt-in posture (54b75c9) on the SteamOS path: the installer keeps
Moonlight compat on by default (`serve --gamestream`, the Deck commonly streams to Moonlight),
but `--no-gamestream` now installs a secure native-only host with no GameStream on-path surface
(plain-HTTP pairing / legacy GCM nonce reuse — security-review #5/#9; native clients only).
Documented in the installer --help; the SteamOS host doc references it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Follows the security audit (#5/#9): the GameStream-compat plane carries inherent on-path weaknesses
that can't be fixed on the wire without breaking stock Moonlight — its pairing runs over plain HTTP
(#9, MITM-able during the pairing window) and its legacy control encryption can reuse GCM nonces (#5,
a passive eavesdropper can recover/forge input). The native punktfunk/1 plane (SPAKE2 PIN pairing +
per-direction AEAD nonces) has neither. So flip the default to secure-by-default:
- `serve` → native punktfunk/1 plane + management API ONLY (no GameStream surface).
- `serve --gamestream` → ALSO the GameStream/Moonlight-compat planes (nvhttp pairing, RTSP, ENet
control, _nvstream mDNS). Opt-in, logged with a trusted-LAN caveat. `--moonlight` is an alias.
- The native plane is now ALWAYS on in `serve` (`--native` is a kept-for-compat no-op); the unified
GameStream+native host is `serve --gamestream`.
`gamestream::serve` gates the GameStream spawns (nvhttp/rtsp/control/mdns) on the flag; the native
plane + mgmt + native-pairing handle always run.
To avoid silently regressing validated Moonlight deployments, the explicit deployment configs PRESERVE
Moonlight via `--gamestream` (each documents dropping it for a secure native-only host): the Linux
systemd unit, the Steam Deck installer, and the Windows service default (DEFAULT_HOST_CMD). The bare
`serve` default (new/manual use) is secure.
Docs swept to match (host-cli, moonlight, quickstart, install, packaging READMEs, CLAUDE.md, README,
…): Moonlight setup now instructs `--gamestream`; native/console refs use bare `serve`. OpenAPI
regenerated (a stale "run `serve --native`" string). fmt + clippy clean; 94 host tests green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Addresses the lower-severity findings from docs/security-review.md (#4-#12). Each fix was
adversarially re-reviewed (5-agent pass); two review catches folded in (the Apple client's
GET /library cert path; an RTSP header-cap bypass + a spawn-panic counter leak).
- #4 [low] mgmt mTLS-paired-cert no longer grants full admin. A paired STREAMING cert authorizes
only a read-only allowlist (GET /host,/compositors,/status,/clients,/native/clients,/library);
every state-changing route and every PIN-exposing route (/pair, /native/pair) requires the
operator's bearer token. New cert_auth_is_a_read_only_allowlist test. (/library kept on the
allowlist — the native clients browse it cert-only; its mutations stay token-only.)
- #6 [low] RTSP pre-auth DoS bounds: a concurrent-connection cap (RAII slot guard), a per-read
timeout (slow-loris), and Content-Length/header/message size caps — closing an unauthenticated
slow-loris / memory-growth / thread-exhaustion vector on TCP 48010.
- #11 [info] A FEC reconstruction failure is now a counted drop (discard the block, keep the
session) instead of being stream-fatal — a lossy link can't be torn down by one bad block.
- #10 [info] Fixed ALPN ("pkf1") on both native QUIC endpoints (defense-in-depth; a deliberate
coordinated client+host upgrade — a new host rejects an ALPN-less old client).
- #8 [info] Constant-time GameStream pairing phase-4 hash compare (crypto::ct_eq).
- #7 [low] New VirtualDisplay::set_launch_command carries the launch command per-session on the
GameStream path (no process-global env stomp under concurrent sessions); native path keeps the
env under today's single-session model (documented; plumb per-session with concurrent sessions).
- #5 [low] Legacy GameStream GCM nonce reuse: documented as inherent to Nvidia's old-style control
encryption (Apollo/Moonlight identical; key is client-known) — unfixable on the legacy wire; the
real fix is V2 control-encryption negotiation. Code comment at control.rs.
- #9 [info] GameStream plain-HTTP pairing: documented (inherent to GFE compat; use punktfunk/1).
- #12 [low] Web global NODE_TLS_REJECT_UNAUTHORIZED: fix designed (undici dispatcher scoped to the
loopback mgmt fetch) but DEFERRED — needs `bun add undici` in the web build env; reverted to keep
the web working. Latent-only (the loopback mgmt fetch is the console's only outbound TLS).
fmt + clippy -D warnings clean; 94 host + core tests green; no C-ABI/OpenAPI drift. (The HDR
Steps 1-2 client work in the tree is the user's parallel WIP — deliberately NOT included here.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Continues docs/hdr-pipeline-plan.md. Steps 0/1 + Step 2 (Windows/Android) already
landed in 3526517; this is Step 2 (Apple) + Step 3 (all clients). Client-only — no
core/host/ABI change (the 0xCE/next_hdr_meta/color_info surfaces shipped in Step 0).
Step 2 — clients APPLY the host's HDR metadata (each remaps from the wire form: ST.2086
G,B,R order, mastering luminance in 0.0001 cd/m2):
- Apple: connect via punktfunk_connect_ex5 (resurrects the previously-dead HDR pipeline);
nextHdrMeta/colorInfo wrappers + HdrMeta SEI-blob builders; the pump drains nextHdrMeta
-> VideoDecoder.setHdrMeta -> CVBufferSetAttachment of MasteringDisplayColorVolume (24B
BE) + ContentLightLevelInfo (4B BE) on each HDR pixel buffer (correct for the
itur_2100_PQ layer; CAEDRMetadata avoided as ambiguous there).
Step 3 — capability-gate: advertise HDR caps ONLY when the display can present it, so an
SDR display gets a proper BT.709 stream instead of PQ it would mis-tone-map; an HDR
display self-tone-maps from the Step-1/2 mastering metadata.
- Windows: present::display_supports_hdr() (DXGI any IDXGIOutput6 colour space == G2084),
ANDed with the user HDR setting in session.rs; logs the SDR drop.
- Apple: NSScreen.maximumExtendedDynamicRangeColorComponentValue>1 (macOS) /
UIScreen.main.potentialEDRHeadroom>1 (iOS) in SessionModel.
- Android: Settings.displaySupportsHdr (Display.getHdrCapabilities HDR10/HDR10+) passed
through a new hdr_enabled jboolean on nativeConnect; session.rs gates the caps.
Validation: Android native (incl. the jboolean gate) builds + clippy clean via cargo-ndk;
fmt clean. Windows (MSVC), Apple (Swift) and the Kotlin side are CI/on-glass validated —
not compilable on the Linux dev box. Deferred to the RTX box: mid-session Reconfigure
SDR-downgrade on monitor move, and confirming the host emits SDR for an SDR client off an
HDR desktop.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two strands, entangled in punktfunk1.rs, committed together (one builds-green tree).
HDR pipeline Step 0 — glass-to-glass colour-metadata transport (docs/hdr-pipeline-plan.md):
- Protocol/ABI: ColorInfo on the Welcome + a 0xCE HdrMeta datagram carry the source colour
space + HDR10 static mastering metadata (quic.rs, abi.rs connect_ex5 fixing caps=0).
- New platform-independent, unit-tested HDR static-metadata helpers (hdr.rs): chromaticities
(1/50000), mastering luminance (0.0001 cd/m2), MaxCLL/MaxFALL in HDR10/ST.2086 units.
- Capture/encode hooks (capture.rs, encode.rs set_hdr_meta) + Linux client / probe plumbing.
Security-audit hardening — top 3 from docs/security-review.md, each adversarially verified:
- #1 [HIGH] Secret file permissions. The host key.pem/cert.pem and both trust stores are now
written owner-only: 0600 + dir 0700 on Unix (mirrors mgmt_token), best-effort
SYSTEM/Administrators/OWNER-only icacls DACL on Windows (%ProgramData% is Users-readable).
Closes a local key-disclosure -> host-impersonation gap. New gamestream::{create_private_dir,
write_secret_file} + a 0600 regression test.
- #2 [HIGH] Native SPAKE2 PIN is single-use. The PIN is consumed the moment the host sends its
key-confirmation (which lets the client test its one guess), before reading the proof, so any
completed attempt -- right OR wrong -- disarms the window. A wrong PIN isn't observable
host-side (the client aborts before sending its proof), so consuming on first attempt is what
delivers the documented "one online guess" instead of an unbounded brute-force of the static
4-digit PIN. Test verifies single-use.
- #3 [MEDIUM] RTSP packetSize is bounded ([64,2048] in stream_config) and VideoPacketizer::new
uses saturating .max(1), killing a PRE-AUTH div-by-zero/underflow panic of the video thread.
Tests for {0,15,16,17} + out-of-range rejection.
fmt + clippy -D warnings clean; full workspace test suite green (93 host tests).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A pass over the apollo-comparison backlog (re-verified against current code).
Lands four items end-to-end plus a Windows-DualSense scoping doc.
- #5/#92/#26 — GameStream paired-cert allow-list. tls.rs surfaces the verified
peer cert to handlers (serve_https + PeerCertFingerprint, now shared with the
mgmt API instead of duplicated); nvhttp gates /launch /resume /applist /cancel
on AppState.paired and reports a real PairStatus; save_paired writes atomically
(temp+rename). Closes the "mTLS accepts any client cert" hole. + regression test.
- #6/#51/#19/#22 — NVENC caps query -> reference-frame invalidation. nvenc.rs
query_caps probes nvEncGetEncodeCaps (max dims / 10-bit / custom-VBV / RFI),
rejecting over-range modes and degrading 10-bit->8-bit instead of an opaque
InvalidParam. New Encoder::invalidate_ref_frames (default false -> caller
keyframes); the Windows NVENC path implements real RFI (multi-ref DPB +
nvEncInvalidateRefFrames, dedup + IDR-on-overflow). control.rs decodes the
0x0301 lost-frame range (Apollo's IDX_INVALIDATE_REF_FRAMES) -> AppState.rfi_range
-> encode loop, falling back to a keyframe. NOTE: the Windows NVENC impl is
RTX-box/CI-pending (can't compile on Linux); adversarially reviewed vs the SDK.
- #43/#72 — media socket QoS + buffer growth. New punktfunk_core::transport::qos:
grow_socket_buffers (factored out the native plane's 32MB SO_SNDBUF growth so the
GameStream sockets reuse it) + set_media_qos (opt-in PUNKTFUNK_DSCP=1: DSCP CS5
video / CS6 audio + Linux SO_PRIORITY, Apollo's scheme). Wired into UdpTransport
and the GameStream video/audio sockets. Windows IP_TOS needs qWAVE (follow-up).
- #8/#45 — GameStream input injection off the ENet service thread. on_receive no
longer injects inline (a slow inject head-blocked ENet keepalive/retransmit); it
forwards to a dedicated injector thread. The hardened InjectorService moved from
punktfunk1 into crate::inject (shared by both planes) + a coalesce step that sums
adjacent relative-mouse/scroll deltas while preserving button/key/abs ordering.
Docs: re-verified apollo-comparison.md status (22 items already done/obsolete since
the snapshot) + windows-dualsense-scoping.md (ViGEm can't emulate a DualSense; real
DS5 on Windows needs a VHF virtual-HID driver — web-research pass pending).
fmt + clippy -D warnings clean; full workspace test suite green; no C-ABI/OpenAPI drift.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The management console is a Nitro `node-server` build (per web/vite.config.ts) — it must be
run with `node`, not `bun`. Run under bun it 500s on every page render with
"Cannot find package 'srvx'": bun mis-resolves Nitro's externalized server deps from the
nested SSR chunk at request time. (This was pre-existing — the old manual pfweb.sh ran it
with bun too.)
- Provision `nodejs` in the pf2 distrobox; run the web service with `node .output/server/index.mjs`.
- Use `enable` + `restart` (not `enable --now`) so re-running the installer actually applies
unit-file changes instead of no-opping against the running service.
Verified on the Deck: web `/login` now returns 200 (was 500), "Listening on http://0.0.0.0:3000",
no srvx error.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
SteamOS is immutable read-only Arch, and the Deck is AMD (VAAPI) — so none of the
checked-in packaging (arch/sysext is NVENC-first + client-oriented, deb/rpm are
soname-mismatched) actually installs a working host on a Steam Deck. The proven path
(distrobox-built native binary + systemd-run units) was 100% manual. Make it one command.
- scripts/steamdeck/install.sh — idempotent installer: ensure the pf2 Debian-trixie
distrobox + toolchain → build host (+web console) → write config (generated web login
password) → raise UDP buffers to 32 MB + udev + input group (sudo, skipped gracefully
if unavailable) → install + start punktfunk-host / punktfunk-web systemd USER services
with linger. Flags: --open (accept unpaired clients), --no-web, --src=DIR. Builds
on-device so a rebuild always matches the running SteamOS (no prebuilt-binary fragility
across OS updates); VAAPI on the Deck's AMD GPU.
- scripts/steamdeck/update.sh — rebuild from current source + restart (config/pairings persist).
- scripts/steamdeck/README.md — deep reference (why on-device, what's installed, gotchas).
- docs-site: new "Steam Deck (Host)" guide + sidebar entry; install.md splits Arch from the
Steam Deck host path; packaging/arch/README points Deck-host users here and corrects the
stale "NVENC-only" note (VAAPI host encode landed).
Live-validated on the Deck: installer runs clean, both services come up, host listens
(QUIC :9777 + mgmt :47990), web serves (302→login); on a client connect it takes over the
Game-Mode gamescope session at the client's mode, captures via PipeWire, and VAAPI-encodes
(hevc_vaapi) — full pipeline confirmed in the host journal.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On a clean link the flat 20% FEC is pure waste: extra wire bytes AND extra
packets. On a packet-rate-bound uplink (the Steam Deck's WiFi tx caps ~22k pps
regardless of bitrate) those extra packets directly cost goodput — measured at
200 Mbps goodput, 20% FEC drove ~10% loss vs ~2.6% at 0% (it saturated the link).
Adaptive FEC closes the loop:
- Client measures the loss FEC is absorbing each ~750 ms window from session stats
(recovered shards / received, + a bump when a frame went unrecoverable) and sends
a periodic `LossReport { loss_ppm }` on the control stream (new message;
`window_loss_ppm` helper, shared + unit-tested). Connector (Apple/Linux/Windows)
and probe both report; suppressed during a speed test so its filler can't skew it.
- Host maps loss → recovery % (`adapt_fec`: ≈ loss×1.4 + 1pt, clamped 1..50) and
applies it live via `Session::set_fec_percent` (the wire is self-describing — each
packet carries its block's data/recovery counts, so the receiver needs no notice).
A clean link decays to ~1%; loss ramps it up and converges.
- `PUNKTFUNK_FEC_PCT`, when set, now PINS FEC static (disables adaptation) so
speed-test / measurement runs keep a fixed, known overhead. Unset ⇒ adaptive,
starting at 10%.
An older host ignores LossReport (unknown control message) and keeps static FEC;
an older client simply never reports and the host holds its start value. Builds +
clippy + fmt + tests green (adapt_fec / window_loss_ppm / loss_report unit tests).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
HDR (10-bit BT.2020 PQ) works end-to-end with the Windows host — it captures
an HDR desktop (WGC FP16 / Desktop-Duplication FP16 for the secure desktop)
and encodes HEVC Main10 to HDR-capable clients (Windows, Android). Only the
Linux host is blocked upstream (no 10-bit compositor capture). Corrected the
roadmap (grid + shipped/blocked), Windows Host page, status, and CLAUDE.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- install (host): add a Windows (NVIDIA) section with signed-installer and
certificate-trust steps; note the .cer is the same across releases.
- install-client: clarify the Windows MSIX certificate is the same every
release (trust once, updates need nothing).
- Move "Project & Internals" out of the public docs site: relocate
implementation-plan, apple-stage2-presenter, gamescope-multiuser,
dualsense-haptics, ci, and gamestream-host-plan to docs/; drop them from
the nav. Move windows-host into Host Setup.
- Rewrite roadmap as a lean public page with an at-a-glance grid and
current statuses (Windows host shipped/beta, Apple incl. tvOS shipped,
Android shipped, concurrent sessions + delegated pairing done).
- Fix status.md link to the now-internal implementation plan.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The punktfunk/1 speed test was unusable across every client/host: at the start of
a burst a little data got through, then everything read as dropped (~10 MB total).
Two compounding bugs:
1. Receive side measured throughput from fully-reassembled FLAG_PROBE *access
units* only. The instant loss crossed the 20% FEC budget no AU completed, so the
figure cliffed to 0 / 100% loss even though most bytes still arrived — a binary
cliff, not a graded measurement.
2. Send side blasted each filler AU (up to 256 KB ≈ 200 packets) into the socket
buffer in one unpaced batch, unlike the real video path which paces. On a small
buffer (e.g. the Steam Deck's 416 KB) a single AU overflowed it, so the test
measured self-inflicted buffer overflow instead of the link.
Fixes:
- Host `run_probe_burst` keeps each AU a small (~16 KB) burst and paces by the byte
budget, mirroring `paced_submit`; reports the WIRE packets the kernel accepted and
the ones the send buffer dropped (stat deltas), separating host-side drops from
link loss.
- `ProbeResult` gains `wire_packets_sent` + `send_dropped` (back-compat decode: a
21-byte pre-wire-stats result still decodes, new fields 0).
- Clients (probe + connector) count delivered traffic at the packet level via
`session.stats()` deltas over the burst window, so throughput/loss degrade
gracefully. Connector freezes the delivered figure when the host report lands so
resumed video can't inflate it. New `ProbeOutcome`/`PunktfunkProbeResult` fields:
`host_drop_pct`, `wire_packets_sent`, `send_dropped`.
Validated on loopback (graded 142→1391 Mbps, host_drop/link_loss split correctly,
no cliff) and live against the Deck: clean to ~200 Mbps goodput / 273 Mbps wire at
0% link loss, host send buffer the wall above that (the lever-#1 target).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add the Windows host (implemented & shipping: DXGI capture, SudoVDA
virtual display, NVENC, ViGEm, WASAPI, LocalSystem service installer;
NVIDIA-only, x64-only) — it was absent entirely.
- Add the Android client (full client: AMediaCodec/HDR10 decode, Oboe
audio + mic, gamepad feedback, discovery, pairing, Compose phone+TV UI;
Google Play internal testing) and drop the stale "scaffolds" item.
- macOS stage-2 presenter: built + live-validated behind the opt-in flag,
not "next".
- Concurrent sessions + delegated pairing approval marked done.
- Layout/CI: note Windows host backends and per-client release workflows.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Refresh the README and documentation for public visitors:
- README: public-facing rewrite with accurate status for all four native
clients (macOS, Linux, Windows, Android) and the Windows host.
- docs site: fix stale client status (Android is a full client, not a
scaffold; Windows client is stage-1 complete + signed MSIX), add the
missing Android client section, correct "which client" guidance.
- Windows host: corrected from "deferred/scoped" to implemented & shipping
(NVIDIA-only, x64-only) across windows-host, roadmap, status,
requirements, running-as-a-service, and the README.
- Remove internal infrastructure from public docs (box names, private IPs,
SSH/token commands, deploy topology); rewrite status.md as a public
project-status page; sanitize ci.md and implementation-plan.md.
- Update clients/android and clients/apple READMEs to current state.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Steam Deck (SteamOS) ships its OWN gaming session — `gamescope-session.target`
driven by `/usr/lib/steamos/gamescope-session`, not Bazzite's `gamescope-session-plus`.
That script `exec gamescope`s with HARDCODED physical-panel args (`-w 1280 -h 800 -O
'*',eDP-1`) and launches Steam via a SEPARATE `steam-launcher.service`, so the existing
managed-session path (which assumes session-plus) couldn't honor the client's mode — an
attach captured the panel's native 1280x800 instead.
Add a SteamOS branch to the managed-session path: detect it, write a `gamescope` PATH-shim
that rewrites the hardcoded args to `--backend headless -W <client> -H <client> -r <hz>`,
drop a transient user `gamescope-session.service.d` override pointing PATH at the shim +
the mode, then RESTART the whole target so `steam-launcher.service` brings Steam up IN the
headless gamescope at the client's resolution. Attach to the one fresh node (the restart
kills any prior gamescope, so no stale-node attach). Restore-on-disconnect removes the
override + restarts the target back to the physical panel (debounced; skipped if the user
switched to a desktop session). All user-level (`systemctl --user`) — no root.
Also widen `build_pipeline_with_retry` to 8 attempts (~90s): a host-managed gamescope
session cold-starting Steam Big Picture takes 30-60s to first frame, and a first-connect
timeout would tear down the warm session (forcing another cold start on reconnect).
Permanent failures still fail fast via `is_permanent_build_error`.
Validated live on a Steam Deck: Game Mode auto-detected, host takes over headless at the
client's mode (720p / 1080p), Steam Big Picture streamed glass-to-glass to the Mac at the
requested resolution. Single-tenant (concurrent clients at different modes still thrash —
a follow-up).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The shared unom CMS is now multi-tenant; the footer global became a per-tenant
collection. Query footers scoped to tenant.slug = punktfunk instead of the
removed /globals/footer.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
WiFi drivers (e.g. ath11k on the Steam Deck) return ENOBUFS — not
EAGAIN/EWOULDBLOCK — when the tx queue is momentarily full. Rust maps
ENOBUFS to ErrorKind::Uncategorized, so `is_transient_io` (which only
matched WouldBlock/ConnRefused/ConnReset) treated it as a real error and
tore the whole stream down on a single transient burst.
This presented as a vicious Heisenbug on the Deck: the native host
streamed flawlessly on loopback and under a debugger (anything slow
enough not to fill the small ~416 KB wlan0 buffer), but died at full rate
cross-machine over WiFi — flaky hang-or-SIGKILL because tx-queue-full is
probabilistic. Diagnosed live via a forced core dump (gdb on the hung
core): the data-plane thread had bailed on a fatal send error.
Treat ENOBUFS (and asynchronous network-path blips ENETUNREACH /
EHOSTUNREACH / ENETDOWN / EHOSTDOWN) as a lossy drop like WouldBlock —
FEC + the next frame recover. Validated: 6/6 back-to-back cross-machine
streams over the Deck's WiFi, host stable, p50 ~4.4 ms (one run dropped
4/300 frames *gracefully*, 0 mismatched — the fix working as intended).
Also surface a data-plane bind/hole-punch failure directly in punktfunk1
(it was previously only reported after teardown, which a stall could
swallow entirely).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Pull the same footer from the shared unom CMS global (cms.unom.io) and render
it globally under both the home and docs layouts. Read-only typed fetch in a
server-side root loader (falls back to null on a CMS hiccup). Root-relative
links target the marketing site, so they're resolved against its origin (the
docs don't host /legal/* etc.); themed with Fumadocs tokens for light/dark.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace the CSS-mask/webp wordmark with the inline vector from
Export/Punktfunk_Logo-Text_No-Border_Dark.svg (white export background
dropped), painted via currentColor — deep-violet on light, light-violet on
dark. Crisp at any size; drops the now-unused funk-wordmark.{webp,png}.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Swap the plain "punktfunk" text in the nav and landing hero for the real
brand wordmark from the marketing site. The source asset is a single
light-violet variant (made for dark surfaces), so it's painted as a CSS
mask and coloured per theme — deep-violet on light, light-violet on dark —
to stay legible with the docs' light/dark toggle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Theme the Fumadocs docs site with the punktfunk identity, mirroring the
marketing site:
- Swap the stock `neutral` preset for `purple`, then override --color-fd-*
with the violet lens-mark palette (#6c5bf3 / #a79ff8). The brand is the
violet, not the site's blue marketing background, so the blue is not used
as a reading surface; dark mode tints the chrome toward the app-icon
violet-dark (#1c1530).
- Adopt @unom/ui's token contract (--brand/--primary/--accent + bg-brand
etc.) as the shared token source, and @source its dist.
- Load Geist (the brand typeface) via @fontsource-variable/geist.
- Add the BrandMark lens to the nav + landing hero, wire the brand
favicon.svg, and add Docs/Website nav links.
- Keep the Fumadocs light/dark toggle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Phase 3 GPU-aware codec mask (6922e1c) probes VAAPI on any non-NVIDIA host.
On a GPU-less box (CI container: no /dev/nvidia* -> `auto` picks VAAPI, but there's
no VA display) the probe returns all-false, so the mask was 0 -- the host
advertised NO codecs, and the serverinfo unit test failed.
Fall back to the static superset when the probe yields nothing (VAAPI wasn't
usable, not "the GPU encodes nothing"); quiet ffmpeg's expected "No VA display"
error during the probe; and assert the test against codec_mode_support() rather
than a hardcoded 65793 so it's deterministic regardless of the build host's GPU.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Polish for AMD/Intel support:
- GameStream serverinfo advertises only codecs the GPU can ACTUALLY encode on
the VAAPI backend (probed once by opening a tiny encoder per codec). AV1
encode is narrow (Intel Arc/Xe2+, AMD RDNA3+/RDNA4) and an old iGPU may lack
HEVC, so a Moonlight client never negotiates a codec the encoder can't open.
NVENC/Windows keep the Moonlight-validated static mask. Validated on a Radeon
780M: h264/h265/av1 all probe true -> mask unchanged (65793).
- Packaging: Recommends mesa-va-drivers + intel-media-va-driver (deb) /
mesa-va-drivers + intel-media-driver (rpm) so the auto-selected VAAPI backend
works out of the box on AMD/Intel; NVIDIA boxes can --no-install-recommends.
(Fedora note: stock mesa-va-drivers disables HEVC/AV1 -- needs the freeworld
variant from RPM Fusion.)
- De-NVIDIA-fy the user-facing encoder log/context strings ("open NVENC" ->
"open video encoder") now that VAAPI is a first-class backend.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 2 of AMD/Intel support: the VAAPI encoder now takes the capture dmabuf
directly and does the RGB->NV12 colour conversion on the GPU's video engine,
eliminating the host-side de-pad + swscale CSC + upload the CPU path pays.
- capture: a vendor-neutral FramePayload::Dmabuf (dup'd fd + fourcc/modifier/
layout). When zero-copy is on, the EGL->CUDA importer is unavailable (any
non-NVIDIA host), and the backend is VAAPI, the capturer advertises LINEAR
dmabuf and hands the raw buffer to the encoder instead of CPU-copying it.
- encode/vaapi: the encoder self-configures from the first frame's payload (no
open_video signature change). The dmabuf arm wraps the buffer as an
AV_PIX_FMT_DRM_PRIME frame and pushes it through a filter graph
buffer(drm_prime) -> hwmap(vaapi) -> scale_vaapi=nv12 -> buffersink; the
encoder takes NV12 surfaces straight from the sink. The Phase 1 CPU-upload
path is kept as the other arm (used when capture produces CPU frames).
Live-validated on a Radeon 780M (real Sway/xdpw desktop capture): correct,
pixel-perfect HEVC, and ~10x less host CPU at 1440p (4.2s -> 0.4s of CPU for
300 frames) -- the de-pad/CSC/upload moves to the GPU. NVIDIA unchanged
(zero-copy still imports to CUDA; the passthrough path only engages on
non-NVIDIA hosts).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The CPU de-pad path trusted PipeWire's MAP_BUFFERS slice (`d.data()`, length =
`data.maxsize`). xdg-desktop-portal-wlr hands MemFd ScreenCast buffers whose
maxsize exceeds the bytes PipeWire actually maps into our process, so reading to
maxsize ran off the end of the mapping and SIGSEGV'd the capture thread —
crashing every CPU-path capture on Sway/wlroots (and thus any non-NVIDIA host,
which has no CUDA zero-copy importer and always falls back to this path).
mmap the fd ourselves, sized to its real length (fstat), for any fd-backed
buffer (MemFd SHM or DmaBuf); fall back to `d.data()` then drop. The existing
`needed > avail` guard now drops cleanly instead of over-reading. This also
subsumes the original "MAP_BUFFERS didn't map a Vulkan dmabuf" fallback.
Verified: fixes real Sway-desktop portal capture -> VAAPI HEVC on a Radeon 780M
(correct image + colours); the NVIDIA zero-copy path (returns before this code)
and the NVIDIA/KWin CPU path (self-mmap, fd_len == maxsize) both still work.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The host hard-linked libcuda.so.1 on Linux (`#[link(name="cuda")]` in
`zerocopy::cuda`), so the binary wouldn't even *start* on a non-NVIDIA box —
the dynamic loader can't resolve the NEEDED libcuda. That blocked running the
new VAAPI (AMD/Intel) path on a machine without the NVIDIA driver.
Resolve the 18 CUDA Driver API symbols at runtime via `libloading` instead.
Same-named wrapper fns forward to the dlopen'd table (call sites unchanged);
when libcuda is absent they return a non-zero CUresult so `context()` fails
cleanly and the capturer falls back to the CPU path. The library handle is
leaked (process-lifetime, like the shared context).
One Linux binary now runs on NVIDIA (CUDA zero-copy -> NVENC) and on AMD/Intel
(VAAPI, no NVIDIA driver). Verified: the NVIDIA dev box still does dmabuf->CUDA
zero-copy; on a Radeon 780M box the host builds with no libcuda present, the
binary has no NEEDED libcuda entry, and VAAPI encode runs with no stub.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Linux host was NVENC/CUDA-only. Add a VAAPI encoder — one libavcodec
backend (h264/hevc/av1_vaapi) covering both AMD (Mesa radeonsi) and Intel
(iHD) — behind the existing `Encoder` trait, and turn `open_video`'s Linux
arm into a vendor dispatcher: `PUNKTFUNK_ENCODER=auto|nvenc|vaapi` (default
auto: NVENC when a CUDA frame or /dev/nvidia* is present, else VAAPI). The
NVIDIA path is unchanged — auto resolves to NVENC on an NVIDIA box and the
bitrate-probe loop moved verbatim into `open_nvenc_probed`.
`VaapiEncoder` mirrors the NVENC hwframes pattern with AV_HWDEVICE_TYPE_VAAPI.
The CPU-input path swscales packed RGB -> NV12 (BT.709 limited, VUI signalled)
and uploads into a pooled VA surface (av_hwframe_transfer_data), preserving the
low-latency model (infinite GOP, on-demand forced IDR, async_depth=1, CBR when
the driver supports it). It works on a non-NVIDIA box with no capture changes:
the capturer already falls back to CPU frames when its EGL->CUDA importer can't
initialise (no libcuda).
Live-validated on a Radeon 780M (RDNA3): hevc/h264/av1_vaapi all encode,
HEVC/H264 decode cleanly with correct BT.709-limited colours, infinite GOP
preserved. Zero-copy dmabuf import (the high-res perf lever) is next.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The host is NVIDIA/NVENC + SudoVDA coupled; Windows ARM64 has neither an NVIDIA
driver nor an ARM64 SudoVDA, so an ARM64 host would install but couldn't encode
or make a virtual display. Document the deliberate x64-only scope so it doesn't
get re-litigated. ARM64 stays client-only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of the persistent ISCC "path not found": ISCC.exe is 32-bit, and the
self-hosted runner runs as SYSTEM, so the checkout lives under
C:\Windows\System32\config\systemprofile\.cache\... . WOW64 file-system
redirection rewrites a 32-bit process's System32 reads to SysWOW64 (where nothing
exists), so ISCC died opening the .iss before it even printed its version line.
(The smoke-test diagnostic compiled fine precisely because it lived at C:\t\out.)
Fix: copy every file ISCC reads (the .iss + host.env.example + README.md) into
the non-redirected build dir C:\t\out and compile from there; BinDir, StageDir,
and OutputDir already live under C:\t. Removed the now-spent smoke diagnostic.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The smoke-test diagnostic proved Inno itself is healthy (a trivial ASCII script
compiled), while the real .iss died before the "Compiler engine version" line —
i.e. at script open, not during compile. The difference: the real .iss was UTF-8
with non-ASCII chars (→, —) in comments, which ISCC 6.4+ rejects without a UTF-8
BOM (and the German-locale runner misreads). Replace them with ASCII (->, -).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
All [Files] sources are validated-present yet ISCC still errors before any
"Compiling" output (no line number) — so it's startup/[Setup]-internal, not a
source path. Add an explicit [Languages] (compiler:Default.isl) to rule out the
auto-added default language, and on ISCC failure dump the Inno install dir +
run a trivial [Setup]-only smoke script to tell "Inno broken" from "my script".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The {#SourcePath} relative-traversal for host.env.example/README kept tripping
ISCC ("path not found", error 2) regardless of the separator, so drop it: compute
the two paths absolutely in pack-host-installer.ps1, Test-Path them (clear PS error
if missing), and pass /DHostEnv + /DReadme. The .iss [Files] now reference the
absolute defines — no {#SourcePath}, no ..\.. traversal. Also prints "source ok"
for each so a future failure is unambiguous.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
windows.yml + windows-msix.yml gain an x86_64/aarch64 target matrix. ARM64 is
cross-compiled on the one x64 Windows runner — the x64 MSVC toolset ships the
ARM64 cross compiler, aarch64-pc-windows-msvc is tier-2 with host tools, and
SDL3/libopus (build-from-source) cross-compile cleanly. The only arch-specific
external dep is FFmpeg's import libs: the matrix points FFMPEG_DIR at a per-arch
tree (x64 C:\Users\Public\ffmpeg, arm64 C:\Users\Public\ffmpeg-arm64, both
FFmpeg 7.x / avcodec-61). Per-arch short CARGO_TARGET_DIR avoids a shared target
dir; fmt + test run only for x64 (aarch64 can't execute on the x64 host).
pack-msix.ps1 gains -Arch x64|arm64 (stamps the manifest ProcessorArchitecture,
arch-suffixes the .msix/.cer); windows-msix.yml matrixes both arches and
publishes ..._x64.msix / ..._arm64.msix. setup-windows-runner.ps1 provisions the
rustup target + the ARM64 FFmpeg tree (idempotent).
Verified live on the runner (home-windows-1): debug+release cross-build green,
clippy -D warnings green, and MSIX pack produces a valid arm64 package (manifest
arch=arm64; bundled exe/SDL3/avcodec/reactor-bootstrap all PE machine 0xAA64).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
On the Windows WGC HDR path the FP16 scRGB capture was fed to NVENC as
R10G10B10A2 (BT.2020 PQ), and NVENC did the RGB→YUV CSC internally on the
contended SM — adding to the encode_ms wall under a GPU-saturating game.
(NVIDIA's D3D11 VideoProcessor can't do RGB→P010 for HDR; that path renders
green, confirmed live — so the convert must be ours.)
New `HdrP010Converter` fuses the tone-map with the BT.2020 RGB→YUV matrix and
emits P010 (10-bit limited range) directly: a luma pass → an R16_UNORM plane
RTV (full-res) and a chroma pass → an R16G16_UNORM plane RTV (half-res, 2x2
box average) of a DXGI_FORMAT_P010 texture. NVENC then takes native P010 and
skips its SM-side convert.
Gated behind PUNKTFUNK_HDR_SHADER_P010 (default OFF → the existing
R10→NVENC path is byte-for-byte unchanged). Colour validated by a new
`hdr-p010-selftest` subcommand: a synthetic scRGB pattern → P010 → readback,
compared to a BT.2020 PQ 10-bit reference — max abs error Y=0.99 / Cb=0.82 /
Cr=0.75 codes on an RTX 4090. Live-validated HDR colours correct (no green).
Build + clippy (--features nvenc -D warnings) green on x86_64-pc-windows-msvc.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Gradle's Exec resolves command[0] via the JVM/daemon's inherited PATH, not
the environment("PATH", …) set on the task (that only reaches the spawned
child). A GUI Android Studio launch — and any daemon it starts — has no
~/.cargo/bin on its PATH, so a bare "cargo" fails with "A problem occurred
starting process 'command 'cargo''". Use the already-computed cargoBin
absolute path; the env PATH still lets cargo/cargo-ndk find their subtools.
Also refresh the README prereqs: add the missing cmake;3.22.1 SDK package
(the cmake crate builds libopus with it) and drop the broken
`brew --prefix openjdk@21` JAVA_HOME hint in favour of `java_home -v 21`.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Fedora RPM build linked punktfunk-host against a synthesized libcuda stub
with a FROZEN symbol list baked into ci/fedora-rpm.Dockerfile. The priority-
stream work added cuCtxGetStreamPriorityRange / cuStreamCreateWithPriority /
cuStreamSynchronize / cuMemcpy2DAsync_v2, which weren't in that list, so the
link failed with "undefined symbol".
build-rpm.sh now regenerates /usr/lib64/libcuda.so.1 from every cu* symbol the
host source references (grep of crates/punktfunk-host/src), before rpmbuild — so
a new cu* call can never silently break the link again. Self-maintaining and
needs no builder-image rebuild (it supersedes the Dockerfile's frozen stub).
Verified the 23 extracted symbols compile and cover the 4 that were undefined.
Also fix the bogus %changelog weekday (Sun -> Mon, Jun 15 2026 is a Monday) that
rpmbuild warned on.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ExposedDropdownMenuBox anchors on a read-only OutlinedTextField, and a text field
captures D-pad focus -- directional keys never escape it, so on a TV/controller you
got stuck on the first select. Replace SettingDropdown with a clickable Surface +
DropdownMenu (no text field): D-pad moves between settings, A opens the menu, A
selects an item. Adds a primary-colour focus border so the focused setting reads
across a room.
Verified locally: ./gradlew :app:assembleDebug BUILD SUCCESSFUL.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The hourly docker-prune could never reclaim the real disk filler: the act_runner
cache server's blob store (cache.dir:"" -> /root/.cache/actcache/cache) lives in
the long-running runner container's WRITABLE LAYER, which docker prune can't see.
It grew to ~66 GB and filled the 125 GB disk on its own.
- New docker-prune.sh holds the logic (inline ExecStart= broke under systemd's
own $-expansion, which emptied $SZ/$(...) before sh ran them — silently no-oping
the burst guard). The unit now just calls the script.
- Caps the actcache: clears the blobs once they exceed ~20 GB (act_runner
repopulates; keys are content-hashed, so only stale entries drop).
- Burst guard lowered 85%->80% and now also clears the actcache.
- Timer hourly -> every 30 min; image/cache `until` 12h -> 6h.
Live: cleared 66 GB on home-runner-1 (93% -> 20%), deployed + verified.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ISPP's {#SourcePath} has no trailing backslash, so {#SourcePath}..\..\scripts
resolved to ...\packaging\windows..\..\scripts (invalid component "windows..")
-> ISCC error 2 "path not found". Add the explicit separator (a double backslash
is harmless on Windows if a future ISPP ever adds the trailing one).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The first CI run failed only on the SudoVDA download: SudoMaker/SudoVDA has no
releases (source-only repo; Apollo embeds the driver in its installer), so there
was nothing to fetch. Vendor the prebuilt SIGNED driver in-repo instead.
- packaging/windows/sudovda/: SudoVDA.inf/.cat/.dll + sudovda.cer (derived from
the .cat signer CN=sudovda@su.mk), pulled from the dev-box driver store.
v1.10.9.289, Class=Display, HWID Root\SudoMaker\SudoVDA, MIT/CC0.
- fetch-sudovda.ps1 -> stage-sudovda.ps1: stage the vendored driver + fetch
nefcon from its real pinned release (v1.17.40, sha256 812bae7e…, x64/nefconc.exe).
- pack-host-installer.ps1: call stage-sudovda.ps1; README updated with the
driver-refresh recipe.
The rest of the pipeline already passed on the first run (host built --features
nvenc via the llvm-dlltool import lib; ISCC + signtool found; signed with the
real CN=unom cert).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Linux zero-copy tiled-GL path can now produce NV12 (BT.709 limited range)
on the GPU and feed NVENC native YUV, deleting NVENC's internal RGB->YUV CSC —
which runs on the SM/3D-compute engine a saturating game pins at 100% (the
game-vs-encode contention headache). Windows already does this via the D3D11
video processor; this closes the Linux gap. See docs/host-latency-plan.md §2A.
Gated behind PUNKTFUNK_NV12 (default OFF → the RGB/BGRx path is byte-for-byte
unchanged; zero regression). Only the tiled EGL/GL path converts; the
LINEAR/Vulkan-bridge (gamescope) path stays RGB.
- zerocopy/egl.rs: Nv12Blit — BT.709 limited Y pass (R8, full-res) + UV pass
(RG8, half-res, GL_LINEAR 2x2 average); both CUDA-registered; import_nv12.
- zerocopy/cuda.rs: two-plane DeviceBuffer (Y W*H@1B + interleaved UV
(W/2)*2 x H/2), paired Y+UV pool, copy_mapped_nv12 + copy_nv12_to_device,
on the per-thread priority stream (dmabuf-recycle sync preserved).
- encode/linux.rs: nvenc_input(Nv12)->NV12; submit_cuda copies two planes into
NVENC's surface; VUI signalled BT.709 limited (colorspace/range/primaries/trc).
- capture/linux.rs: gate (PUNKTFUNK_NV12 && tiled), report format Nv12.
- main.rs + zerocopy/mod.rs: `nv12-selftest` subcommand.
Validated on RTX 5070 Ti two ways: (1) nv12-selftest — synthetic RGBA->NV12
round-trip vs a BT.709 reference, max abs error Y=0.56/U=0.33/V=0.26 LSB;
(2) live capture->NV12->NVENC->decode of animated red content matches the RGB
path's colour (avg RGB 230,18,18 vs 231,18,20). build/clippy/fmt green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The cfg(windows) code can't be lint-checked on the Linux dev box, so three
-D warnings slipped through (caught by windows.yml; the FFI + shaders compiled
fine):
- gpu.rs: SetMultithreadProtected returns a must-use BOOL -> `let _ =`.
- video.rs: drop the unused GpuFrame::ten_bit field (present keys off `hdr`;
the value is still computed locally for the first-frame log).
- present.rs: GpuView::frame is an RAII keep-alive (its Drop returns the decoder
surface to the pool), never read -> #[allow(dead_code)].
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The client was pure software HEVC decode + CPU swscale->RGBA + a full-frame
dynamic-texture upload every frame -- the reason performance was poor on a GPU
box (the GPU sat idle while the CPU churned). This adds a hardware path, HDR,
and a GUI pass.
Performance -- D3D11VA zero-copy:
- gpu.rs (new): one D3D11 device (hardware + VIDEO_SUPPORT, WARP fallback,
multithread-protected) shared by decoder and presenter via a Send/Sync
OnceLock. Sharing is mandatory -- a decoded texture is only bindable on the
device that created it. windows-rs COM interfaces are !Send/!Sync, so the
unsafe impl is sound only under the multithread protection + disjoint
decode(video ctx)/present(immediate ctx) split.
- video.rs: D3d11vaDecoder (raw FFI mirroring the Linux VAAPI module). The
COM-typed AVD3D11VA{Device,Frames}Context are declared here (stable FFmpeg
ABI) to avoid ffmpeg-sys binding the d3d11 headers; get_format builds a frames
ctx with BindFlags=SHADER_RESOURCE so the NV12/P010 array slices are
sampleable. av_frame_clone guard keeps each surface out of the reuse pool
until the presenter drops it. Software decode stays as the fallback
(DecoderPref Auto/Hardware/Software; auto falls back on init/decode error).
- present.rs: shared device; per-plane SRVs over the array slice
(NV12->R8/R8G8, P010->R16/R16G16) + three pixel shaders (RGBA passthrough,
NV12/BT.709, P010/BT.2020-PQ). present() now takes the frame by value so the
GPU surface survives re-presents.
HDR:
- Detected in-band (transfer == SMPTE2084), same signal as the other clients.
Swapchain flips to R10G10B10A2 + ST.2084 + HDR10 metadata. New Settings toggle
gates advertising VIDEO_CAP_10BIT|HDR; host still gates 10-bit behind its own
PUNKTFUNK_10BIT + actual-HDR-content checks.
GUI (windows-reactor):
- Host cards with accent-monogram avatars + colored status pills, InfoBar for
errors/pairing hints, ToggleSwitch settings (+ HDR, decoder, bitrate), button
icons, a richer connecting screen, and a stream HUD with GPU/CPU-decode + HDR
status chips.
Not yet on-glass validated: the Linux dev box can't compile the cfg(windows)
code (ffmpeg/windows crates unfetched; WARP has no hw decode) -- only
cargo fmt checks it here. API shapes verified against the windows-rs/reactor
source and the YUV->RGB coefficients checked by hand, but D3D11VA + shaders +
the GUI need a real build (Windows CI / build VM) and on-glass test on the RTX
box. The host-side HDR encode path is unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Make a controller drive the Compose UI when not streaming, so the menus work on a TV
remote AND on a controller paired to a phone:
- MainActivity maps gamepad face buttons to the keys Compose's focus system
understands (A -> DPAD_CENTER to activate, B -> BACK); D-pad *keys* already move
focus and pass through untouched.
- For controllers whose D-pad reports as HAT axes (or to navigate with the left
stick), dispatchGenericMotionEvent converts AXIS_HAT_X/Y / AXIS_X/Y into discrete
D-pad key events, edge-detected so a held direction moves focus exactly once.
- HostCard draws a clear primary-colour focus border (the default state layer is too
subtle across a room on TV).
All gated on "not streaming" -- during a stream the controller still forwards to the
host unchanged. Compile-verified (./gradlew :app:assembleDebug); the focus behaviour
itself needs on-device validation (no KVM here for a TV emulator).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Verified, prioritized analysis in docs/host-latency-plan.md (multi-agent
investigation + adversarial verification). Lands the two low-risk tiers:
Tier 2B — Linux scheduling hygiene:
- boost_thread_priority now nices the capture/encode (-10) and send (-5)
threads on Linux (setpriority, best-effort; no-op without CAP_SYS_NICE),
and the wrong "gamescope caps the game" doc-comment is corrected.
- CUDA context created with CU_CTX_SCHED_BLOCKING_SYNC (frees a core on the
shared box instead of busy-spinning on completion).
- Copies moved off the default stream onto a per-thread highest-priority
CUDA stream (cuStreamCreateWithPriority, graceful NULL-stream fallback)
with a per-stream sync that no longer blocks on the other worker thread's
in-flight copies. Stream priority is measure-then-keep (NVIDIA Linux may
ignore it); never regresses.
Tier 3A — Windows session tuning (new session_tuning.rs, raw C-ABI FFI,
no-op off Windows): once-per-process 1ms timer + DwmEnableMMCSS + HIGH
priority class; per-thread MMCSS "Games" + keep-display-awake. Wired into
both the native (boost_thread_priority) and GameStream (stream.rs) paths.
We had zero session tuning before (Apollo streaming_will_start parity).
Tier 2A (Linux NV12 convert) is specified but intentionally not landed:
it is colour-correctness-critical and needs A/B validation on a GPU box
with a display (green-screen risk). Builds + clippy + fmt green on Linux.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
MSIX (the client's format) can't install the host's LocalSystem secure-desktop
service or the SudoVDA kernel driver, so the host ships as a signed Inno Setup
setup.exe that runs elevated and delegates to the existing idempotent
`punktfunk-host service install`.
- packaging/windows/punktfunk-host.iss: lay exe into Program Files, optional
SudoVDA driver task, run service install/start; [Code] stops+waits the service
before file copy on upgrade; uninstall runs service uninstall.
- pack-host-installer.ps1: cert (reuses MSIX_CERT_PFX_B64 / self-signed CN=unom),
sign inner exe + setup.exe, fetch/stage SudoVDA, run ISCC, export public .cer.
- fetch-sudovda.ps1 / install-sudovda.ps1: pinned SudoVDA + nefcon download, cert
import, gated device-node create (no phantom dup), pnputil install (warn-not-abort).
- nvenc/: synthesize nvencodeapi.lib via llvm-dlltool from a 2-export .def so
--features nvenc links with no GPU/SDK at build time.
- .gitea/workflows/windows-host.yml: build (nvenc) -> clippy -> ISCC -> sign ->
publish setup.exe + .cer to the generic registry pkg punktfunk-host-windows.
Tag host-win-v* -> X.Y.Z (+ latest/ alias); main push -> rolling 0.2.<run>.
- setup-windows-runner.ps1: provision Inno Setup; docs: installer instructions.
SudoVDA/nefcon release URLs+SHA-256s in fetch-sudovda.ps1 are placeholders
(baseline v0.2.1) — fetch warns + prints the computed hash until pinned.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Decky Loader is a PyInstaller binary; it puts its bundled (older) libssl/libcrypto
on LD_LIBRARY_PATH via its /tmp/_MEI* unpack dir, and that env leaked into the
backend's `flatpak run`/`flatpak kill` subprocess. The SYSTEM flatpak's libcurl
+ libostree need newer OPENSSL symbols (3.2/3.3/3.4), so pairing failed with
"libssl.so.3: version OPENSSL_3.3.0 not found". _flatpak_env() now restores
each LD_*_ORIG PyInstaller saved, or drops the var, so the system loader uses
system libs. Reproduced + verified on the Deck (SteamOS 3.8.10, Flatpak 1.16.6).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Settings: flat list -> Display / Host / Audio / Overlay sections in outlined
cards (SettingsGroup + ToggleRow helpers) with section headers.
- ConnectScreen: connection errors now show in a filled errorContainer banner
(was plain red text lost in the layout), and a "Searching the local network..."
spinner appears while discovery is active but nothing's turned up yet.
Verified locally: ./gradlew :app:assembleDebug BUILD SUCCESSFUL.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Crash: DisposableEffect.onDispose called nativeClose(handle) (Box::from_raw frees
the SessionHandle) while the SurfaceView's surfaceDestroyed independently called
nativeStopVideo/Audio/Mic on the same handle -- whichever ran after the close
dereferenced freed memory (SIGSEGV: the consistent back-navigation crash). Add a
one-shot `closed` guard: onDispose marks it before freeing; surfaceDestroyed skips
the native calls once closed (backgrounding still stops the threads when it wins).
Polish:
- Branded Material You theme (Theme.kt): dynamic colour on Android 12+, punktfunk
brand violets as the pre-12 fallback, replacing the generic darkColorScheme().
- ConnectScreen: "Connecting..." was rendered in error-red with no spinner; now a
neutral spinner while connecting, red reserved for actual errors.
Verified locally: ./gradlew :app:assembleDebug BUILD SUCCESSFUL (both ABIs + the
Compose changes), debug APK assembles.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ndk's DataSpace derives Copy/PartialEq/Eq and impls Display (no Debug), so the
{ds:?} in the HDR dataspace log statements wouldn't compile under cargo-ndk.
Host clippy can't catch it — decode.rs is android-gated. Switch to {ds}.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Install failed with "GPG verification enabled, but no signatures found" on the
commit: the deploy step only ran build-update-repo (signs the summary). Add
`flatpak build-sign` to sign the commit objects too — clients with
gpg-verify=true verify the commit, so summary-only signing isn't enough.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mirrors the Apple client's HDR path so the Android client can display HDR from a
Windows HDR host:
- nativeConnect now advertises VIDEO_CAP_10BIT | VIDEO_CAP_HDR (was 0), so the
host upgrades to a Main10 / BT.2020 PQ encode.
- decode.rs detects HDR reactively from the decoder's reported output colour
(color-transfer ST2084=6 / HLG=7, color-range) -- the AMediaCodec analogue of
VideoToolbox's format description on Apple -- and signals the Surface dataspace
(Bt2020[Itu]Pq / Bt2020[Itu]Hlg) so the compositor/display switch to HDR.
AMediaCodec decodes Main10 from the in-band SPS; no profile override needed.
Also fixes the Android build: set_frame_rate (added in 5262e28) is gated on the
ndk `nativewindow` + `api-level-30` features, which weren't enabled -- so that
commit could not compile under cargo-ndk. Enable
features = ["media","audio","nativewindow","api-level-31"] (minSdk 31): covers
set_frame_rate (api-30), set_buffers_data_space + the DataSpace module (api-28),
and ANativeWindow (nativewindow).
Verified host-side: fmt --all + clippy --workspace (the caps advertise + JNI
surface). The android-gated decode + NDK gating verified against the ndk 0.9
sources; android.yml (cargo-ndk) is the compile gate, and real HDR display needs
an HDR device + Windows HDR host.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Apple is TestFlight-only (no App Store) — link the join URL; drop the App Store
placeholder. Add the live Google Play listing for io.unom.punktfunk.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per-device install steps in one place: Linux (Flatpak via flatpak.unom.io +
native apt/rpm/Arch), Steam Deck, Windows (signed MSIX from the registry),
macOS (notarized DMG from releases), and iOS/Android (store/beta links). Adds
it to the Connecting nav and cross-links clients.md, whose Linux/Flatpak bullet
now points at the hosted flatpak.unom.io repo instead of the bundle README.
Mobile store/TestFlight URLs are placeholders pending the public listings.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Stats HUD (mirrors the Apple client): the decode thread accumulates FPS, receive
throughput, and capture->client latency (p50/p95, skew-corrected) in Rust
(clients/android/native/src/stats.rs); nativeVideoStats drains a snapshot ~1 Hz
over JNI as a DoubleArray. StreamScreen renders a Compose overlay
(W*H@Hz / fps / Mb/s / latency, + dropped-under-loss), toggled by a Settings
switch (persisted, default on) or a 3-finger tap.
Performance (decode.rs):
- ANativeWindow_setFrameRate(refresh_hz): align display vsync to the stream rate
(no 60-in-120 judder); safe since minSdk 31 >= API 30.
- Raise the decode thread toward URGENT_DISPLAY (best-effort setpriority) so
background work can't preempt it under load.
- Codec low-latency hints KEY_PRIORITY=0 (realtime) + KEY_OPERATING_RATE.
Verified host-side: cargo build/clippy/fmt --workspace (the ungated stats + JNI
accessor). The android-gated decode.rs (NDK) and the Kotlin build only in CI
(android.yml: gradle + cargo-ndk) -- APIs verified against crate sources.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The CI added --default-branch=stable, so the repo ref is
app/io.unom.Punktfunk/x86_64/stable. build-bundle defaults to `master` when no
branch is given → "Refspec app/io.unom.Punktfunk/x86_64/master not found". Pass
`stable` explicitly in both flatpak.yml and the local build-flatpak.sh.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
stoppedHandler/resetHandler are non-optional closures on the CI SDK
((StoppedReason)->() and ()->()), so assigning nil fails to compile
(apple.yml). Assign no-op closures to disarm them before engine.stop()
-- same re-entrancy guard intent, type-correct.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The CI only shipped a single-file .flatpak bundle, which has no remote — users
couldn't `flatpak update`. Keep the bundle (Decky fallback) but also sign the
OSTree repo flatpak-builder already produces and publish it to a shared,
reusable unom-wide remote.
- flatpak.yml: pin --default-branch=stable; import the signing key and
build-update-repo --gpg-sign; generate unom.flatpakrepo + the app .flatpakref
+ index.html; rsync the repo to unom-1 and bring up a static Caddy container.
The step no-ops until FLATPAK_GPG_PRIVATE_KEY/DEPLOY_* exist (build stays green).
- packaging/flatpak/server/: compose.production.yml + Caddyfile (static file
server on :3230, mirrors docker.yml deploy-docs).
- unom-flatpak.gpg: committed public signing key (base64 -> GPGKey= in the descriptors).
- README: hosted repo is now the recommended install; documents the one-time
infra (edge Caddy vhost, infra port 3230, DNS, the GPG secret).
Edge Caddy vhost + infra port allowlist + the secret are applied out-of-band.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two bodies of work in one commit (the rename moved files the fixes also touched).
Naming/structure cleanup (pre-launch):
- Host modules m3.rs->punktfunk1.rs, m0.rs->spike.rs; CLI m3-host->punktfunk1-host,
m0->spike; bare `punktfunk-host` now prints help. Types M3Options/M3Source->
Punktfunk1Options/Punktfunk1Source.
- Clients consolidated out of crates/ into clients/: punktfunk-client-rs->
clients/probe (crate punktfunk-probe), client-linux->clients/linux,
client-windows->clients/windows, punktfunk-android->clients/android/native
(crate punktfunk-client-android; kept [lib] name=punktfunk_android so the JNI
contract is unchanged). crates/ now holds only core + host.
- Milestone codes M0-M4 purged from code/CLI/CLAUDE.md/README/docs/docs-site,
kept only in docs/implementation-plan.md. docs/m2-plan.md->
docs/gamestream-host-plan.md. CI/gradle/flatpak paths updated.
Client loss-recovery (video froze and never recovered after a brief drop):
- Export punktfunk_connection_frames_dropped through the C ABI (the core already
tracked it for the client keyframe-recovery loop; it was never reachable from
the ABI clients). Regenerated punktfunk_core.h.
- Apple (StreamPump + Stage2Pipeline) and Android (decode.rs) now poll
frames_dropped and request a keyframe when it climbs -- the same loss-driven
recovery Linux/Windows already had. Under infinite GOP the decoder silently
conceals reference-missing frames, so the decode-error trigger rarely fires.
Apple rumble robustness (worked then went spotty -- DualSense + Xbox):
- Add CHHapticEngine stopped/reset handlers (rebuild on app background / audio
interruption / server reset) and drop the permanent `broken` latch on a
transient drive failure; latch only when the controller truly has no haptics.
- Surface swallowed SDL set_rumble errors on Linux/Windows + diagnostic logging.
Verified: cargo build/clippy/fmt --workspace, C-ABI harness, header drift.
Not runnable on this box (verify in CI): Gitea workflows, gradle/Android,
flatpak, Swift/decky.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
r0adkll/upload-google-play hides real API errors behind "Unknown error
occurred." Proved the full upload sequence (insert edit -> upload bundle ->
track update -> validate) succeeds with the service account, so the failure was
r0adkll's opaque error handling and/or a base64-encoded SERVICE_ACCOUNT_JSON
secret.
clients/android/ci/play-upload.py does the same sequence with stdlib + openssl
(no pip), reuses the SERVICE_ACCOUNT_JSON secret, tolerates it being raw JSON or
base64, auto-retries commit with changesNotSentForReview, and prints Google's
actual error. Locally dry-run-validated against the live app (both secret forms).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Build a universal release APK alongside the AAB and push both to the public
generic registry (punktfunk-android/<run_number>/) before the Play upload, so
artifacts are downloadable even while the Play step is still failing. Matches
windows-msix.yml / deb.yml (REGISTRY_TOKEN, user enricobuehler).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
So Windows mic passthrough works without the user installing anything: when no virtual-mic
device is present, install Steam Remote Play's SteamStreamingMicrophone.inf (ships under
Steam\drivers\Windows10\{arch}\ next to the speakers INF Apollo uses) via DiInstallDriverW
loaded from newdev.dll — the same mechanism Apollo uses for Steam Streaming Speakers — then
re-find the device. Needs admin (the host runs as SYSTEM); best-effort and safe (no-op if
Steam absent / INF not found / PUNKTFUNK_NO_MIC_INSTALL), falling back to the manual-install
guidance (VB-Audio Cable) otherwise.
Not yet built/validated on the box (down); FFI cross-checked against windows-0.62. Whether
Steam ships SteamStreamingMicrophone.inf at that path is to be confirmed on the box.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The host received the client's mic uplink (0xCB Opus) but dropped it on Windows ("requires
Linux"). Windows has no user-mode way to CREATE a capture endpoint, so target an existing
virtual audio device and write the decoded mic PCM into its RENDER endpoint — the device's
CAPTURE endpoint then surfaces as a microphone host apps record from (the inverse of a
virtual cable). New audio::wasapi_mic::WasapiVirtualMic: finds the device by friendly-name
(Steam Streaming Microphone / VB-Audio CABLE Input / VoiceMeeter / "virtual", override with
PUNKTFUNK_MIC_DEVICE), opens a WASAPI shared event-driven RENDER client (48 kHz stereo f32,
autoconvert), and a dedicated COM thread writes a bounded (~80 ms drop-oldest) inject queue
with silence-fill. open_virtual_mic() gets a Windows arm; mic_service_thread (Opus decode →
push) now compiles for windows too (opus is already a windows dep). Clear error + install
guidance when no virtual device is present.
Linux/cross-platform side cargo-checks; the Windows path is built/validated when the box is
back (the wasapi render API was cross-checked against the docs + the existing capture path).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Apollo runs its capture thread at CRITICAL and its encoder thread at ABOVE_NORMAL; we set
none. Our GPU work is already HIGH priority, but the GPU scheduler can only favour commands
we've SUBMITTED — a normal-priority thread descheduled by a CPU-heavy game submits the
convert/encode late, so the HIGH GPU priority never bites (consistent with the measured
"NVENC engine idle yet the encode waits ~15 ms"). Raise the WGC helper's capture+encode
loop and the single-process capture+encode loop to THREAD_PRIORITY_HIGHEST, and the
transmit thread to ABOVE_NORMAL, via a cross-platform boost_thread_priority() (Windows-only
effect — the Linux host caps the game via gamescope so its threads aren't starved).
Not yet built/validated on the GPU box (it's down); the cross-platform side compiles
(cargo check) and the Windows calls are cross-checked against the windows-0.62 API.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When a client requests a bitrate above the GPU's HEVC/AV1 level ceiling, NVENC rejects
initialize_encoder. The old probe stepped the rate down by ×¾ each retry, undershooting
the real ceiling badly (a 1 Gbps request landed ~300 Mbps even with the level cap near
800). Replace it with a binary search over [floor, requested] that converges (±20 Mbps)
on the HIGHEST rate NVENC accepts and clamps to that — so the stream uses the full
codec-level bitrate. Factored the session open/config/init into try_open_session() for
the probe; split-encode rejection is disambiguated from a bitrate-cap rejection (retry
once with split disabled) and the floor fallback also tries split-disabled.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
NVENC defaulted to Main tier, whose per-level bitrate ceiling at 5K (HEVC Level 6.2
Main ≈ 240 Mbps) made initialize_encoder reject a high client bitrate; the existing
probe-and-step-down then silently dropped a ~1 Gbps request by ×¾ to ~240-320 Mbps —
visible color/motion compression on fast scenes. Set HIGH tier (≈800 Mbps for HEVC,
higher for AV1) + autoselect level so the requested bitrate goes through. `tier`/`level`
are u32 (HIGH=1, AUTOSELECT=0) shared across the HEVC/AV1 union offset; the step-down
remains as a safety net. Not yet built/validated on-box (box offline).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Windows host capped at ~60 fps with 35-40 ms latency on a GPU-heavy game:
the per-frame capture→encode path shared the 3D engine with the game and got
scheduled behind it. Rework to minimize 3D-engine work per frame:
- VideoConverter (D3D11 video processor): capture → NVENC-native NV12/P010 so
NVENC skips its internal RGB→YUV (a 3D/compute step). Wired into both DDA
(dxgi.rs) and WGC (wgc.rs). New PixelFormat::Nv12/P010 + NVENC YUV input.
- GPU scheduling hardening (Apollo-style): D3DKMTSetProcessSchedulingPriorityClass
HIGH, absolute SetGPUThreadPriority, SetMaximumFrameLatency(1).
- WGC SDR zero-copy (hold pool frames; no CopyResource). DDA keeps a fast
CopyResource to decouple its single-frame acquire/release from the async convert.
- Pipelined helper encode loop (PUNKTFUNK_ENCODE_DEPTH, default 1) + perf split
(cap_wait / encode / write).
Live on the RTX 4090: hard 60 fps ceiling removed (now scene-scaling 40-200+),
latency much reduced. Residual cap in GPU-pinned scenes is the irreducible RGB→YUV
convert (no fixed-function unit on NVIDIA — VideoProcessing engine reads 0%) waiting
behind an uncapped game under WDDM context time-slicing; Linux avoids it via
gamescope capping the game to the display refresh.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The plugin was a QAM launcher whose stream never appeared, with no
pairing. Three fixes, plus a headless --pair mode on the GTK client:
- Stream actually starts (MoonDeck's proven mechanism): gamescope only
focuses the process tree Steam launched via reaper, so a flatpak
spawned from the (root) backend is invisible. The frontend now
registers ONE hidden non-Steam shortcut pointing at bin/punktfunkrun.sh,
passes the host as the shortcut's Steam launch options, and starts it
with SteamClient.Apps.RunGame — gamescope then fullscreen-focuses it.
The wrapper execs `flatpak run io.unom.Punktfunk --connect <host>`.
- Fullscreen page: routerHook.addRoute("/punktfunk") — host list,
per-host Pair/Stream, and a settings section (resolution/refresh/
bitrate/gamepad/mic, written to client-gtk-settings.json).
- Pairing: a gamepad-navigable PIN keypad. The host shows the PIN; the
backend runs the SPAKE2 ceremony headlessly via the client's new
`--pair <PIN> --connect host` CLI mode (app.rs), persisting the host
as paired so the stream then connects silently. Same flatpak =>
shared identity store, verified live (ceremony against a real host).
- Backend (main.py): discover / pair / runner_info / get_settings /
set_settings / kill_stream; uses DECKY_USER_HOME so paths resolve to
the deck user's flatpak install regardless of the plugin's root flag.
CI (decky.yml) and the sideload packager now ship bin/punktfunkrun.sh.
The Steam-shortcut launch and headless-pairing env follow MoonDeck
exactly but need a Deck in Gaming Mode to fully confirm.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On the Steam Deck there was no way out of fullscreen — no F11 key, and the
header bar (with the fullscreen button) is hidden while fullscreen.
- Controller: a Moonlight-style escape chord (L1+R1+Start+Select) held
together leaves fullscreen and releases input capture. The gamepad
service latches the chord (fires once per press) and signals the stream
page over an async channel; four simultaneous buttons no game uses as a
deliberate combo, so it can't trigger during play.
- Keyboard: F11 already toggled fullscreen (checked before input
forwarding, so it works while captured) — now surfaced.
- Discoverability: entering fullscreen flashes a 4s hint listing both
exits (F11 · L1+R1+Start+Select).
The escape future is aborted on page-hidden so a stale session can't act
on the shared window.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The lock prune (a5b99b2) stopped flatpak-builder full-cloning windows-rs
(disk-fill), but exposed the next layer: `cargo --offline --locked -p
punktfunk-client-linux` resolves the WHOLE workspace, so it still tried
to load the now-un-vendored windows-rs source for the
punktfunk-client-windows member (its windows-rs git deps are
cfg(windows)-gated, but cargo resolves all targets regardless) and
failed: "can't checkout ... you are in the offline mode".
Drop the Windows client from the workspace members inside the sandbox
build (sed on the copied Cargo.toml — the flatpak never compiles it) and
remove --locked (the lock no longer matches the reduced member set;
--offline still pins every crate to the vendored cargo-sources.json, so
the build stays reproducible). android stays — it has no git deps.
Verified locally: removing the member, `cargo build -p
punktfunk-client-linux --offline` Finishes with zero windows-rs
involvement; manifest YAML still valid.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
cargo fmt --all over the merged connect() call-sites (the video_caps/
launch args landed without a fmt pass). Comment-alignment only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The flatpak CI was failing at "Downloading sources" with "No space left
on device": flatpak-cargo-generator walks the whole workspace Cargo.lock
and emits a `type: git` source for the windows-rs crates (windows +
windows-reactor + ~12 sub-crates, pinned by punktfunk-client-windows),
and flatpak-builder then FULL-clones that multi-GB repo — for a bundle
that only ever compiles `-p punktfunk-client-linux` and never touches a
windows-* crate.
New packaging/flatpak/prune-windows-lock.py writes a copy of Cargo.lock
with the windows-rs git packages stripped (matches on the `source =`
line, so a crate that merely lists a windows dependency is kept;
dependency-free so it also runs on the Deck's stock python). Both the CI
and build-flatpak.sh feed that pruned lock to the generator. The
committed Cargo.lock is untouched — cargo --offline only needs vendored
sources for the crates it actually builds, and the windows-rs crates are
not in the Linux client's dependency closure.
Verified locally: 14 crates pruned (507 -> 493 packages), zero windows-rs
`source =` lines remain, output parses as TOML, all Linux-client deps
(gtk4/ffmpeg-sys-next/sdl3/pipewire) intact.
This unblocks the flatpak build carrying the VAAPI green-screen fix
(64b1679) for the Steam Deck.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First AMD test (Steam Deck, Mesa radeonsi) showed a mostly-green image
with red whites — the classic fingerprint of NV12 chroma read as 0.
Root cause (confirmed against FFmpeg/GTK/mpv source): FFmpeg's VAAPI
export uses VA_EXPORT_SURFACE_SEPARATE_LAYERS unconditionally, so an
NV12 surface comes back as TWO single-plane layers — layers[0]=R8
(luma), layers[1]=GR88 (chroma) — sharing one object/fd, the UV plane
reached via offset. map_dmabuf took layers[0] only and used its format
(R8) as the GTK fourcc, so GdkDmabufTexture got a luma-only texture
with the chroma plane dropped → chroma defaults to 0 → green field,
red highlights.
Fix (matches mpv's dmabuf_interop_gl flatten pattern):
- Derive the combined fourcc from the decoder's sw_format
(AVHWFramesContext.sw_format → NV12 → DRM_FORMAT_NV12), NOT from the
per-plane component formats. The frame format is absent from the
separate-layer descriptor and must be deduced from sw_format.
- Flatten every plane across every layer in declared order (Y then UV),
each with its own fd (objects[plane.object_index].fd), offset, pitch.
- One-time descriptor dump (objects/layers/formats/modifier) so a new
driver's real layout is visible in the logs.
- Unit test locks the DRM FourCC magic numbers (NV12=0x3231564e).
Software decode (swscale, reads colorspace from the VUI) was always
correct, which isolated the bug to this path. PUNKTFUNK_DECODER=software
is the immediate workaround on an un-rebuilt binary. Awaiting Steam Deck
reconfirm (no AMD VAAPI on the NVIDIA dev box to live-verify).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Light up the dormant 10-bit/HDR path end to end on the Windows client.
- core: NativeClient::connect gains a video_caps param threaded into the Hello. The Windows
client advertises VIDEO_CAP_10BIT | VIDEO_CAP_HDR; every other caller (the C ABI shim,
Linux, Android, host test connects) passes 0, so the 8-bit BT.709 path is unchanged. The
host already gates a Main10/PQ encode on these bits + PUNKTFUNK_10BIT.
- video.rs: a PQ frame (color_trc == SMPTE2084) converts 10-bit YUV → X2BGR10 (== DXGI
R10G10B10A2) with the BT.2020 matrix via sws_setColorspaceDetails; swscale applies only
the matrix + range, so the PQ-encoded samples pass through untouched.
- present.rs: on an HDR frame the swapchain flips in place (ResizeBuffers) to R10G10B10A2 +
DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020 + HDR10 metadata; the passthrough shader is
unchanged and the compositor maps PQ→display. Switched to ALPHA_MODE_IGNORE so the 10-bit
padding bits don't render transparent. SDR stays 8-bit B8G8R8A8.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The first real run on a display surfaced three issues the headless/dev-VM build never hit:
- Route each hook-using screen (hosts/pair/stream) as its own component() instead of
calling it with the shared cx. Calling hooks on the parent cx changed the hook order
when the screen flipped, tripping reactor's Rules-of-Hooks guard and aborting the moment
you navigated to the stream page.
- Mouse: replace the absolute path (which swallowed WM_MOUSEMOVE and so froze the OS cursor,
snapping the host pointer back to one point) with proper pointer lock — hide + ClipCursor
+ recentre, shipping relative MouseMove scaled by the Contain-fit factor. Ctrl+Alt+Shift+Q
now actually toggles capture: track modifier state from the hook's own event stream
(GetAsyncKeyState doesn't see keys we suppress in our own LL hook), and flush held
keys/buttons on release so nothing sticks on the host.
- Add the stats HUD overlay (mode · fps · Mb/s · capture→client/decode latency), mirroring
the Apple client. Stats live in root state and reach the stream page as a prop (a child's
own async-state update is pruned when props are unchanged), fed by a small poll thread.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The secure-desktop DDA leg went black with HDR on: legacy DuplicateOutput (the SDR-era
API) can't capture an FP16/HDR desktop, and dropping the SudoVDA out of HDR is denied on
the Winlogon desktop (so the SDR-drop attempt just churned and stayed black).
Instead capture HDR natively on the DDA path — the capturer already has the full
FP16→BT.2020 PQ→R10G10B10A2 conversion (hdr_fp16 path), it just never requested FP16.
Thread a want_hdr flag into duplicate_output: for an HDR session request
DuplicateOutput1 with FP16 first and retry it (5×) instead of bailing to the
HDR-incapable legacy fallback. The secure-desktop mux now reads the monitor's real HDR
state and opens DDA in HDR when set — no advanced-color toggling at all. The
normal-desktop DDA overlay/flip issues that pushed us to WGC don't apply to the composed
Winlogon UI. want_hdr is threaded through every (re)duplication incl. ACCESS_LOST recovery.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Keep HDR OFF for the DDA (secure-desktop) path rather than bailing to WGC: the DDA
capturer is SDR-only (BGRA8), so an HDR SudoVDA makes the Winlogon capture black.
On the secure transition, drop the monitor out of HDR and VERIFY it took (re-read
advanced_color_enabled, retry up to 6×200ms) before opening DDA — the CCD toggle can
transiently fail (rc=5) or lag. Restore HDR on return to the WGC normal-desktop leg.
Logs clearly if the drop can't be applied (e.g. denied on the Winlogon desktop).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
HDR streamed nothing and "didn't persist" because build() forced the SudoVDA's
advanced-color state to match the handshake bit_depth on every build — with an
8-bit-negotiated session (the common case: clients advertise no 10-bit cap) that
meant set_advanced_color(false) on every connect, wiping a user's deliberate
Windows HDR toggle on the virtual display.
But the whole pipeline already follows the monitor's REAL HDR state: WGC captures
FP16 when HDR is on, NVENC forces Main10 + BT.2020 PQ from the 10-bit capture
format regardless of the negotiated depth (encode/nvenc.rs), and the client
auto-detects PQ from the HEVC VUI. So the negotiated bit_depth must NOT drive the
monitor's colorspace.
- build(): only ever ENABLE HDR (proactively, for a negotiated 10-bit session);
never force it off. A user-enabled HDR session now persists and flows end-to-end.
- secure-desktop mux: gate the HDR→SDR drop (for the DDA leg) on the monitor's
ACTUAL advanced-color state at switch time, not bit_depth — so an HDR session
with an 8-bit handshake still drops correctly for Winlogon and restores after.
- sudovda: add advanced_color_enabled() reader (DISPLAYCONFIG_GET_ADVANCED_COLOR_INFO).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Final cleanup after the DDA-parity work, plus an end-user service to replace the
PsExec/VBS/scheduled-task launch chain.
Cleanup (behavior-preserving):
- sudovda.rs: drop the dead legacy GDI isolate_displays/restore_displays (CCD is
the sole isolation path), the always-empty Monitor.isolated field, and the
vestigial reassert_isolation + PUNKTFUNK_ISOLATE_DISPLAYS knob; fix stale comments.
- dxgi.rs: downgrade leftover debug warns/infos (DuplicateOutput1 retry, FALLBACKS,
hook-hits, AcquireNextFrame idle timeout) to debug!; remove the PUNKTFUNK_NO_CURSOR
per-frame test knob.
Windows service (src/service.rs, `punktfunk-host service`):
- SCM supervisor (windows-service crate) that duplicates its LocalSystem token,
retargets it to the active console session, and CreateProcessAsUserW's the host
there (Sunshine/Apollo model) — relaunching on exit and console session switch,
inside a kill-on-close job object so a service crash never orphans the host.
- install/uninstall/start/stop/status subcommands: one elevated `service install`
registers an auto-start LocalSystem service + firewall rules + a default host.env.
- Config moves to %ProgramData%\punktfunk\host.env; config_dir() now resolves to
%ProgramData%\punktfunk on Windows (replacing the APPDATA=C:\Users\Public hack),
with a PUNKTFUNK_CONFIG_DIR override. Logs land in %ProgramData%\punktfunk\logs\.
- merged_env_block (shared with the WGC helper) now also carries RUST_LOG.
- docs/windows-service.md + scripts/windows/host.env.example; windows-host.md updated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
User: tearing down + recreating the monitor per session is wrong both ways — a
fixed GUID collides on overlapping sessions, but a per-session GUID makes a new
screen on every reconnect; host-lifetime would leave a phantom display for
physical-screen users. Correct model = rock-solid state machine.
Replace the per-session create/REMOVE with a host-level reference-counted
manager (global MGR):
- States: Idle / Active{refs} / Lingering{until}.
- Connect (acquire): Idle→create; Lingering→reuse (cancel teardown, reconfigure
if the mode changed) — the quick-reconnect reuse, no new screen/PnP chime;
Active→refs++ (concurrent / Reconfigure-overlap), reconfigure on a mode change.
- Disconnect (release, via the MonitorLease keepalive Drop): refs-- ; at 0 →
Lingering(now + PUNKTFUNK_MONITOR_LINGER_MS, default 10s).
- Background timer: Lingering past its deadline → REMOVE the monitor → Idle, so a
physical screen returns ~10s after streaming stops.
Eliminates BOTH the cross-session REMOVE collision (teardown only at refs==0 +
expired grace) and the new-screen-on-reconnect, without a persistent phantom
display. The control-device handle is opened once (host-level) — a handle, not a
screen. SudoVdaDisplay is now a marker; the old create() body is create_monitor.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
User: re-adding WGC brought back the teardown/recreate bug (audible disconnect/
connect on the secure<->normal switch). Cause: the secure->normal switch called
build() = vd.create() = IOCTL_REMOVE old SudoVDA monitor + IOCTL_ADD new one +
respawn the helper — the same teardown/recreate kernel stress we just eliminated
from DDA, now on the mux path.
Apply the same learning (reuse, don't tear down): the SudoVDA monitor and WGC
helper persist for the whole session; only the host-DDA leg opens (on secure)
and closes (on normal). On returning to normal, RESUME the still-alive helper
(drain its secure-dwell backlog + request a keyframe) instead of rebuilding.
The HDR-session colorspace restore (set_advanced_color(true) + helper rebuild)
is kept ONLY for bit_depth>=10 — an SDR session never changed the colorspace, so
it needs no rebuild at all. The secure switch already reuses the monitor
(open_dda on the existing target).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Re-test still broken: the WGC helper captured HDR FP16 BT.2020 PQ from the FIRST
frame (before any switch), feeding the 8-bit SDR encoder → broken normal-desktop
image. Root cause: the SudoVDA's advanced-color (HDR) state PERSISTS on the
monitor across sessions, so the 8-bit session inherited HDR left enabled by the
earlier broken toggle — and gating the per-switch toggles can't undo a state
that's already on at start.
Fix: in build() (runs on initial create + every mode-switch/return-from-secure
rebuild), force set_advanced_color(target, bit_depth>=10) BEFORE spawning the
WGC helper, with a 250ms settle if it changed. An 8-bit session now always
captures SDR via WGC (matching the encoder); 10-bit keeps HDR.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Re-enabling the WGC relay brought back a broken image on the secure->normal
switch. Log root cause: on returning to the normal desktop the relay called
set_advanced_color(target, true) to 'restore HDR', so the rebuilt WGC helper
captured HDR FP16 BT.2020 PQ while the session encoder is 8-bit SDR -> format
mismatch (the 'HDR gets restored when flipping back to WGC' bug).
Gate BOTH set_advanced_color toggles on bit_depth>=10. An SDR (8-bit) session
now stays SDR across WGC<->DDA switches (no HDR force, no needless topology
change); HDR sessions keep the drop-on-secure / restore-on-normal behavior.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Remove 4 unused imports (PCWSTR in composed_flip, anyhow macro + SizeInt32 in
wgc, Write in wgc_relay).
- DuplicateOutput1 retry defaults to N=1 (immediate legacy): on the secure
desktop DuplicateOutput1 is LOGON_UI-only so it always refuses, and the
release-before-reduplicate + gentle recovery keep the legacy dup stable;
retrying there only blocked. Still env-tunable (PUNKTFUNK_DUP_RETRY_N/_MS).
- Throttle the 'using legacy DuplicateOutput' warning (expected + once-per-gentle-
recovery on secure) so a lock dwell doesn't flood the log.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
User's observation: entering UAC/lock works instantly, but clicking OUT of it
breaks (with the disconnect sound) — Apollo's enter and leave are symmetric.
Root cause: attach_input_desktop() (SetThreadDesktop to the current input
desktop) was gated behind is_secure_desktop() in recreate_dupl, so:
- Default->Winlogon (enter): is_secure==true -> re-attach to Winlogon -> works.
- Winlogon->Default (leave): is_secure==false -> SKIP re-attach -> the capture
thread stays stuck on the now-gone Winlogon desktop -> every rebuild fails ->
no frames -> client timeout -> session ends -> SudoVDA removed (the disconnect
sound).
Fix: call attach_input_desktop() UNCONDITIONALLY on every rebuild (Apollo calls
syncThreadDesktop before every duplicate), so leaving secure re-attaches to the
returned desktop. reassert_isolation stays secure-only. Also stop leaking the
HDESK (CloseDesktop right after SetThreadDesktop, like Apollo) so calling it on
every recovery is safe.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per the user's insight: on the secure (Winlogon) desktop the duplication dies on
every independent-flip, and our tight recovery loop tore it down + recreated it
hundreds of times/sec — that release/recreate cycle is the real kernel stress,
and it stalled the send thread long enough that the client timed out ('display
disconnected'). Normal-desktop streaming is already solid (per-session GUID
killed the collision); this only changes the loss-recovery cadence.
Gentle recovery (user chose 'keep session alive'):
- cap the cheap re-duplicate to PUNKTFUNK_RECOVER_MS (default 250ms, was 5ms)
- cap the heavy new-device rebuild to PUNKTFUNK_REBUILD_MS (default 1500ms, was
250ms) — it's the costliest teardown, throttled hardest
- repeat the last frame between attempts (no busy-spin, no 8ms sleep)
~200/s -> ~4/s teardown/recreate during a secure dwell. The session survives
lock/UAC (frozen/laggy secure screen, then clean resume on unlock) instead of
churning the kernel into a disconnect. Both cadences env-tunable.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
User observed: 'display disconnected' + freeze with NO context change, and
'first switch happy, subsequent slower, then chaos under stress'. Log shows the
cause: MONITOR_GUID was a FIXED constant, so overlapping sessions (a client
RECONNECTING after a freeze before the old session tore down, or concurrent
sessions) all map to the SAME SudoVDA monitor (same GUID -> IOCTL_ADD reuses
target 257). When the old session ends, its IOCTL_REMOVE tears the monitor down
OUT FROM UNDER the live session -> 'display disconnected' + the late
E_INVALIDARG/MODE_CHANGE failures (output vanished mid-session) -> cascade.
Fix: next_monitor_guid() returns a unique GUID per (process, session) [base GUID
with low 48-bit node = pid<<16 | session#]; create() threads it into AddParams
AND the keepalive (which REMOVEs by it). Each session now owns its own monitor;
one ending can't kill another. (The 200ms DuplicateOutput1 retry confirmed
working — 'succeeded on retry' logged; the residual failures were this
collision, not the race.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The old-dup kernel teardown takes ~200ms (Apollo waits exactly that), so the
previous 2-16ms retries were too short and still fell through to the churning
legacy dup. Bump to PUNKTFUNK_DUP_RETRY_MS (default 200) x PUNKTFUNK_DUP_RETRY_N
(default 6) so the robust DuplicateOutput1 dup wins the race. Env-tunable for
on-box dialing without a rebuild.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
User's insight, and it fits the evidence exactly: in duplicate_output the FIRST
DuplicateOutput1 (called microseconds after the caller releases the old
duplication via self.dupl=None) returns E_ACCESSDENIED, but the legacy
DuplicateOutput a beat later SUCCEEDS — the only difference is TIMING. The
kernel-side teardown of the just-released duplication is async, so the immediate
DuplicateOutput1 races it ('output still duplicated' -> E_ACCESSDENIED). We then
fell straight through to legacy DuplicateOutput, which 'succeeds' into a FRAGILE
dup that churns ACCESS_LOST/MODE_CHANGE every few ms on this cross-GPU IDD
(causing the post-login freeze + UAC-confirm drop).
Fix: retry DuplicateOutput1 up to 5x with escalating 2/4/8/16 ms waits before
falling back to legacy, so the teardown finishes and the ROBUST DuplicateOutput1
dup succeeds (no churn). Bounded (~30 ms worst case) so a genuine failure still
falls back quickly. This is exactly Apollo's 2x/200ms retry rationale.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
DuplicateOutput1 returned E_ACCESSDENIED ~8815x even with PER_MONITOR_AWARE_V2
confirmed on the capture thread (thread_is_v2=true) — so DPI was NOT the cause.
The real cause: DXGI permits only ONE IDXGIOutputDuplication per output, and on
ACCESS_LOST you MUST release the old one before re-duplicating. Our recovery
(try_reduplicate / recreate_dupl) created the NEW duplication while the OLD
self.dupl was still alive → the output stayed held → DuplicateOutput1
E_ACCESSDENIED and the legacy fallback returned a BORN-LOST dup. It never
converged because there was always exactly one stale dup alive at creation
time. The initial open() works precisely because there's no prior dup; Apollo
is clean because it releases (dup.reset()) before every re-DuplicateOutput.
Fix: make self.dupl an Option and set it to None (drop → release the output)
BEFORE duplicate_output in try_reduplicate and before reopen_duplication in
recreate_dupl, then Some(new). acquire() gets a None-guard that synthesizes
ACCESS_LOST (routes into recovery) so a transient None can't panic. All
ReleaseFrame/AcquireNextFrame sites updated for the Option.
This is the documented DDA recovery requirement and the one thing that
distinguished our failing DuplicateOutput1 from Apollo's working one.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The remaining born-lost ACCESS_LOST storm traces to ONE thing: our
IDXGIOutput5::DuplicateOutput1 returns E_ACCESSDENIED (0x80070005) ~4370x, so
we fall back to legacy DuplicateOutput, which yields a BORN-LOST duplication on
this hybrid box. Apollo's DuplicateOutput1 SUCCEEDS on the identical
desktop/output/4090-device → a working dup, clean capture.
Root cause: DuplicateOutput1 REQUIRES Per-Monitor-Aware-V2. At startup our
SetProcessDpiAwarenessContext(PER_MONITOR_AWARE_V2) FAILS with E_ACCESSDENIED
('already set' — a manifest/runtime locked the process to a lower awareness),
and GetAwarenessFromDpiAwarenessContext reports 2 for BOTH Per-Monitor V1 and
V2, so the earlier 'awareness=2' was misleading — the process is likely V1,
which DuplicateOutput1 rejects with E_ACCESSDENIED. (Legacy DuplicateOutput has
no V2 requirement, so it 'worked' but born-lost.)
Fix: SetThreadDpiAwarenessContext(PER_MONITOR_AWARE_V2) on the capture thread
in open() — a per-thread override that takes regardless of the process default,
so DuplicateOutput1 can succeed (the working dup Apollo gets). Logs set_ok +
thread_is_v2 (via AreDpiAwarenessContextsEqual) to confirm V2 actually applied.
Topology fixes (sole display, no MODE_CHANGE) and the recovery backstops stay.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Live result of the previous build: the MODE_CHANGE_IN_PROGRESS storm was FIXED
(0 occurrences) by dropping primary-promotion — but it exposed the regression
the review predicted: a non-primary EXTENDED SudoVDA is NOT DWM-composited on
this box, so DDA gets born-lost ACCESS_LOST (0x887a0026) + black frames. The
IDD genuinely must be the sole/primary/composited display here.
Apollo reaches that end state ('Virtual Desktop: 5120x1440', sole display) via
Windows AUTO-promoting the real WDDM display over the box's leftover 1024x768
basic display — but Windows does NOT auto-promote for us, leaving the IDD
extended. So make it sole explicitly, the clean way:
- create(): deactivate the other display(s) via the atomic CCD path
(isolate_displays_ccd) by DEFAULT (opt out with PUNKTFUNK_NO_ISOLATE). Drop
the legacy per-device GDI detach from the path (it misses iGPU-attached
monitors and churns; kept #[allow(dead_code)] for reference).
- set_active_mode(): CDS_UPDATEREGISTRY only — set the mode in place, NO
CDS_SET_PRIMARY / CDS_GLOBAL / DM_POSITION. A sole display is already primary,
so there's nothing to contest → no MODE_CHANGE storm (that storm came from
promoting primary at (0,0) WHILE the basic display was still active).
Net: sole SudoVDA → primary → composited → capturable, with no topology
contest. Keeps the prior MODE_CHANGE-as-transient handling + removed born-lost
escape as backstops.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
ROOT CAUSE (verified by multi-agent compare vs Apollo + adversarial review):
set_active_mode() applied the SudoVDA mode with CDS_UPDATEREGISTRY | CDS_GLOBAL
| CDS_SET_PRIMARY + DM_POSITION(0,0) — promoting the freshly-added IDD to
PRIMARY at the virtual-screen origin and persisting it globally. On this box
(baseline active display = a 1024x768 basic 'WinDisc') that primary-promotion
contests the existing display so the desktop topology never reaches a stable
fixed point → every DuplicateOutput/AcquireNextFrame during the unending
settle returns DXGI_ERROR_MODE_CHANGE_IN_PROGRESS (0x887A0025). Apollo, live
on this EXACT box with an empty config, never promotes primary and captures
the same SudoVDA at 5120x1440 with zero DXGI errors. (Ruled out earlier on the
live box: win32u hook, DPI, independent-flip/overlay, isolation, render pin.)
Fixes (subtractive, gated per adversarial review):
- sudovda.rs set_active_mode: default to CDS_UPDATEREGISTRY only (no primary
promotion, no GLOBAL, no DM_POSITION) = Apollo-parity for the multi-display
default. Promote to primary (CDS_GLOBAL|CDS_SET_PRIMARY+DM_POSITION) ONLY
when PUNKTFUNK_ISOLATE_DISPLAYS=1 (sole display, where a blank extended IDD
would otherwise yield no frames). Avoids regressing headless/isolated +
mid-stream Reconfigure.
- dxgi.rs acquire: treat MODE_CHANGE_IN_PROGRESS (0x887A0025) as a TRANSIENT
(Ok(None), repeat last frame, wait it out) instead of falling through to the
fatal Err arm → cold-rebuild → create()→set_active_mode (which re-issued the
mode change and amplified the storm).
- dxgi.rs acquire: remove the born-lost cold-rebuild escape — it re-created the
SudoVDA (IOCTL REMOVE/ADD = the audible PnP chime the user heard) and never
converged; now repeat last frame in-process (never tear the IDD down mid-
session, like Apollo). Overlay + cheap-spin/HDR recovery left intact.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
GROUND TRUTH from Apollo streaming live on this exact box (empty config):
captures the SudoVDA at 5120x1440@240 on the RTX 4090 with ZERO ACCESS_LOST /
born-lost / MODE_CHANGE -- clean, no overlay, no isolation, no render pin. That
disproves the independent-flip theory (a sole SudoVDA captures fine here) and
points at something WE do that Apollo doesn't.
The concrete culprit: we call SET_RENDER_ADAPTER, which this driver IGNORES
(logs 'render adapter DIFFERS from pinned add=0x23664 pinned=0x15768') and the
IDD ends up rendering on adapter 0x23664 while its DXGI output is enumerated
under the 4090 (0x15768) where we create the capture device -- a cross-GPU
mismatch that is the real source of the perpetual ACCESS_LOST +
MODE_CHANGE_IN_PROGRESS (0x887A0025) storm. Apollo never pins (empty config),
so its IDD stays on its natural adapter, aligned with capture.
Make the render pin OPT-IN (PUNKTFUNK_RENDER_ADAPTER=<name>); default to NOT
pinning, matching Apollo. The startup log now shows the resulting AddOut LUID
so we can confirm the IDD lands on the 4090.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CONFIRMED on the live RTX4090+iGPU box: hook fires+verified, DPI=2, overlay
running, yet the stream STILL freezes -- born-lost dropped but MODE_CHANGE_IN_
PROGRESS (0x887A0025) churn took over (2284x) and frames go stale. Root cause
is the topology itself: create() makes SudoVDA the SOLE active display
(CDS_SET_PRIMARY + isolate_displays + isolate_displays_ccd), and a sole display
on a hybrid box goes into fullscreen independent-flip / MPO that Desktop
Duplication cannot capture.
Apollo is rock solid on this EXACT box because it does the opposite: it keeps
the physical monitor ACTIVE and arranges the virtual display alongside it
(rearrangeVirtualDisplayForLowerRight, 'Do not change the primary'). Multi-
display is DWM-composited, so the output never independent-flips.
Make isolation OPT-IN (PUNKTFUNK_ISOLATE_DISPLAYS=1) and default to NOT
isolating -- match Apollo's multi-display topology. SudoVDA stays primary (so
it carries the shell -> frames) but other monitors stay active, which disables
independent-flip. reassert_isolation honors the same flag (re-isolating mid-
stream would itself trigger the storm). Keeps the overlay + born-lost escape
as belt-and-suspenders.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
NsdManager service discovery needs NEARBY_WIFI_DEVICES on Android 13+. The app DECLARED it but
never REQUESTED it, so on a real device the permission stayed denied and discoverServices silently
found nothing — no prompt, no hosts. (It only worked on the emulator because the permission was
granted via `adb pm grant`.) Request it (mirroring the mic RECORD_AUDIO flow) when the connect
screen appears, and start/restart discovery once granted; on API < 33 discovery starts immediately
(the permission doesn't apply there). The advertised hosts the Apple clients already see will then
appear here too.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
CONFIRMED root cause via instrumented build: hook_hits=1+ (win32u hook fires,
verified-patched) and DPI awareness=2 (PER_MONITOR), yet the born-lost
ACCESS_LOST storm persists with 100% DuplicateOutput1 E_ACCESSDENIED. That
rules out reparenting (the hook works) and DPI -> it is fullscreen
independent-flip / MPO: the SudoVDA virtual display, isolated as the SOLE
active output, scans out one plane on one display, bypassing DWM composition,
so Desktop Duplication gets a born-lost duplication.
Apollo never hits this because it runs WITH a physical monitor attached
(multi-display is already DWM-composited); we isolate to sole-display, so we
must force composition ourselves. The fix already existed (ForceComposedFlip,
a tiny topmost layered overlay that disqualifies independent-flip) but was
only wired into the WGC relay path's secure branch, which PUNKTFUNK_NO_WGC=1
disables. Wire it into virtual_stream unconditionally (DDA owns the normal
desktop here, where the storm is). Held for the session; Drop tears it down;
PUNKTFUNK_FORCE_COMPOSED=0 disables.
Keeps the prior build's born-lost escape as a safety net.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
On the test phone's launcher the standard (colored) adaptive foreground rendered noticeably larger
than the themed (monochrome) layer — identical geometry, but the launcher insets/scales the two
differently — so the colored circles overflowed the circle mask. Shrink only the foreground group
(scale 0.105 → 0.073, re-centred) to match the correctly-sized monochrome; the monochrome layer is
unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The hybrid RTX4090+iGPU box storms DXGI_ERROR_ACCESS_LOST (0x887A0026) +
MODE_CHANGE_IN_PROGRESS (0x887A0025) ~3s after first frame: every rebuilt
duplication is born-lost (created OK, first AcquireNextFrame instantly
ACCESS_LOST), seeds black, retries forever. The steady-state m3 loop calls
try_latest()->acquire() which returns Ok(None) on every recovery, so the
cold-rebuild escape (MAX_CAPTURE_REBUILDS) was unreachable -> frozen stream.
Multi-agent root-cause + adversarial review point at the win32u GPU-pref hook
being ineffective (patched on the main thread, no FlushInstructionCache, never
verified) rather than the synthesis's independent-flip theory (Apollo has no
overlay yet is stable on this exact box).
This build instruments + applies the safe, high-probability fixes:
- Hook: FlushInstructionCache after the inline patch (cross-thread i-cache);
read back the 12 patched bytes and error! if they didn't land; per-call hit
counter (hybrid_hook_hits) logged after open -- hits==0 proves the hook is
off DXGI's reparent path.
- DPI: log SetProcessDpiAwarenessContext result + effective awareness (need
2=PER_MONITOR for DuplicateOutput1; explains the 100% E_ACCESSDENIED).
- SetThreadExecutionState(ES_CONTINUOUS|ES_DISPLAY_REQUIRED|ES_SYSTEM_REQUIRED)
at capture open, restored on Drop -- stop IDD idle-invalidation (Apollo does
this too).
- Born-lost escape: count consecutive born-lost rebuilds; on the NORMAL desktop
(never the secure/Winlogon dwell) escalate to Err after ~5s so the m3 loop
cold-rebuilds the whole pipeline instead of freezing on the last frame.
Diagnostic-forward: one test now tells us hook-hits + DPI awareness + whether
ExecutionState/desktop-sync alone fixes it, and the stream self-recovers
instead of wedging.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the placeholder system icon with the Punktfunk brand mark (two overlapping violet circles,
from the shared logo in clients/apple/.../punktfunk_Logo.icon).
- drawable/ic_launcher_foreground.xml: the violet logo (3 exact paths) scaled + centered into the
108dp adaptive-icon safe zone via a group transform.
- drawable/ic_launcher_monochrome.xml: single-tone silhouette for Android 13+ themed icons
(Material You) — the launcher recolors it to the wallpaper.
- mipmap-anydpi-v26/ic_launcher{,_round}.xml: adaptive-icon (background + foreground + monochrome);
dark-indigo background (@color/ic_launcher_background) so the violet pops.
- Manifest: android:icon=@mipmap/ic_launcher + roundIcon (was @android:drawable/sym_def_app_icon).
minSdk 31 → anydpi-v26 covers every device (no legacy PNG mipmaps needed). Verified on a physical
phone (Android 16): the icon renders centered + circle-masked; the themed-icon layer is wired.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The freeze on context change is the lock/login rendering on a PHYSICAL monitor
instead of the captured SudoVDA display. Root cause: the legacy isolate_displays
(EnumDisplayDevices + ChangeDisplaySettings) found NOTHING to detach on this hybrid
box (4090 + AMD iGPU) — an iGPU-attached monitor isn't flagged ATTACHED_TO_DESKTOP
in the GDI enum, so it's never detached and the secure desktop lands on it while the
virtual output freezes. (Log: isolate ran, logged zero "detaching" lines.)
Add CCD-based isolation (QueryDisplayConfig(QDC_ONLY_ACTIVE_PATHS) + SetDisplayConfig)
— the API Apollo uses, which sees every active path. Deactivate all active paths
except the SudoVDA target's, leaving the virtual display the sole desktop so ALL
content (incl. Winlogon) renders to it. Runs alongside the legacy pass (now a no-op
fallback); the original topology is saved and restored on teardown before REMOVE.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The one major capture-API difference left vs Apollo: punktfunk used legacy
IDXGIOutput1::DuplicateOutput; Apollo uses IDXGIOutput5::DuplicateOutput1 with a
format list, the modern path that's more robust to overlay/format changes (a
candidate for the SudoVDA-on-hybrid 0x887A0026 churn). Add a duplicate_output()
helper used at all 3 duplication sites (open, reopen_duplication, try_reduplicate):
QI to IDXGIOutput5 and DuplicateOutput1, falling back to legacy DuplicateOutput.
DuplicateOutput1 requires per-monitor-v2 DPI awareness, so set that at process
start alongside the GPU-pref hook (matches Apollo).
Format list is BGRA8-only for now (SDR test): DuplicateOutput1 returns the first
format it can CONVERT to, so FP16-first would hand back FP16 even on SDR and trip
the HDR path. Real FP16/HDR capture (with IDXGIOutput6 colorspace detection) is the
follow-up once the churn is settled.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The win32u hook only works if it patches before DXGI caches the hybrid preference.
It was installed in DuplCapturer::open (first capture), but the SudoVDA
render-adapter selection creates a DXGI factory during virtual-display setup —
seconds earlier — so the preference was already cached and the hook had no effect
(churn persisted; log showed "render adapter chosen" at :02, "hook installed" at
:04). Call install_gpu_pref_hook() at the top of real_main(), before any command
runs, so it beats the first DXGI factory. (open() still calls it too; Once makes
the earliest call win.) Also fix the cosmetic function-cast-as-integer warning.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Root cause of the ACCESS_LOST (0x887A0026) churn + context-change freeze, found
live: the box is a HYBRID system (RTX 4090 + AMD Radeon iGPU + SudoVDA). DXGI does
hybrid GPU-preference resolution and REPARENTS the SudoVDA output between adapters
(SET_RENDER_ADAPTER is ignored — the IDD lands on the iGPU 0x23664 while we
duplicate on the 4090 0x15768), which constantly invalidates Desktop Duplication.
Apollo runs fine on this same box because it hooks this away.
Port Apollo's hook: replace win32u.dll!NtGdiDdDDIGetCachedHybridQueryValue to always
report D3DKMT_GPU_PREFERENCE_STATE_UNSPECIFIED, so DXGI skips preference resolution
and never reparents the output → DDA stays on one adapter. Installed once before the
first DXGI factory/enumeration (DuplCapturer::open). We fully replace the function
(never call the original) so a 12-byte absolute-jmp prologue patch suffices — no
detour crate / C length-disassembler dependency, just VirtualProtect.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
recreate_dupl called reassert_isolation (a display-TOPOLOGY change via
isolate_displays) + attach_input_desktop on EVERY ACCESS_LOST rebuild — 200×
in a 6 s SDR session. A topology change itself invalidates the freshly-rebuilt
duplication, so the next acquire is ACCESS_LOST → recreate → reassert → a
self-feeding 0x887A0026 churn that freezes the stream and never recovers across
context changes (lock / login / post-login).
Gate both behind is_secure_desktop(): the heavy topology work runs only on the
actual Winlogon (secure/login) desktop — where a physical monitor can grab the
secure desktop off our virtual output. Routine churn, the lock screen, and
post-login are all on the normal desktop, so they take a light re-duplicate with
no topology meddling. Apollo isolates once at startup; its recovery just
re-duplicates — this matches that.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two freeze drivers found live on the RTX box (DDA-only, 5K@240 HDR SudoVDA):
Step 1 — the per-frame format-change check (995db69) mis-fired EVERY frame in HDR
(827+/session): self.hdr_fp16 is derived from the duplication ModeDesc (FP16
scanout mode), but legacy DuplicateOutput always hands back 8-bit BGRA, so the
acquired-texture format never equals hdr_fp16 → a rebuild storm (each rebuild
re-inits device+NVENC → freeze). Make the acquire check SIZE-only; a real
HDR<->SDR toggle still arrives as ACCESS_LOST → recreate_dupl re-detects it.
Step 3 — ACCESS_LOST (0x887A0026) churn: HDR overlay/MPO flips invalidate the
duplication continuously and the recovery loop had no rate limit (the 250ms
throttle guarded only the full rebuild, not the cheap try_reduplicate), so it
spun DuplicateOutput + up-to-16ms Acquire and starved the encode thread. Add a
last_recover throttle capping ALL recovery attempts to ~one per 5ms; between
attempts return None so the caller repeats the last frame, paced at the frame
interval (no busy-spin, encode thread keeps running).
Real FP16 HDR capture (DuplicateOutput1) + per-loss desktop-reisolation cleanup
are the next steps; validate this in SDR first.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
DDA only re-read the duplication format/size on rebuild (recreate_dupl) and
initial open. A mid-stream HDR<->SDR flip (FP16<->BGRA — e.g. the SudoVDA output
dropping out of HDR for the secure desktop) or a resolution change that does NOT
raise ACCESS_LOST left hdr_fp16/width/height stale, so present_acquired copied
into a mismatched-format/size target — the secure-desktop "works once, then HDR
breaks" symptom. Re-read the acquired texture's desc every frame (as Apollo does)
and rebuild on a real change instead of presenting a mismatched frame; throttled
like the ACCESS_LOST path so a flapping toggle can't hammer DuplicateOutput.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A single test flag to bring up / validate DDA on its own and as the base for the
secure-desktop work. When set it (1) skips WGC in capture_virtual_output (forces
dxgi::DuplCapturer, same as PUNKTFUNK_CAPTURE=dda) and (2) makes should_use_helper
return false, so even a SYSTEM host bypasses the two-process WGC relay and captures
in-process with one DDA capturer for both the normal AND the secure desktop —
Apollo's model. All the WGC / relay code stays compiled; unset the flag to restore.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When the DDA-on-secure path is enabled (PUNKTFUNK_SECURE_DDA=1), the mux now
toggles the SudoVDA's advanced-color (HDR) state via the CCD API
(sudovda::set_advanced_color → DisplayConfigSetDeviceInfo +
DISPLAYCONFIG_SET_ADVANCED_COLOR_STATE): on entering the secure (Winlogon)
desktop it disables HDR so the lock/UAC renders SDR/composed (no fullscreen
independent-flip → DDA can duplicate it instead of storming ACCESS_LOST/black),
opens DDA fresh on the now-SDR output; on returning to normal it re-enables HDR
and rebuilds the helper so WGC re-detects the restored colorspace.
Also debounce the DesktopWatcher (publish a Default↔Winlogon change only after it
is stable ~80ms) so transient flaps during the transition don't thrash the mux.
Default (no flag) is unchanged: WGC stays live through a lock, no DDA switch.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Under infinite GOP the punktfunk/1 plane has no periodic IDR — the only recovery
keyframe is one the client requests. But the reassembler drops unrecoverable AUs
silently (frames_dropped) and hands the decoder reference-missing delta frames
that libavcodec conceals and returns Ok for, so keying recovery off a decode
error mostly never fires under real loss → a long/permanent freeze.
Surface the data-plane pump's Session.frames_dropped to NativeClient via a shared
atomic (NativeClient::frames_dropped()), updated every pump iteration so it stays
current through a total-loss drought. The Linux and Windows client video loops
watch it and call request_keyframe() when it climbs, throttled to 100 ms (the
decode stays wedged for several frames until the IDR lands). macOS already does
this; client-rs doesn't decode.
Resolves reliability backlog #2.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two steady-state faults previously bubbled a bare `?` to conn.close / silently
muted the rest of a session. Recover in place instead.
#4 — capture loss (virtual_stream): a mid-session capture stall/disconnect
(`try_latest` Err: PipeWire/compositor thread ended, virtual output gone) ended
the whole session — and the native client has no reconnect path, so it had to
cold-restart the handshake. Now rebuild the pipeline IN PLACE at the current
mode via build_pipeline_with_retry (same primitive the mode/session switch uses),
force a keyframe, and only propagate when the bounded retry is exhausted. A
consecutive-rebuild cap stops a flapping source from looping the client through
endless cold IDRs. Track the live mode so a rebuild after a mode switch targets
the right mode (also fixes the session-switch rebuild using the stale mode).
#3 — native audio thread (audio_thread): broke the loop on ANY next_chunk Err,
spawned once per session and never restarted, so a transient 5 s quiet-sink
timeout permanently muted a multi-hour session. Make a quiet sink return an empty
chunk (not an Err) in both backends so only a genuinely dead capture thread is an
Err, and reopen-with-backoff (INJECTOR_REOPEN_BACKOFF) on death, keeping the Opus
encoder + monotonic seq. Documents the next_chunk contract; also makes the
GameStream audio sender survive quiet sinks for free.
Resolves reliability backlog #3 and #4.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Regression fix. The DDA-on-secure mux + force-composed overlay + rebuild-on-switch
made the stream worse than just staying on WGC: DDA can't reliably capture the
secure desktop's HDR independent-flip (storms ACCESS_LOST → instant black), and
rebuilding the output on every Default↔Winlogon flip thrashed (frequent freezes).
Meanwhile the WGC helper STAYS LIVE through a lock/UAC.
So make the DDA-on-secure path OPT-IN (PUNKTFUNK_SECURE_DDA=1, or the test
toggle). By default the mux keeps WGC the whole session — the DesktopWatcher and
the force-composed overlay aren't even started — so a lock/UAC no longer black-
screens or freezes the stream. The DDA-secure machinery stays in the tree for
future experimentation behind the flag.
(Reverts the rebuild-on-every-switch change 3f191ba via 555ec2a; this gates the
remaining switch.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Key insight (from the user): a fresh RECONNECT shows the secure desktop but the
live transition does not — so the difference is what a fresh session does that
the live switch skipped. A reconnect runs build() = REMOVE + fresh ADD of the
SudoVDA monitor + re-isolate + a fresh capturer; the live transition instead
reused the session-start output (created while on the NORMAL desktop), which goes
born-lost (ACCESS_LOST storm → black) on the secure desktop.
Fix: virtual_stream_relay now calls build() on EVERY source switch (both WGC→DDA
and DDA→WGC), then opens DDA on the new target for secure / uses the fresh helper
for normal. This makes each transition equivalent to the reconnect that works —
fixing both the WGC→DDA cutover (secure desktop now in the clean output state DDA
can duplicate) and the DDA→WGC cutover (a fresh helper's first frame is its
opening IDR, so await_idr clears immediately instead of waiting on a wedged
helper). Costs a ~1-2s rebuild per transition, acceptable for UAC/lock events.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The secure (Winlogon: UAC/lock/login) desktop presents via fullscreen
independent-flip/MPO — it scans out bypassing DWM composition, so DXGI Desktop
Duplication returns born-lost DXGI_ERROR_ACCESS_LOST (the client sees black; the
UAC only "flashes" during the brief composed transition). Confirmed live: stable
4090 LUID across the storm (NOT reparenting) on an FP16 HDR output, recovering
only when the screen changes.
Fix (non-input, no system-wide registry change): capture/composed_flip.rs keeps a
tiny click-through near-invisible TOPMOST LAYERED window alive on the current
input desktop. Any visible window on the output disqualifies independent-flip →
DWM composites → DDA can capture. A dedicated thread follows the input desktop
(Default↔Winlogon) and recreates the window there on each switch (a window is
bound to its desktop), re-asserting topmost + pumping messages every 200ms.
Started for the two-process stream's lifetime; gated by PUNKTFUNK_FORCE_COMPOSED
(default on, =0 to disable). Needs GENERIC_ALL on OpenInputDesktop for
DESKTOP_CREATEWINDOW (0x80070005 otherwise). Validated: overlay creates on the
Default desktop; live lock test pending.
Also includes SET_RENDER_ADAPTER (sudovda.rs, Apollo item #16): pins the IDD
render GPU to the NVENC GPU before ADD — issued + accepted live, though the
secure-desktop storm was proven to be independent-flip (stable LUID), not
reparenting, so it's correctness/hygiene here rather than this bug's fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The GameStream video sender did one send() syscall per packet on Windows
(the #[cfg(not(target_os="linux"))] sendmmsg_all fallback), capping
throughput at high packet rates. Wire it to UDP Send Offload (the Windows
analogue of Linux GSO) so each paced 16-packet burst goes out in one
WSASendMsg(UDP_SEND_MSG_SIZE) syscall instead of 16, preserving the
microburst pacing.
Expose a reusable punktfunk_core::transport::send_uso_all (Windows-only)
that reuses the proven native-plane USO primitive (send_one_uso + the uso
on/off latch + uso_unsupported), with the same uniform-size guard and
≤512-segment chunking as UdpTransport::send_gso. It returns how many leading
packets it sent via USO; the GameStream sendmmsg_all sends any remainder
(USO off via PUNKTFUNK_GSO=0, a size-mixed burst, or a frame's short final
packet) with per-packet send. On-wire packet boundaries are unchanged.
Resolves#4 in docs/apollo-comparison.md. Linux build unaffected;
punktfunk-core type-checks for x86_64-pc-windows-msvc. Host Windows compile
deferred to CI / dev box.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
#13 (two-pass alpha+XOR cursor) implemented in capture/dxgi.rs. #21
(composite moved cursor without a new desktop frame) is already handled:
DXGI returns S_OK for pointer-only updates so punktfunk recomposites in
present_acquired; the original premise (stutter via timeout) was incorrect.
Adds status banner + per-item resolution notes in Part 4 and Part 3.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A single DXGI cursor shape can need BOTH an alpha-blended layer AND a
screen-inverting (XOR) layer at once — a masked-color text I-beam (opaque
hot-spot + inverting bar) or a monochrome cursor mixing opaque and invert
pixels. The old path produced ONE BGRA image per shape and picked ONE blend
(cursor_invert) for the whole shape, so such mixed cursors rendered wrong
(masked-color opaque pixels forced through the invert blend; monochrome
(AND=1,XOR=1) invert pixels approximated as solid black).
Port Apollo/Sunshine's decomposition: convert_pointer_shape now returns a
CursorShape with optional alpha/xor layers; CursorCompositor holds tex_alpha
+ tex_xor and draw_layer renders each with its own blend (alpha = src-over,
HDR-scaled; XOR = inversion, unscaled — it operates on the framebuffer
reference). The CPU software path blends both layers too. Empty layers are
never uploaded or drawn. Removes the single cursor_invert flag.
Fixes#13 in docs/apollo-comparison.md. Independently reviewed (ship);
Windows-only code — compile verified by CI / dev VM.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Sign every MSIX build with one STABLE self-signed cert instead of a fresh per-build cert, so the
Trusted People import is a one-time, per-machine step that survives upgrades (a fresh cert each build
forced a re-import every time). The cert (CN=unom, SHA-1 CD1EFDEE…E941, valid to 2036) lives in the
MSIX_CERT_PFX_B64 / MSIX_CERT_PASSWORD Actions secrets; its public half is checked in as
packaging/punktfunk-codesign.cer and published next to each .msix.
pack-msix.ps1 now always exports the signing cert's public .cer (extracted from a supplied pfx too,
not just the ephemeral-generated path) and warns if the cert subject != manifest Publisher (the
mismatch Add-AppxPackage would otherwise reject). Documents the path to a publicly-trusted
(no-import) cert: swap the two secrets + pass a matching -Publisher.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Root cause (confirmed live: "black until I pressed a key, then the image came
back"): the secure desktop (lock/login/UAC) is STATIC, and DXGI Desktop
Duplication only emits a frame on CHANGE. On the normal→secure switch the
duplication is rebuilt (recreate_dupl / try_reduplicate), and we then SEEDED A
BLACK frame as last_present — which the static secure desktop never replaced
(no change-frame) until the user pressed a key. So we streamed black.
Fix: after rebuilding the duplication, CAPTURE the current desktop frame instead
of seeding black. A freshly-created duplication's first AcquireNextFrame returns
the full current desktop; grab it and present it. New `present_acquired` factors
the frame-processing out of `acquire`; both recovery paths now call it:
- recreate_dupl: after adopting the new duplication, acquire+present the real
frame (born-lost ACCESS_LOST / no-initial-frame → seed black as fallback and
let the 250ms-throttled caller retry — a brief flash, then real content).
- try_reduplicate: adopt-first, then capture its probe frame (was discarded).
Also (independently-correct safe fixes, per the adversarial review):
- DesktopWatcher computes the current desktop synchronously in start() before
returning, so a session that begins on the secure desktop (reconnect to a
locked box) doesn't relay one stale normal-desktop frame (the "flash").
- DuplCapturer::open reasserts SudoVDA isolation at open time (mirrors
recreate_dupl) — forces the secure desktop back onto the virtual output if a
lock/UAC re-attached a physical monitor.
- Instrumentation: dbg_black_seeds counter + a throttled warn when black is
seeded, and an info when a real secure-desktop frame is captured on recovery.
Pending: the user's real-lock smoke test on the 4090 (a headless PsExec
LockWorkStation runs as SYSTEM and can't lock an interactive session, so this
must be validated with an actual lock).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The binary had no windows_subsystem attribute, so it linked as a console (CUI) app and Windows
opened a console window alongside the WinUI window on every launch (incl. the MSIX). Add
#![cfg_attr(windows, windows_subsystem = "windows")] so the windowed/MSIX launch is window-free
(verified: the built exe's PE subsystem flips from WINDOWS_CUI=3 to WINDOWS_GUI=2). To keep the CLI
paths usable, main now calls AttachConsole(ATTACH_PARENT_PROCESS) at startup — it binds to an
existing parent console only (never creates one), so --headless/--discover still print to the
launching terminal while Explorer/MSIX launches stay console-free. Adds the Win32_System_Console
windows feature.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Map Apollo's architecture for future agents and compare against punktfunk,
with a deep-dive on the Windows host (the focus area). Produced by the
apollo-vs-punktfunk multi-agent workflow; every claim carries file:line into
both codebases.
Contents: Apollo architecture map + Apollo->punktfunk file index; subsystem
parity; a reference-grade Windows-host deep-dive (DXGI/WGC capture, cursor
compositing, HDR, NVENC-on-D3D11, SendInput/ViGEm, SudoVDA, SYSTEM/secure
desktop); and a prioritized 96-item improvement backlog (89 Windows-host,
24 high-severity). Top confirmed Windows gaps: GameStream TLS accepts any
client cert (verify_client_cert returns assertion()), no NVENC reference-frame
invalidation, SudoVDA watchdog ignores its ioctl result, absolute-mouse mapping
discards the virtual-desktop rect.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Package the Windows client as a signed MSIX (Start tile, clean install/uninstall) and publish it to
Gitea's generic registry, mirroring the host's .deb/.rpm and the Mac's DMG. Validated end-to-end on
the build VM: cargo build --release -> makeappx pack (16 payload files, 58 MB) -> signtool ->
Add-AppxPackage deploy -> framework-dependency resolution all green.
- packaging/AppxManifest.xml: full-trust Win32 app (Windows.FullTrustApplication + runFullTrust),
templated {VERSION}/{PUBLISHER}. windows-reactor packages cleanly despite being built "unpackaged"
because it calls MddBootstrapInitialize2 with OnPackageIdentity_NOOP — under MSIX identity the
bootstrapper no-ops and the App SDK resolves from the manifest's PackageDependency on
Microsoft.WindowsAppRuntime.2 (reactor pins MAJORMINOR 0x20000 = 2.0).
- packaging/pack-msix.ps1: assemble layout (exe + reactor/SDL3 auto-staged DLLs + resources.pri +
FFmpeg DLLs + tile assets), makeappx, signtool. Cert precedence: MSIX_CERT_PFX_B64 secret, else an
ephemeral self-signed cert whose .cer is published alongside (swap in a real cert later, no
manifest change).
- assets: tile/store logos rasterized from packaging/flatpak/io.unom.Punktfunk.svg.
- .gitea/workflows/windows-msix.yml: runs on the Windows runner on main pushes + win-v* tags +
dispatch. MSIX version is 4-part numeric — win-vX.Y.Z -> X.Y.Z.0, else 0.2.<run>.0. shell: pwsh +
CARGO_TARGET_DIR=C:\t like windows.yml.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The injector reattached the input desktop (OpenInputDesktop + SetThreadDesktop,
two syscalls) before EVERY event. Now it stays bound to its desktop and only
reattaches on a SendInput short write (the input desktop switched into UAC/lock)
+ retries once — Sunshine's model. No steady-state per-event overhead; still
follows the desktop across the secure boundary, serving both desktops.
Validated on the RTX 4090 (host as SYSTEM): client-rs --input-test injected for
~6s with no "blocked desktop" errors. Completes all 6 steps of the two-process
secure-desktop build; only a real-UAC user smoke test remains.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A WGC-helper exit (crash, or a console disconnect killing its session) used to
end the stream. Now virtual_stream_relay rebuilds the output + helper and resumes
on the new helper's opening IDR. Rebuild — not respawn-on-the-old-target —
because an abruptly-killed helper leaves the SudoVDA's DXGI output briefly
unresolvable ("no DXGI output for target N yet"), and a console reconnect needs
a fresh output in the new session; `build` (the same path reconfigure uses)
recreates both. Bounded: 500ms backoff per attempt, give up after
MAX_HELPER_FAILS (20) consecutive failures; the counter resets on the first
relayed frame.
Live-validated on the RTX 4090 (host as SYSTEM): force-killed the helper PID
mid-stream → exactly one "WGC helper exited — rebuilt output + helper fails=1" →
the stream recovered and client-rs decoded 645 HEVC Main-10 frames continuously
across the kill (an earlier respawn-on-stale-target attempt storm-failed with
"no DXGI output", which the rebuild fixes).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
With the BOM fixed (shell: pwsh), the build got far enough to compile audiopus_sys, which
does a CMake-from-source build of libopus. The runner's host workdir sits deep under
C:\Windows\System32\config\systemprofile\.cache\act\<hash>\hostexecutor\, so target\debug\build\
audiopus_sys-*\out\build\CMakeFiles\CMakeScratch\TryCompile-*\...\.tlog overran Windows' 260-char
MAX_PATH and MSBuild's tracker failed to create its .tlog (DirectoryNotFoundException -> MSB6003,
"CL.exe konnte nicht ausgeführt werden"). Pointing CARGO_TARGET_DIR at C:\t shortens every nested
build path well under the limit (fixes audiopus_sys + SDL3, both CMake-from-source).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
PUNKTFUNK_SECURE_TEST_PERIOD_MS=N drives a square-wave secure/normal toggle in
virtual_stream_relay (instead of the real DesktopWatcher), to exercise the
mid-session helper↔DDA mux without a live UAC/lock. Gated behind the env var,
in the style of PUNKTFUNK_VIDEO_DROP / PUNKTFUNK_FEC_PCT.
Live-validated on the RTX 4090 (host as SYSTEM): with a 4s toggle the mux
switched secure(DDA)↔normal(WGC relay) cleanly 5× in one session and the client
decoded 308 HEVC Main-10 frames continuously across every switch — the
wait-for-IDR latch held with no decode break. The real Winlogon DDA capture is
pre-proven by the single-process secure path (f4b4a6c); the toggle exercises the
new surface (the mux). Doc updated with the validation + the SYSTEM-mode audio
caveat.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Windows PowerShell 5.1's Out-File -Encoding utf8 prepends a UTF-8 BOM, corrupting the first
GITHUB_ENV line so CARGO_WORKSPACE_DIR silently never got set -> windows-reactor build.rs panic
-> CI build failed (runs 8765/8768). pwsh 7 writes UTF-8 without a BOM. Installed PowerShell 7.6.2
MSI on the runner and put C:\Program Files\PowerShell\7 on the daemon wrapper PATH so jobs find
pwsh; switched all windows.yml steps to shell: pwsh. (Reproduced locally with CARGO_WORKSPACE_DIR
set: the build is green in 2m37s — the BOM was the only issue.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CreateProcessAsUserW gives the spawned helper the *user's* environment block, so
the host's PUNKTFUNK_ENCODER=nvenc (and ZEROCOPY/PERF/…) were dropped and the
helper fell back to the software (H.264-only) encoder — the client negotiated
H265 → "WGC helper exited". `merged_env_block` now parses the user block, strips
any PUNKTFUNK_* it carried, overlays this (host) process's PUNKTFUNK_* vars, and
passes the merged UTF-16 block.
Validated live on the RTX 4090 (host as SYSTEM): the helper spawns via
CreateProcessAsUserW, runs WGC with no hang (HDR FP16 BT.2020 PQ), opens NVENC
(D3D11 Main10), and relays AUs over the pipe — client-rs decoded 411 HEVC
Main-10 frames over the LAN. Step 4 (spawn + relay) complete.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`virtual_stream_relay` now muxes the AU source by input desktop. A DesktopWatcher
(SYSTEM-only Winlogon-name poll) drives it: the user-session WGC helper relay
feeds the normal (Default) desktop; the host's OWN DDA capturer+encoder — opened
lazily on the first secure transition, on the same SudoVDA target with a no-op
keepalive (the host still holds the real isolation owner) — captures the secure
(Winlogon: UAC/lock/login) desktop that WGC can't see. Every switch latches
"wait for IDR" and forces the now-active source to emit a keyframe (the two
encoders keep independent infinite-GOP state, so the client must resume on an
IDR); returning to the helper also drains its stale buffered AUs first.
Reconfigure drops the stale-target DDA; keyframe requests route to the live
source. Send path (FEC/seal/paced-send) unchanged.
Also: wgc_relay gains try_recv (drain on switch-back); open_dda takes dims as
args (avoids a closure borrow of the reassigned cur_mode); the forward! macro
returns bool with `break 'outer` at the call site (no in-macro label hygiene).
cfg-gated windows-only. Live validation (UAC switch over a session) pending.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The SYSTEM host now sources the normal-desktop video from a user-session WGC
helper instead of capturing in-process (WGC won't activate as SYSTEM). New
`capture/wgc_relay.rs`: `HelperRelay::spawn` launches `m3-host wgc-helper` in the
interactive user session via CreateProcessAsUserW (WTSQueryUserToken →
DuplicateTokenEx(TokenPrimary) → lpDesktop="winsta0\\default", CREATE_NO_WINDOW)
with three anonymous pipes — stdout (framed Annex-B AUs → parsed back to
RelayAu), stdin (control: force-keyframe), stderr (helper logs → host tracing).
The host holds the SudoVDA keepalive (sole isolation/topology owner); the helper
captures by GDI name only.
m3.rs: `virtual_stream` dispatches to the new `virtual_stream_relay` when
`should_use_helper()` (running as SYSTEM, or PUNKTFUNK_FORCE_HELPER; disable with
PUNKTFUNK_NO_HELPER). The relay loop feeds the existing send thread — same
FEC/seal/paced-send path. Reconfigure rebuilds the output + re-spawns the helper;
keyframe requests forward over the control pipe; helper pts_ns (same-machine
monotonic clock) is used directly as capture_ns. Disconnect ends the stream
(step 6 adds the relaunch watchdog).
wgc_helper.rs: reads the stdin control byte to request an IDR; --bit-depth flag
threaded through so SDR 10-bit (Main10) negotiation reaches the helper's encoder.
cfg-gated windows-only; Linux/macOS build unaffected. Step 5 (DesktopWatcher mux
to host DDA on the Winlogon secure desktop) is next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Re-add the paths filter (the trigger was never the problem — the runner was registered at the
wrong scope, so org-repo runs found 'no fitting runner' despite the runner showing idle). Document
in setup-windows-runner.ps1 that the registration token must be GLOBAL (Site Administration ->
Actions -> Runners), like the Linux runner. CARGO_WORKSPACE_DIR is set via GITHUB_ENV in a step
(the job-env ${{ github.workspace }} form didn't resolve, leaving it unset -> reactor build.rs
panic).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`m3-host wgc-helper --target-id N --gdi NAME --mode WxHxHz --bitrate K`: the
USER-session half of the two-process secure-desktop design
(docs/windows-secure-desktop.md). Opens WGC on the EXISTING SudoVDA output by
GDI name only (never creates a virtual output — a second topology owner re-trips
the ACCESS_LOST born-lost storm), encodes via NVENC, and ships framed Annex-B
AUs on stdout for the SYSTEM host to relay onto the live QUIC session:
`[u32 magic "PFAU"][u32 len][u64 pts_ns][u8 keyframe][data]`. tracing → stderr so
stdout stays the pure AU stream. cfg-gated windows-only; Linux build unaffected.
scripts/headless/win-build.cmd: the canonical box build script (sets
PUNKTFUNK_BUILD_VERSION so build.rs stamps the version + the NVENC LIB path).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mirror apple.yml's shape — drop the job-level env + defaults blocks; set CARGO_WORKSPACE_DIR
from $GITHUB_WORKSPACE in a step (Gitea can't resolve github.workspace at job-env-eval time)
and use per-step shell: powershell.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The paths filter wasn't dispatching the run on the newly-added workflow (the runner is healthy
and 'declare successfully', but received no task). Match apple.yml: trigger on every push to main
+ PRs. Also set NO_COLOR in the daemon wrapper so runner.log is plain text (the ANSI spinner
garbled it).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Polls the input-desktop name (OpenInputDesktop + GetUserObjectInformationW(UOI_NAME)) on its own
thread → Default/Winlogon atomic; the authoritative normal-vs-secure signal for the capture mux +
input path (WTS notifications miss UAC). Not yet wired into the mux (needs the SYSTEM host + WGC
helper, steps 3-5 in docs/windows-secure-desktop.md). NOTE: detecting the secure desktop requires the
host to run as SYSTEM (a user-token process can't OpenInputDesktop the Winlogon desktop).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
runs-on: windows-amd64 (home-windows-1, host mode). Build + clippy(-D warnings) + fmt + test the
WinUI 3 client. The toolchain is baked into the runner's daemon env; the workflow only sets
CARGO_WORKSPACE_DIR=${{ github.workspace }} (windows-reactor's build.rs needs it). Triggers on
changes to the windows crate / core / Cargo / this workflow.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Validated design for adding secure-desktop (UAC/lock/login) coverage on top of the shipped WGC
animation fix. Key verified constraint: WGC won't activate under SYSTEM (0x80070424) even with
thread-level ImpersonateLoggedOnUser, and DDA+SendInput on Winlogon need LOCAL_SYSTEM — so one
process can't do both. Architecture: SYSTEM host (QUIC + SudoVDA + DDA-secure + SendInput + AU mux)
+ a USER-session WGC helper (CreateProcessAsUser) that relays encoded Annex-B AUs over a named pipe;
the host muxes helper-AUs (normal desktop) vs its own DDA encoder (secure desktop), switched by a
desktop-name watcher. No shared GPU texture (rejected — MIC/keyed-mutex pain); just AU bytes.
docs/windows-secure-desktop.md has the ordered, box-testable steps.
The impersonate_active_user() in wgc.rs is kept as a harmless no-op (under a user-token process
WTSQueryUserToken fails → no impersonation → WGC works natively); it does NOT make WGC work under
SYSTEM (the two-process design uses a real user process for WGC instead). + Win32_System_RemoteDesktop.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two fixes after live setup on home-windows-1: register from $RunnerHome (act_runner writes
.runner relative to CWD, so it must run there — it had landed in the SSH home and the daemon
couldn't find it), and run the daemon under cmd-level redirect (>> runner.log 2>&1) so its native
stderr stays out of PowerShell's error stream. Runner is live: windows-amd64:host, SYSTEM
scheduled task, "declare successfully" against git.unom.io.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
PowerShell 5.1 reads .ps1 in the system code page; an em-dash inside a string literal misparsed
(its bytes look like a quote) and the non-ASCII username in the daemon wrapper would have been
mangled. Drop the em-dash and copy rustup toolchains to C:\Users\Public\.rustup so the wrapper
carries no non-ASCII path. Prep validated: act_runner 1.0.8 + Node 20 + config generated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Windows analogue of scripts/ci/setup-macos-runner.sh: downloads act_runner (gitea-runner)
in host mode, bumps Node 20 via nvm4w (actions/checkout@v4), registers against git.unom.io with
labels windows-amd64:host, and installs a SYSTEM scheduled task that keeps the daemon alive
across reboots. The daemon's env wrapper hard-codes this box's MSVC/WinUI toolchain (cargo/rustup,
NASM, CMake, LLVM, FFmpeg, the ASCII CARGO_HOME SDL3's PCH needs) so the Windows workflow inherits
a working toolchain. Idempotent; token (from org unom -> Settings -> Actions -> Runners) not
persisted.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The capture-architecture reset from the research: add a Windows.Graphics.Capture (WGC) backend that
captures the COMPOSED desktop — including the overlay/independent-flip/MPO planes DXGI Desktop
Duplication misses — which structurally fixes the frozen HDR animations + video (proven live: a WGC
frame decodes to the real 5120x1440 HDR content DDA freezes on). It reuses the whole pipeline
unchanged: the WGC frame's GPU texture → same scRGB→BT.2020-PQ shader → NVENC zero-copy; the OS
composites the cursor (IsCursorCaptureEnabled) so no manual cursor pass. crates/punktfunk-host/src/
capture/wgc.rs; find_output/make_device/HdrConverter/nudge_cursor_onto made pub(crate) for reuse.
Reliability findings + mitigations (live on the RTX 4090):
- WGC can't activate under the SYSTEM account (0x80070424) — it needs the interactive user token. The
host must run as the user for WGC (run.cmd: drop PsExec -s). DDA still needs SYSTEM for the secure
desktop — that token reconciliation (impersonation) is the remaining task.
- WGC's Direct3D11CaptureFramePool::CreateFreeThreaded intermittently HANGS on the headless SudoVDA
(IddCx) display, correlated with accumulated SudoVDA churn (failed REMOVEs leaving lingering
displays); clean-state opens reliably. Since it's a blocking hang, capture_virtual_output runs WGC
open on a watchdog thread with a 5s timeout and falls back to DDA on hang/error — the session is
NEVER left black: WGC when it opens (fixed animations), DDA otherwise. First-frame nudge added (WGC
fires FrameArrived on change; a static desktop otherwise never delivers the first frame).
- Default WGC; PUNKTFUNK_CAPTURE=dda forces DDA. DDA path unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The first cut was a flat stack of buttons. Reworked the chrome to match the windows-reactor
gallery's look:
- Mica backdrop on the window.
- A centred, scrollable, max-width column (`page()` helper) instead of full-width sprawl.
- Card surfaces (`border` + `ThemeRef::CardBackground`/`CardStroke`, rounded, padded) grouping
content, with all-caps section labels.
- Host rows are clickable cards: name (semibold) + address + a PIN/Open/Paired badge + chevron,
laid out with a grid so the badge/chevron sit right; tap to connect.
- Header row with title + Settings button; a ProgressRing while searching / connecting; settings
as grouped "Stream" / "Audio" cards; the pairing screen is a centred card.
Pure styling/layout — no logic change. Build + clippy + fmt green on x86_64-pc-windows-msvc.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The winit-commit docs claimed "Reactor rejected, no SwapChainPanel hatch" — that was wrong.
windows-rs PR #4499 added the SwapChainPanel widget; the client now uses WinUI 3 via
windows-reactor. Update CLAUDE.md M4, the bootstrap-doc status banner (reactor integration:
pinned git dep, CARGO_WORKSPACE_DIR, App-SDK build.rs, LL-hook stream input), and the
docs-site clients page (WinUI 3, launch-and-pick-a-host).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
windows-reactor exposes no raw key-down/up or pointer-position/wheel events (only keyboard
accelerators + pointer button-state), so the WinUI 3 stream page captures input below XAML via
WH_KEYBOARD_LL / WH_MOUSE_LL, installed on the UI thread when the stream page mounts and removed
on unmount (held keys/buttons flushed). The SwapChainPanel fills the window, so the pointer maps
through the client rect (Contain-fit into the negotiated mode); keys carry the native Windows VK
directly (the wire contract — no table needed). While captured, events inside the video area are
swallowed so Alt+Tab/Win reach the host; Ctrl+Alt+Shift+Q toggles capture; clicks on the title
bar (outside the client rect) pass through. Mouse buttons (L/M/R/X1/X2), vertical + horizontal
wheel, and absolute motion all forwarded. Build + clippy + fmt green on x86_64-pc-windows-msvc.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replaces the winit + raw-HWND-D3D11 shell with a native WinUI 3 UI via windows-reactor (a
declarative React-like framework backed by WinUI). The earlier "Reactor can't host a
swapchain" read was wrong — PR #4499 (merged 2026-06-01) added a SwapChainPanel widget with
`set_swap_chain` over `CreateSwapChainForComposition`. Builds + clippy + fmt green on
x86_64-pc-windows-msvc.
- Cargo: drop winit/raw-window-handle; add windows-reactor + the `windows` crate, both pinned
to the SAME windows-rs commit (b4129fcc) so the `IDXGISwapChain1` handed to `set_swap_chain`
satisfies reactor's `windows_core::Interface`. Reactor's build.rs downloads the Windows App
SDK NuGets + stages the bootstrap DLL/resources.pri — it requires `CARGO_WORKSPACE_DIR` set
(now in the VM build env); /temp + /winmd gitignored.
- present.rs: composition swapchain (B8G8R8A8 FLIP_SEQUENTIAL premultiplied) bound to the
SwapChainPanel; WARP fallback, runtime D3DCompile shaders, dynamic RGBA texture, Contain-fit
letterbox; driven by reactor's per-frame `on_rendering`.
- app.rs: the WinUI 3 shell — host list (live mDNS + saved + manual), settings (resolution/
refresh/mic combos+toggle), in-app SPAKE2 PIN pairing screen, and the stream page. Trust gate
mirrors the GTK client (pinned → silent, pair=optional → TOFU, else PIN); a pinned-fp
mismatch routes to re-pair. The session pump + decoded-frame handoff cross to the UI thread
via a Mutex side-channel + thread-locals (the SwapChainPanel sample's pattern).
- gamepad: `ctl` sender now `Arc<Mutex<…>>` so GamepadService is `Sync` (shared across the UI
and session-pump threads). main.rs: windowed = in-app UI; `--headless`/`--discover` keep the
CLI paths.
Not yet wired: raw stream keyboard/mouse input (next commit — reactor exposes no raw key/
pointer events, so it needs Win32 low-level hooks or Microsoft.UI.Xaml bindings). On-glass
validation pending a display (the dev VM is headless/GPU-less).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds the SDL3 gamepad service (near-verbatim port of the GTK client's — SDL3 is
cross-platform) and wires it into the winit app: per-session capture (buttons/axes,
DualSense touchpad + motion 0xCC), feedback (rumble, lightbar, raw DualSense effects),
single-pad-forwarded model with auto pad-type from the physical controller. Built from
source on Windows (no system SDL3).
- gamepad.rs: GamepadService (app-lifetime SDL thread) attach/detach on session
connect/end; auto_pref resolves "Automatic" to the attached pad's type.
- app.rs: hold the service, attach on Connected, detach on Ended/Failed/close. Also
simplify the keydown path (drop the identical if/else arms).
- main.rs: start the service for the windowed path, resolve GamepadPref from settings +
the physical pad.
Build gotcha documented + fixed in the dev loop: SDL3's build-from-source MSVC
precompiled-header chokes on the `ü` in the dev box's username embedded in the cargo
registry path (MSB8084/C4828) — CARGO_HOME must be an ASCII path
(C:\Users\Public\.cargo). Unrelated to our code.
Docs: CLAUDE.md M4 + docs/windows-client-bootstrap.md status banner (winit-not-Reactor
rationale, CARGO_HOME gotcha, what's pending) + docs-site clients.md "Windows desktop
client (in development)". Crate is build + clippy + fmt + test green on
x86_64-pc-windows-msvc.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Builds on the prior headless scaffold (which was committed but never VM-built — its
audio.rs had two non-compiling wasapi calls). This makes the whole crate build + clippy
+ fmt + test green on x86_64-pc-windows-msvc and adds the windowed client.
- Fix audio.rs: `DeviceEnumerator::new()?.get_default_device(...)` (the free fn doesn't
exist) and the 3-arg `write_to_device` (wasapi 0.23). WASAPI shared-mode event-driven
render + mic capture now compile and link.
- present.rs: D3D11 renderer with WARP fallback (GPU-less dev box), runtime-compiled
fullscreen-triangle shaders, dynamic RGBA video-texture upload, Contain-fit letterbox
draw, and a flip-model swapchain on the window HWND.
- app.rs: winit 0.30 ApplicationHandler — present loop + Moonlight-style click-to-capture
input (keyboard via the physical-KeyCode→VK keymap, absolute mouse, wheel, F11), held
state flushed on release/focus-loss.
- keymap.rs: winit physical KeyCode → Windows VK (layout-independent positional mapping,
the analogue of the Linux client's evdev table).
- main.rs: windowed default + `--headless` counting mode, `--discover` (mDNS list),
`--pair PIN` (SPAKE2 ceremony), `--pin HEX`/known-host/TOFU trust, settings-backed
CLI defaults.
UI decision: winit + raw D3D11 (the bootstrap doc's sanctioned fallback), confirmed by a
research pass — windows-rs "Reactor" ships no SwapChainPanel / SetSwapChain escape hatch,
so it can't host the presenter; winit+WARP validates on the GPU-less VM. Native-chrome
host-list/settings GUI + D3D11VA hardware decode + 10-bit/HDR present are follow-ups.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The macOS sessionView branch was missing the .ignoresSafeArea() its iOS/tvOS
siblings have, so in fullscreen the stream was laid out in the safe area below the
notch; the aspect-fit video then scaled down to that smaller area and left black
borders. Add .ignoresSafeArea() so the stream fills the whole display including
behind the camera housing (a thin top-center strip occluded — normal fullscreen-
video behavior); at the display's native mode it's now a 1:1 fill. Inert in
windowed mode and on non-notched displays. NSPrefersDisplaySafeAreaCompatibilityMode
is deliberately not used (it shrinks the whole window with borders on all sides).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GCDeviceHaptics.createEngine returns a CHHapticEngine (the only controller-rumble
API on Apple platforms); starting it spins up CoreHaptics, which looks up the
system audio-analytics daemon over Mach. The App Sandbox denies that global-name
lookup and the framework's precondition turns the denial into a hard crash
("Process is sandboxed but com.apple.security.exception.mach-lookup.global-name
doesn't contain com.apple.audioanalyticsd") the moment a controller's rumble
engine starts.
Add the documented, App-Store-acceptable temporary-exception whitelisting exactly
that one service. Verified embedded into the signed binary (codesign -d
--entitlements) alongside the existing entitlements. macOS-only (iOS/tvOS reject
temporary-exception keys and don't need it). App Store: declare it in App Sandbox
Entitlement Usage Information.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The "broken animations in HDR" was an encode-throughput cliff, not the ACCESS_LOST churn. Measured at
5120x1440@240 HEVC Main10 on the RTX 4090: forced 2-way split-encode = 7.6 ms/frame (~131 fps, well
over the 4.17 ms/240fps budget → choppy), while SINGLE engine = 2.8-3.9 ms/frame (~256-357 fps, fits
240). The split/merge overhead dominates for 10-bit; a single Ada NVENC engine already handles 5K@240
Main10 comfortably. So the split decision now forces DISABLE for Main10 (bit_depth >= 10), keeping the
existing forced-2 only for 8-bit above 1 Gpix/s. PUNKTFUNK_SPLIT_ENCODE still overrides. Added a
split-mode log line.
Validated live on the 4090: encode_us_p50 7.6 ms → 3.9 ms at 5K240 HDR with no env override.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The HDR path produced a constant ACCESS_LOST churn during real desktop activity (window resize /
Start menu / DWM transitions): the duplication keeps getting invalidated but the OUTPUT stays valid
(probe passes — 0 born-lost over 72 rebuilds). The old recovery did a FULL rebuild (new device +
factory) on every loss, which re-inits NVENC + seeds black + was throttled to 4x/s → mostly-frozen,
re-init churn = "broken animations".
Now recovery is tiered (mirrors Sunshine): try_reduplicate() does a fresh DuplicateOutput on the
EXISTING device+output — no new device, so NO encoder re-init, NO black seed, gpu_copy/HDR
textures/last_present kept → frames resume immediately. Only a genuine output loss (secure-desktop
switch) or a dead device (DEVICE_REMOVED/RESET) falls back to the full, throttled recreate_dupl.
Both paths probe the new duplication and reject a born-lost one.
Validated synthetically (1080p60 + 5120x1440@240 HDR): pipeline stable, 0 churn, frames flow. The
real-desktop churn needs live validation (can't synthesize DWM animations). Secure-desktop "UI never
appears in-session" is a separate issue (output gone in-session; only a fresh monitor re-add works) —
still open.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- HDR cursor: sRGB→linear decode + scale to HDR graphics white (PUNKTFUNK_HDR_CURSOR_NITS, default
203 per BT.2408) in the FP16 cursor composite, so it's no longer ~2.5x too dim. SDR path unchanged;
the masked-color (I-beam) inversion blend left unscaled. Cursor cbuffer widened 16→32 + bound to PS.
(Validated live: cursor now correct brightness in HDR.)
- Secure-desktop recovery: recreate_dupl now PROBES the rebuilt duplication with a 50ms
AcquireNextFrame and only adopts it when live (Ok/WAIT_TIMEOUT); a born-lost one (immediate
ACCESS_LOST) is dropped so the caller repeats the last frame + retries. Plus reassert_isolation()
re-detaches physical displays on every recovery (re-routing the secure/HDR desktop to the virtual
output, the delta a fresh reconnect has). NOTE: the born-lost ACCESS_LOST storm in HDR is NOT yet
resolved by these — still under investigation (animations/secure-UI/cursor-trail in HDR remain).
- docs/windows-client-bootstrap.md: handoff for the native Windows Rust client (windows-rs Reactor +
WinUI 3 SwapChainPanel, D3D11VA decode, WASAPI audio, SDL3 input; ports crates/punktfunk-client-linux;
10-bit/HDR present; dev boxes + gotchas).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Priority inversions (Thread Performance Checker): the Apple client drains every
plane on .userInteractive threads (video pump, audio, gamepad feedback) and
connects on a .userInitiated Task, but the connector's producer threads ran at
the default QoS — so a high-QoS consumer parked waiting on a lower-QoS producer.
Pin the connector's producers (outer worker thread, all tokio runtime threads via
on_thread_start, and the data-plane spawn_blocking pump) to .userInteractive on
Apple so they match the consumers. #[cfg(target_vendor = "apple")] helper using
the existing libc dep; no-op off Apple, no Swift-side change (no latency
regression).
GamepadFeedback.swift: the init's MainActor hop captured self implicitly-strong
while the inner $active sink captured it weakly — capture [weak self] in the hop
too (the sink stays weak to avoid the retain cycle).
StreamPump.swift: the @Sendable pump-thread closure captured the non-Sendable
AVSampleBufferDisplayLayer. enqueue/flush are documented thread-safe and only the
pump thread drives it after start(), so assert that with nonisolated(unsafe).
cargo build/test/clippy/fmt green (core + host); xcframework rebuilt; swift build
+ iOS/tvOS targets clean with both warnings gone. Runtime confirmation of the
inversion warnings needs a GUI run under Xcode's Thread Performance Checker.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds true HDR (BT.2020 PQ) and 10-bit (HEVC Main10) streaming, negotiated so an
8-bit/SDR client is never sent a stream it can't decode, plus a robust fix for the
capture losing the stream across a secure-desktop transition.
Protocol (punktfunk-core/quic.rs):
- Hello gains `video_caps` (VIDEO_CAP_10BIT / VIDEO_CAP_HDR), Welcome gains `bit_depth`,
both as optional trailing bytes (back-compat). client-rs advertises 10-bit via
PUNKTFUNK_CLIENT_10BIT; the connector advertises 0 for now (in-band detection drives
the native clients). Regenerated punktfunk_core.h.
Windows host:
- 10-bit Main10: host enables it only when the client advertised VIDEO_CAP_10BIT AND
PUNKTFUNK_10BIT is set; threaded through open_video → NVENC (profile Main10,
pixelBitDepthMinus8).
- HDR: when the captured desktop is scRGB FP16 (R16G16B16A16_FLOAT, HDR on), copy it to
an FP16 surface, composite the cursor there, convert scRGB → BT.2020 PQ 10-bit
(R10G10B10A2) via a shader, and encode HEVC Main10 with the BT.2020/PQ colour VUI
(ABGR10 input). Fixes the freeze + cursor-trail that came from feeding FP16 into the
BGRA path. Reacts dynamically to the HDR toggle.
- Capture recovery: rebuild is now a single NON-BLOCKING attempt, throttled to ~4×/s,
repeating the last good frame between attempts (format-tagged last_present). During a
secure-desktop dwell SudoVDA's output is gone; the old blocking 12 s retry starved the
send loop for seconds so the client timed out and disconnected — now the session stays
fed (frozen) until the desktop returns. Also seeds a black frame on recovery.
Apple client (PunktfunkKit):
- Detects HDR in-band from the stream VUI (PQ transfer function), decodes to 10-bit P010,
and presents via an rgba16Float + BT.2020 PQ CAMetalLayer with EDR; SDR path unchanged.
Switches automatically on a mid-session HDR toggle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The macOS Settings window had outgrown one scrolling pane — split it into a tabbed
preferences window (General / Display / Audio / Controllers / Advanced). Each
settings group is now a shared @ViewBuilder section, so iOS keeps its single
grouped Form and tvOS its pushed-picker layout, each defined once. No setting
moved or dropped.
New statistics-overlay controls (Settings → Display → Statistics): a show/hide
toggle (DefaultsKey.hudEnabled) and a corner picker (HUDPlacement /
DefaultsKey.hudPlacement) — the HUD moves to the chosen corner and aligns its text
to that edge.
A Scene-level "Stream" menu (StreamCommands) carries Show/Hide Statistics (⌘⇧S)
and Disconnect (⌘D). Disconnect moved off the HUD button into the menu so it
survives the overlay being hidden, wired via .focusedSceneValue. On iOS a
material-backed exit chip appears when the HUD is hidden (touch users have no
menu/⌘D); tvOS disconnect is unchanged (Siri-Remote Menu button).
Builds on macOS/iOS/tvOS; swift test green. Adversarially reviewed (8 findings
refuted, 2 minor — the iOS exit-chip contrast fix is included here).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Windows host pegged the GPU 3D engine at ~97% during high-fps desktop streaming — measured (per-
process GPU-engine counters) as OUR process, not DWM. Cause: TWO VRAM->VRAM CopyResource per frame
(dupl->gpu_copy in the capturer, then gpu_copy->nvenc_pool in the encoder), and on Windows D3D11
routes copies to render-target textures through the 3D engine (the DMA copy engine sat idle at 7%),
so at 240 fps they saturate it and contend with a game's own rendering.
Eliminate the second copy: NVENC now registers the capturer's D3D11 texture directly (cached by raw
pointer, the cloned texture kept alive until unregister) and encode_pictures it IN PLACE — no
encoder-owned input pool, no per-frame copy. Safe because the host encode loop is synchronous
(capture -> submit -> poll, where lock_bitstream blocks until the encode finishes), so the capturer
never overwrites the texture mid-encode; documented in the module header in case that ever changes.
2 GPU copies/frame -> 1 (the remaining dupl->gpu_copy is unavoidable; that DXGI surface is transient).
Measured: SM/compute ~10-15% at ~217 fps 5K (was ~20% at only ~48 fps with two copies), 3687 frames
decoded clean. Windows-only; Linux/macOS unaffected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Windows host couldn't sustain high-throughput / high-fps streams — two gaps vs the Linux host,
both found via live RTX 4090 measurement (PERF timing + nvidia-smi per-engine attribution):
- UDP Send Offload (USO). punktfunk-core's UdpTransport sent one packet per `send` syscall on
Windows (send_batch/send_gso were Linux-only), capping throughput at high packet rates. Add a
Windows `send_gso` override using `WSASendMsg` + `UDP_SEND_MSG_SIZE` (the Windows analogue of
Linux UDP GSO) via windows-sys — one syscall segments a coalesced <=512-segment super-buffer to
the connected peer. On by default with auto-fallback (PUNKTFUNK_GSO=0 disables, error latches
off); plugs into the existing paced send path. SO_SNDBUF (32MB) was already cross-platform.
- NVENC 2-way split-frame encoding. A single Ada NVENC session tops out ~0.8 Gpix/s, so 5K@240
(1.77 Gpix/s) took ~8 ms/frame -> a ~125 fps ceiling at high motion (the in-game stutter). Set
NV_ENC_INITIALIZE_PARAMS.splitEncodeMode = TWO_FORCED above ~1 Gpix/s (matching the Linux
libavcodec split_encode_mode path) to use both 4090 encoders — measured ~8 ms -> ~4 ms/frame at
throughput. Env override PUNKTFUNK_SPLIT_ENCODE; init-failure fallback disables it (e.g. H264).
Windows-only paths; Linux/macOS unaffected. Builds clean on x86_64-pc-windows-msvc.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Display name capitalized: app_name (launcher label + permission dialogs) and the connect-screen
header are now "Punktfunk". Package/applicationId/service names stay lowercase.
- Settings: removed the redundant "Done" button (the bottom tab bar is the navigation; system Back
still returns to Connect). Dropped the now-unused imports.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Polish pass on the connect screen.
- Host cards: ElevatedCard with a colored letter-avatar (Apple-contact style), name + address, a
colored status pill (Paired / PIN pairing / Trust on first use), and an overflow menu with Forget
on saved hosts. Tapping a card connects. Unifies the old saved/discovered rows into one HostCard.
- Manual connect moved behind an "Add host" ExtendedFloatingActionButton that opens a
ModalBottomSheet with the Host/Port form (the current M3 pattern) — declutters the list.
- Empty state when there are no saved/discovered hosts; single scrollable column; removed the
"core ABI v2" footer.
- Status bar: enableEdgeToEdge driven explicitly dark (transparent bars + light icons) so the
status/nav bars blend with our always-dark surface instead of showing a black band (the no-arg
edge-to-edge had picked the system light/dark theme).
Verified live (emulator screenshots): cards render with avatars + status pills + Forget menu; the
FAB opens the bottom-sheet form; the status bar blends with light icons.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Live-validated Mac <-> RTX 4090 at the display's native 5120x1440@240:
- Resolution: set_active_mode enumerates the IDD's advertised modes and sets the
requested resolution at the best supported refresh (keeps 5120x1440@240; no more
silent fallback to the 1080p OS default when an exact mode is briefly unavailable).
- Bitrate auto-cap: NVENC init probes and steps the average bitrate down to the GPU's
codec-level max so a high client bitrate connects (matches the Linux host; we do not
split NVENC sessions).
- Mouse cursor: DXGI duplication excludes the HW cursor; capture the pointer
shape/position (GetFramePointerShape) and GPU-composite it before NVENC. Color cursors
alpha-blend; masked-color (the text I-beam) uses an INV_DEST_COLOR inversion blend so
the caret inverts the screen and shows on any background (no black box); monochrome
handled too.
- Secure desktop (lock / login / UAC): run as SYSTEM in the interactive session, follow
the input desktop via SetThreadDesktop, and on the WinSta switch recreate the D3D11
device and re-resolve the virtual output's GDI name from the stable SudoVDA target id
(the name changes across the topology rebuild; the old failure hunted the stale
\\.\DISPLAYn and dropped). ACCESS_LOST / INVALID_CALL / device-removed are recoverable,
and a mid-stream resolution change is followed (capturer + NVENC re-init at the new
size). isolate_displays detaches other monitors so Winlogon renders to the virtual
output. One real session recovered 1012 desktop switches and completed cleanly.
Windows-only backends; Linux/macOS unaffected. Builds clean on x86_64-pc-windows-msvc.
Deployment (windowless SYSTEM launch via PsExec + hidden VBScript) documented in
docs/windows-host.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The signing rollout is confirmed end to end: the latest published RPM (0.2.0-0.ci1089) carries
a header GPG signature (added by `rpm --addsign`) and passed the in-CI `rpmkeys --checksig`
self-verify before publishing (a bad/unsigned build fails that gate and never reaches the
registry). So flip every .repo snippet from gpgcheck=0 to gpgcheck=1 and add the package-signing
public key (served from the generic registry, committed at packaging/rpm/RPM-GPG-KEY-punktfunk) to
gpgkey= alongside the Gitea metadata key — dnf/rpm-ostree imports both. Covers rpm/README,
packaging/README, the bootc Containerfile, and the docs-site bazzite/fedora-kde install pages;
rpm/README's signing section reframed from "dormant/enabling" to active (+ key-rotation notes).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Microphone uplink (client → host's virtual mic, 0xCB) and a cleaner connect screen.
Mic (Rust-heavy, mirrors the audio playback path in reverse):
- crates/punktfunk-android/src/mic.rs: AAudio LowLatency **input** → realtime callback hands
captured f32 to a channel → a worker thread Opus-encodes 20 ms stereo frames (48 kHz, VOIP,
64 kbps) and calls NativeClient::send_mic. MicCapture owns the stream + encode thread (RAII stop).
- session.rs: SessionHandle gains a `mic` slot; nativeStartMic/nativeStopMic JNI (mirror of audio);
stopped in Drop. NativeBridge: the two externs.
- Settings: a `micEnabled` flag + a Microphone toggle in SettingsScreen that requests RECORD_AUDIO
(denied → stays off). StreamScreen starts the mic only if enabled AND the permission is held.
Connect-screen redesign:
- One scrollable Column (was a fixed centered layout that could clip with the new tab bar);
host rows render via forEach (no nested LazyColumn). Colored section labels ("Saved hosts",
"Discovered on the network", "Connect manually"), full-width host cards / fields / Connect button,
a header + subtitle, and a muted footer.
Verified live (emulator pf_phone -> home-worker-2): toggling mic requests RECORD_AUDIO; with it
granted, a session sends mic frames (client "mic: sent=250 … peak=0.439" — real audio) and the host
logs "client datagram stream ended … mic=276". Redesigned screen confirmed via screenshots.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace the ad-hoc screen switching with a Material3 bottom NavigationBar. Two top-level
destinations — Connect (Home icon) and Settings (gear) — persist across tab switches; the
immersive stream view is shown full-screen, outside the bar. Settings is now a tab, so its
button is dropped from the Connect screen.
- app/build.gradle.kts: + androidx.compose.material:material-icons-core (tab icons).
- MainActivity: Screen sealed interface -> Tab enum; App() wraps the tabs in a Scaffold with a
NavigationBar bottomBar (streamHandle != 0 -> StreamScreen full-screen); ConnectScreen drops
the onOpenSettings param + the Settings button.
Verified live (emulator): the bar renders with Connect/Settings; tapping a tab swaps content and
moves the selected indicator; the bar persists on both tabs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The previous CI fix bumped the pinned platform to android-37, but the runner's sdkmanager has no
such package yet ("Failed to find package 'platforms;android-37'"), failing the SDK step before it
could install CMake. Revert to platforms;android-36 (AGP auto-installs the compileSdk-37 platform
during the build, as it did before) while keeping the cmake;3.22.1 package that fixes the libopus
cross-build.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A managed list of known/paired hosts on the connect screen — one-tap reconnect + forget —
and a fix for the discovered-vs-manual trust-key split.
- kit/security: KnownHostStore (replaces the fp-only PinStore) stores KnownHost{address, port,
name, fpHex, paired} keyed by address:port, persisted as JSON in SharedPreferences. So a
discovered and a manually-typed connection to the same host now share ONE trust record (the old
PinStore keyed discovered hosts by the mDNS instance id, manual by host:port — pairing via one
path wasn't seen by the other).
- MainActivity: connect() looks up trust by (address, port); on a successful TOFU or PIN pairing
the host is saved (paired flag set for the PIN path). A "Saved hosts" section lists them (name,
address:port · paired/trusted, fp) with tap-to-reconnect (silent, pinned) and a Forget button.
Verified live (emulator -> home-worker-2): pair -> host appears under "Saved hosts" as paired;
tap -> silent reconnect (new host session, no dialog); Forget -> removed. Trust now shared across
the discovered + manual paths by construction.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This session's push storm refilled the runner to 100% WITHIN the prune timer's 24h window
(it only trims >24h), so a build hit ENOSPC and actions/cache saved a truncated target/ ->
`error[E0463]: can't find crate for shlex` in ci.yml's clippy. Two fixes:
- Bump cargo-target-v2- -> v3- in ci.yml + deb.yml so the poisoned tarball is bypassed (a
suffix bump can't — restore-keys falls back to the old prefix; same as the v1->v2 fix).
- Harden scripts/ci/docker-prune: run HOURLY (was 6h) with a burst guard — if the disk is
still >85% after the normal until=12h trim, prune ALL idle images + build cache (in-use
protected). A fast push-burst can fill 99 GB inside any time window, so the disk-pressure
trigger, not the age filter, is the real backstop. Applied live on home-runner-1 (reclaimed
95%->66%) and checked in.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The android.yml runner installed the NDK but not cmake/ninja, so cargo-ndk's audiopus_sys
(libopus via CMake) failed with "is `cmake` not installed?" — broken since the audio increment
added the libopus dependency. kit/build.gradle.kts prepends $ANDROID_SDK/cmake/3.22.1/bin to
PATH (the same SDK CMake that makes local builds work); install cmake;3.22.1 (cmake + ninja) so
that path exists in CI too. Also pin platforms;android-37 to match compileSdk (AGP auto-installs
it otherwise).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The connect mode was hardcoded to 720p60 — violating the "native client resolution, no
scaling" invariant. Derive the device's real display mode (landscape, long edge = width) and
add a Settings screen to tune the stream, mirroring the Linux/Apple clients.
- crates/punktfunk-android: nativeConnect gains bitrateKbps + compositorPref + gamepadPref
(CompositorPref/GamepadPref wire bytes via from_u8); these were hardcoded Auto/Auto/0.
- app/Settings.kt: Settings (width/height/hz/bitrate/compositor/gamepad; 0 = native/auto) +
a SharedPreferences store + nativeDisplayMode (Display.mode, landscape-swapped) +
effectiveMode + the UI option tables.
- app/SettingsScreen.kt: dropdowns for resolution / refresh / bitrate / compositor / controller.
- MainActivity: App owns the settings + a Settings screen; ConnectScreen resolves the effective
mode (Native = the display), shows it on the Connect button, and threads the prefs through
nativeConnect.
Mic + codec selection deferred (mic uplink isn't wired yet; the decoder is HEVC-only).
Verified live (emulator pf_phone -> home-worker-2): default -> host mode=2400x1080@60 (the
emulator's native display, was 720p); Settings 1920x1080 + 20 Mbps + DualSense -> host
mode=1920x1080, requested_kbps=20000, gamepad=dualsense (host created a UHID DualSense).
Settings persist across screens; pinned reconnect stays silent.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
deb.yml builds the punktfunk-web .output in the rust-ci image, but that image had no bun
(only ci.yml's web/docs jobs use the oven/bun image) -> "bun: not found". Bake bun (+ unzip
for its installer) into ci/rust-ci.Dockerfile, and bootstrap it in the deb web step too so the
job is green against the previous image (docker.yml rebuild lag) — mirroring the rpm.yml fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two CI fixes:
- rpm signing (2nd bug): overriding %__gpg_sign_cmd via --define reached gpg with
%{__plaintext_filename}/%{__signature_filename} UNEXPANDED ("No such file or directory").
Stop overriding it — use rpm's default signer (which expands those correctly) and just set
_gpg_name; a passphrase-less key + loopback in gpg.conf makes gpg sign headless. (Requires a
passphrase-less signing key, as the runbook's %no-protection key is.)
- flatpak: the job runs in fedora:43 which has no node, so actions/checkout (a JS action) failed
with "node: not found". Install nodejs in a plain `run:` step (shell, no node needed) before
checkout. Also scope the heavy flatpak-builder run to client/core/manifest changes (+ tags) so
it stops rebuilding on every unrelated docs/host push (tag pushes still build — paths filters
only branch pushes).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The first signed CI run failed at the Sign step: `%{__gpg} gpg ...` expands to `<gpgpath> gpg ...`,
so gpg got a spurious `gpg` filename arg ("no command supplied", options "not considered"). Dropped
the literal `gpg` → `%{__gpg} --batch ...`. Validated locally: the corrected invocation parses as a
sign command (fails only with "No secret key", which is present in CI). The checksig gate did its
job — nothing published, installs stayed safe.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The dedicated EdDSA signing key (AF245C506F4E4763, "punktfunk packages <packages@unom.io>")
whose private half is now the RPM_GPG_PRIVATE_KEY CI secret. Committing the public half so
clients can fetch it (raw URL) for gpgcheck=1. This push triggers a rpm.yml run that signs
0.2.0~ciN via packaging/rpm/sign-rpms.sh (no longer a no-op); the gpgcheck=1 flip follows once
that signed build is confirmed published.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
TOFU let anyone who could reach the host click "Trust" and stream, which defeats the point
on a LAN. Make SPAKE2 PIN pairing the default and only way to trust a NEW host; TOFU survives
as an explicit HOST opt-in (for fully trusted networks), advertised over mDNS so clients render
their trust UI from the host's policy rather than offering trust on faith.
Contract:
- Host advertises pair=required (default) or pair=optional. pair=required rejects unpaired
clients at the handshake; pair=optional accepts them (TOFU).
- Clients: a pinned host whose fingerprint matches connects silently; a pinned host whose
fingerprint CHANGED forces re-pairing via PIN (no re-trust shortcut); a NEW host is offered
TOFU only if it advertised pair=optional, otherwise PIN pairing is mandatory; a manually-typed
or unknown-policy host is always PIN.
Host (crates/punktfunk-host/src/main.rs):
- m3-host now REQUIRES pairing by default (was open by default). New --allow-tofu opts into
accepting unpaired clients + advertising pair=optional; pairing is always armed (PIN logged at
startup). serve --native was already secure-by-default (serve --open). The mDNS advert and the
accept loop already mapped require_pairing -> pair=required + reject; only the m3-host CLI
default + help text changed.
Clients honor the advertised policy:
- Android (MainActivity.kt): TOFU only for a discovered pair=optional host; manual/unknown -> PIN;
fp-change -> re-pair only (dropped the "Forget & re-TOFU" shortcut).
- Apple (HostDiscovery/SessionModel/ContentView/HostCards/HostStore): new allowsTofu
(pair==optional, distinct from unknown); connect() gates .awaitingTrust on it; unpinned
non-optional hosts route to the PIN sheet; "Forget Identity" re-pairs rather than re-TOFUs.
- Linux (app.rs/ui_hosts.rs/session.rs): ConnectRequest.pair_required -> pair_optional;
initiate_connect routes pinned/fp-changed/optional/else; manual + --connect unknown -> PIN; a
pinned connect rejected on trust grounds re-pairs.
Docs (CLAUDE.md, README.md, docs-site/content/docs/pairing.md): describe the gated model — PIN is
the default, TOFU an explicit opt-in with an impostor warning.
Verified: host cargo check/clippy/fmt clean; Android built + live (emulator -> home-worker-2):
a manual connect now opens the PIN dialog (no Trust button) and the PIN ceremony streams; Apple
swift build clean; Linux clippy -D warnings + fmt clean on the Linux box.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The audit's signing recommendation, scoped to RPM (apt's signed Release metadata already
covers .debs; bootc cosign deferred). packaging/rpm/sign-rpms.sh GPG-signs dist/*.rpm and
self-verifies (rpmkeys --checksig), run from rpm.yml between build + publish.
Safe to ship: the step is a NO-OP (exit 0, unsigned as today) until RPM_GPG_PRIVATE_KEY is
set as a CI secret — so it can't break current CI, and when enabled a bad macro fails loudly
via the in-step checksig rather than shipping bad signatures. rpm/README gains the one-time
enablement runbook (generate a dedicated passphrase-less key, add the secret, publish the
public key, flip gpgcheck=1 only after a signed build lands) and notes step-ca is for TLS,
not OpenPGP (it can't sign RPMs).
Also fixes the rpm/README version staleness the doc review caught: rolling is 0.2.0-0.ciN
(outranks the stray 0.1.1, no pin needed), host releases use host-v* not the client's v*.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Make the host docs match the real distribution path and the actual CLI. Reviewed by a
multi-agent pass (6 editors against one verified fact sheet + an accuracy reviewer); its
findings (a wrong client-Recommends claim, a native-concurrency overstatement) folded in.
- Install front door: new README "Install (host)" method-picker + docs-site/install.md
(+ nav), routing each distro to its package registry; source build demoted to a fallback.
- Registry-first install: ubuntu-gnome/ubuntu-kde now lead with the apt registry (not a
cargo build); bazzite leads with the Gitea RPM registry (was COPR/source). Source builds
moved to an appendix.
- CLI accuracy: serve --native arms pairing from the web console (NOT --allow-pairing, which
with --require-pairing/--max-concurrent is m3-host-only); --open disables mandatory pairing.
host-cli/configuration/pairing/quickstart/troubleshooting corrected; mgmt API documented as
always HTTPS+token. Native host serves one session at a time (extras queue) — not multi.
- Firewall: real ports documented (native UDP 9777 + the ephemeral data port caveat +
GameStream ports) for Debian + Arch (ufw + nftables), not just Bazzite.
- Sync/accuracy: punktfunk-client (GTK4) presented as a shipping client (not "roadmap"),
punktfunk-client-rs as the headless tool; host Recommends punktfunk-web only (not the
client); COPR chroots f43/44; bootc header says Gitea registry not COPR.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- spec: narrow ExclusiveArch to x86_64 — no aarch64 build is produced/published (NVENC is
desktop-NVIDIA), so claiming aarch64 advertised an arch we never ship.
- build-deb.sh: ship punktfunk-kde-session.service (ExecStart repointed to the packaged
run-headless-kde.sh) + host.env.kde, matching the RPM/Arch — the deb README's "mirrors the
Fedora RPM" claim now holds.
- audit.yml: weekly + Cargo.lock-change `cargo audit` over the network-facing crypto dep tree
(RustSec advisories); ignore unfixables via .cargo/audit.toml.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Record the on-box native build path (fast iteration vs build-on-VM):
full MSVC C++ tools incl. CRT libs (a partial VS install → LNK1104;
fix via the GUI, headless setup.exe fails), build from an ASCII path
(non-ASCII username → LNK1201 PDB write fail), nasm/cmake/NVENC import
lib + CMAKE_POLICY_VERSION_MINIMUM. Validated: native build → 720p60
NVENC, 174/174 frames, p50 2.5 ms.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The stale code a default install/upgrade got was a TAG LEAK: deb.yml/rpm.yml shared
`tags: ['v*']` with the Apple-client release.yml, so the v0.1.0/v0.1.1 tags cut to ship
the macOS app ALSO published host packages versioned 0.1.1 — which outranks every rolling
0.0.1~ciN / 0.0.1-0.ciN build in both registries (dpkg/rpm version compares confirm), so
`apt install`/`rpm-ostree install` silently fetched ~99-commits-stale code while the READMEs
claimed auto-tracking. Two fixes:
- Decouple host publishing from Apple `v*` tags: deb.yml/rpm.yml now trigger on `host-v*`
only, so a client tag can never poison the host channel again.
- Bump the rolling base 0.0.1 -> 0.2.0 (deb `0.2.0~ciN`, rpm `0.2.0-0.ciN`): sits ABOVE the
stray 0.1.1 yet BELOW a future 0.2.0 tag, and still climbs monotonically by run number — so
`apt upgrade`/`rpm-ostree upgrade` genuinely move forward. Spec default + build scripts +
PKGBUILD pkgver bumped to match.
Build provenance (so a stale/shadowed host is detectable): build.rs stamps PUNKTFUNK_BUILD_VERSION
(set by CI = the full package version, e.g. 0.2.0~ci120.g802e98d; falls back to the crate version
for a plain `cargo build`) into the binary via rustc-env. Surfaced in `punktfunk-host --version`,
the startup log, and the mgmt /health + /host `version` field (was a hardcoded CARGO_PKG_VERSION).
Deliberately env-driven, not git-derived — the RPM builds from a git-archive tarball with no .git.
Version computed BEFORE the build in deb.yml; the spec %build exports it from %{version}-%{release}
(and gains --locked for reproducibility parity with the .deb path). Validated: plain build reports
0.0.1, env-stamped build reports 0.2.0~ci999.gdeadbee.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M4 Android stage 1 (trust). The client now presents a persistent self-signed identity on
every connect, pins host certs trust-on-first-use, and runs the SPAKE2 PIN pairing
ceremony — parity with the Apple/Linux clients. The Rust connector already exposed this;
this wires it through the JNI + a Keystore-backed Kotlin store + the connect UI.
- crates/punktfunk-android: nativeGenerateIdentity (mint), nativeConnect gains
certPem/keyPem/pinHex (identity + TOFU/pinned), nativeHostFingerprint, nativePair
(SPAKE2). hex32/parse_hex32 helpers.
- kit/security: IdentityStore (AndroidKeyStore AES-256-GCM-wrapped PEM blob; StrongBox
with TEE fallback; four-state load so a decrypt failure never shadow-mints), PinStore
(host-id -> fp-hex in SharedPreferences). obtainIdentity mints once on genuine first run.
- app: ConnectScreen loads/mints the identity, looks up the stored pin, and gates connect
on a trust decision — TOFU prompt (first connect), fingerprint-changed warning, PIN dialog.
- AndroidManifest: allowBackup=false (Keystore keys don't restore; a restored device
re-mints rather than carrying a dead blob).
Verified live (emulator -> home-worker-2, synthetic m3-host):
- identity: host logs the presented client fingerprint; stable across an app restart.
- TOFU: first-connect prompt -> Trust -> pins the observed host fp -> pinned reconnect
skips the prompt.
- SPAKE2: PIN ceremony -> "pairing complete — client trusted" -> auto-connect under
--require-pairing; wrong PIN / host down -> "Pairing failed".
Known follow-up: trust is keyed by mDNS instance id for discovered hosts but by
"host:port" for manually-typed ones, so pairing via one path isn't recognized by the other.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The punktfunk-web management console (pairing + status) shipped only via apt. Extend it
to the other HOST packaging methods, mirroring the Debian punktfunk-web .deb (flatpak is
the client, correctly excluded):
- rpm/punktfunk.spec: new noarch `punktfunk-web` subpackage (the .output bundle + a
/usr/bin/punktfunk-web-server node launcher + both systemd --user units + web-init.sh +
web.env.example), gated behind `%bcond_with web`. OFF by default because building the
Nitro/Node SSR bundle needs `bun`, which a plain rpmbuild / COPR mock chroot lacks. Host
package weak-Recommends punktfunk-web.
- ci/fedora-rpm.Dockerfile: install bun (+ unzip) so the CI builder can build the console.
- rpm.yml: build `PF_WITH_WEB=1` (Prep bootstraps bun to stay green pre-image-rebuild); the
publish loop already globs the new noarch rpm into the registry. build-rpm.sh: `--with web`
when PF_WITH_WEB=1.
- bootc/Containerfile: install from the Gitea RPM registry (which carries punktfunk-web)
instead of COPR — `dnf5 install punktfunk punktfunk-web`.
- arch/PKGBUILD: opt-in `punktfunk-web` split member (PF_WITH_WEB=1 appends it + bun) so a
default makepkg still builds host+client with no JS tooling — matching the spec's bcond.
- docs: packaging/README, rpm/README, copr/README (the no-bun caveat), bazzite/README
(Path B rewritten COPR→Gitea registry), arch/README — enable + journal-password steps.
Reviewed across methods by an adversarial multi-agent pass (rpm/ci/arch/bootc/consistency
lenses, each blocking finding 3x-verified); fixed the two it confirmed real — the Arch
bun-mandatory regression (now opt-in) and the stale COPR wording in bazzite Path B.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M4 Android stage 1 (discovery). Kotlin-only — browse _punktfunk._udp and present a
tappable host list above the manual Host/Port fields.
- clients/android/kit: HostDiscovery — NsdManager browse + resolve (registerServiceInfoCallback
on API 34+ for reliable TXT, legacy resolveService on 31-33), MulticastLock while running, and
a pure parseTxt(proto/fp/pair/id). Exposes the live host set via an onChange callback (NSD
callbacks land on the main thread). DiscoveredHost(name, host, port, fingerprint?, pairingRequired).
+ a JVM unit test of parseTxt.
- clients/android/app: ConnectScreen renders discovered hosts (tap -> fill host/port + connect);
discovery scoped to the screen (start on enter, stop on connect/leave). Manifest adds
CHANGE_WIFI_MULTICAST_STATE + ACCESS_WIFI_STATE (NEARBY_WIFI_DEVICES already declared). Trust
stays TOFU (pin=None); fp shown advisory; pairingRequired shown (SPAKE2 PIN wiring is later).
Verified: parseTxt unit test (5/5 green); on the emulator a loopback NsdManager.registerService of
a fake _punktfunk._udp host was discovered + resolved + TXT-parsed and rendered as a card
(name/host:port/TOFU/fp) -- the full browse->resolve->parse->UI path. Real cross-LAN discovery
needs a physical device on the host LAN (the emulator's SLIRP NAT drops mDNS multicast).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mark DXGI capture + NVENC as live-validated (720p60/1080p60), record the
real-GPU test box (192.168.1.174), the Session-0→Session-1 Interactive
scheduled-task launch, the VM-built-exe-runs-with-driver-DLL trick, and
the SudoVDA-output-under-the-rendering-GPU gotcha. Refresh remaining gaps
(SendInput in-session, ViGEm input/rumble, Moonlight-on-GPU, static-frame
pacing).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Validated live on an RTX 4090 (Windows 11) host streaming to the Rust
reference client over the LAN: SudoVDA virtual display → DXGI Desktop
Duplication (D3D11 zero-copy) → NVENC HEVC → punktfunk/1. 720p60 and
1080p60 both clean (181 / 177 frames, 0 mismatched, p50 1.6 / 3.45 ms
cross-machine), coexisting with Apollo. Two real-hardware bugs the
GPU-less VM couldn't surface:
- DXGI capturer: the SudoVDA virtual monitor's DXGI output is enumerated
under the GPU that *renders* it (the 4090, LUID 0x15df6), NOT under the
SudoVDA "adapter" LUID SudoVDA reports (0x23276). Restricting the output
search to that LUID found nothing → "adapter has no output named
\\.\DISPLAYn". Now search ALL adapters for the GDI name, bind the D3D11
device to whichever adapter exposes it (NVENC then shares that device),
with a settle-retry (the output appears a beat after display creation)
and topology logging.
- native_pairing / apps: keyed config paths off raw $HOME, which a Windows
service/scheduled-task context doesn't set → "HOME unset" hard-fail at
m3-host startup. Route both through gamestream::config_dir(), which falls
back to %APPDATA% on Windows (cert/paired/apps now under AppData\Roaming).
clippy -D warnings + build green on x86_64-pc-windows-msvc (default and
--features nvenc) and Linux (78/78 tests).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The self-hosted runner filled its disk (95%, builds failing on ENOSPC): every CI
push builds a sha-<commit>-tagged Docker image per pipeline, and since those tags
are never dangling a plain `docker image prune` skips them — they piled up to 589
images / ~85 GB plus 18 GB of build cache. Two parts:
- scripts/ci/docker-prune.{service,timer}: a host-level systemd timer (every 6h,
Persistent) that prunes images/build-cache/containers older than 24h — in-use
images stay protected. Checked in (the runner is hand-provisioned and shared
across orgs) and already installed live; reclaimed 89 GB -> 39 GB (95% -> 42%).
- ci.yml / deb.yml: bump the `cargo-target-<rustc>-*` cache key to `-v2-`. The
disk-full build let actions/cache save a truncated target/ (a dep's .rmeta went
missing -> "error[E0463]: can't find crate for pem_rfc7468" while compiling der).
A suffix bump is useless here — restore-keys would fall back to the poisoned
prefix — so the prefix is versioned to force one clean rebuild. cargo-home is
untouched (sources were intact; the failure was a missing build artifact).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Every user needs the console for pairing, so ship it via apt, auto-wired to the
host — no manual bun/env setup. New punktfunk-web .deb (Architecture: all,
Depends: nodejs >= 20 — runs the node-server build under apt-native node, no
bundled bun):
- packaging/debian/build-web-deb.sh: stages web/.output (server + public) + a
/usr/bin/punktfunk-web-server wrapper (node) + the systemd --user units + the
web.env template + docs. Refuses a bun bundle (Bun.serve) as a wrong-preset guard.
- scripts/punktfunk-web.service: --user unit on :3000, EnvironmentFile sources the
host's ~/.config/punktfunk/mgmt-token (the shared bearer) + the generated
web-password; sets PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990 +
NODE_TLS_REJECT_UNAUTHORIZED=0 (loopback self-signed cert). Restart=on-failure
rides out the host-writes-token-first ordering.
- scripts/punktfunk-web-init.service + web-init.sh: --user one-shot that generates
the login password (a .deb postinst runs as root → wrong $HOME) and surfaces it
to the journal.
- build-deb.sh: punktfunk-host now Recommends punktfunk-web (apt pulls it by
default; headless boxes opt out with --no-install-recommends).
- deb.yml: build the web console + smoke-boot it under node (gate the .deb on a
real /login 200) + build-web-deb.sh; the publish loop globs it automatically.
- web/{.env.example,web.env.example}: document the auto-wiring vs a manual deploy.
End state: `apt install punktfunk-host` pulls punktfunk-web; enable both --user
services; the console logs in (password from the journal) and proxies the host's
HTTPS mgmt API with the shared token — zero hand-edited env. Local .deb build +
node smoke-boot verified.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Switch the Nitro build preset from `bun` to `node-server` so the built
.output/server is a standalone HTTP server runnable by apt-native `node`
(validated: `node .output/server/index.mjs` → Listening, /login 200 on node
v25.9.0). This lets the upcoming punktfunk-web .deb depend on `nodejs (>= 20)`
instead of vendoring the bun binary. CI still BUILDS with bun; only the runtime
target changes, and bun still runs a node-server build, so existing
`bun run .output/server/index.mjs` deployments keep working. `vite dev` is
unaffected. Prereq for bundling the web console into the apt install.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The mgmt API already always serves HTTPS (the host identity cert), but on a
loopback bind with no token it ran unauthenticated — any local process could
drive it. Make auth required ALWAYS:
- new mgmt_token::load_or_generate(): token precedence is --mgmt-token > env
PUNKTFUNK_MGMT_TOKEN > persisted ~/.config/punktfunk/mgmt-token > freshly
generated 32-byte hex, persisted 0600 in KEY=VALUE form (so the bundled web
console can source it directly as a systemd EnvironmentFile — one source of
truth). config_dir() made pub(crate).
- parse_serve() resolves the token via load_or_generate() when unset, so a bare
`serve` Just Works with auth on and no operator step.
- mgmt::run() drops the loopback no-token exemption and requires a token;
require_auth()'s unauthenticated fallback now returns 401. The paired-cert
(mTLS) branch is unchanged — Apple client + library auth unaffected.
- web /api proxy: 503 (legible) instead of forwarding an empty bearer.
- tests: test_app/test_app_native default a token, send() auto-attaches the
bearer; blank-token test asserts the new "no token" refusal. 80 pass.
- docs: mgmt module doc + host.env.example reflect always-on auth + auto-gen.
Compiles, clippy/fmt clean, openapi no drift. Part B (bundle the web console into
apt, auto-wired to this token) follows.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M4 Android stage 1 (DualSense feedback, host->client). Two Kotlin poll threads drain the
connector's rumble (0xCA) + HID-output (0xCD) planes via blocking native pulls and render
in Kotlin (Option B — no JNI upcalls, Android APIs stay in Kotlin).
- crates/punktfunk-android: feedback.rs — nativeNextRumble (returns (low<<16)|high, or -1)
+ nativeNextHidout (writes [kind][fields] into a caller's direct ByteBuffer). Ungated; no
new Cargo deps (next_rumble/next_hidout are on the quic feature already).
- clients/android: GamepadFeedback.kt — rumble -> VibratorManager (two-motor amplitude),
HID Led -> lightbar + PlayerLeds -> player LED via LightsManager (API 33+), adaptive
triggers parsed + logged (no public Android API); resolves the connected pad, emulator ->
logged no-op. Started/stopped in the StreamScreen lifecycle (stop + join before nativeClose).
Verified live (emulator -> synthetic host, PUNKTFUNK_TEST_FEEDBACK=1): client received +
decoded the full burst -- rumble low=16384 high=32768, Led r=10 g=20 b=30, PlayerLeds bits=4
player=1, Trigger which=1 mode=0x21 -- matching the host hook exactly. Rendering is a logged
no-op on the emulator (no controller); real haptics/lightbar/player-LED need a physical pad.
Deferred (need a physical DualSense + device enumeration): client->host rich input
(touchpad/motion send_rich_input) and DualSense controller-type negotiation.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M4 Android stage 1 (gamepad). One controller forwarded as pad 0; mirrors the
Linux/Apple gamepad mapping (byte-identical GamepadButton/GamepadAxis events).
- crates/punktfunk-android: 2 JNI fns (nativeSendGamepadButton/Axis) building the
GamepadButton/GamepadAxis InputEvents (flags = pad index 0).
- clients/android: Gamepad.kt — BTN_*/AXIS_* wire constants, KEYCODE_*->BTN_* map, and
an AxisMapper (joystick MotionEvent -> sticks +-32767 +y-up / triggers 0..255 /
HAT->BTN_DPAD_* with on-change gating + release-all reset). MainActivity routes
gamepad-source KeyEvents in dispatchKeyEvent (DPAD only when from a gamepad, so
keyboard arrows still map to VK) and adds dispatchGenericMotionEvent for joystick axes.
Verified live (emulator -> gamescope host, `adb input gamepad keyevent`): host created
the virtual X-Box 360 uinput pad (index=0) and received the gamepad datagrams (input=22).
Axes can't be adb-injected (joystick MotionEvents) -- build/clippy + code-review this
increment; live stick/trigger test deferred to a physical controller. Deferred: device
enumeration/selection, controller-type negotiation, DualSense rich input.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Steam `LibraryProvider` keyed off `$HOME` + Linux paths, so the game
library was empty on Windows. Add Windows discovery: the default Steam
install dirs under Program Files (`ProgramFiles(x86)`/`ProgramFiles`/
`ProgramW6432`), with games on other drives picked up via each root's
`libraryfolders.vdf` — whose Windows values are backslash-escaped, so
unescape `\\` → `\`. The existing root-scan/dedup logic is shared via a
new `steam_roots_existing` helper. The custom store (mgmt JSON CRUD) was
already cross-platform; only Steam auto-discovery was Linux-only.
Not yet covered: a non-default Steam install dir (the registry
`Valve\Steam\InstallPath`). Degrades gracefully — no Steam → empty list.
clippy -D warnings + library tests green on Windows and Linux.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`serve` gave Moonlight clients no audio on Windows: the GameStream audio
stream thread was Linux-only (a non-Linux stub errored). Widen the
stereo path to Windows — the encode/RTP/AES-CBC/hand-rolled-RS(4,2)-FEC
logic is platform-neutral and already live-validated byte-identical on
Linux, and it now runs over the WASAPI capturer + the (already
cross-platform) `opus` crate. The cfg gates go from `linux` to
`any(linux, windows)`; only the surround path stays Linux-only because
its libopus *multistream* encoder needs `audiopus_sys` (a Linux dep) —
on Windows a surround request fails cleanly with a "use stereo" error.
Linux stays byte-identical (the `SessionEncoder::Surround` variant and
its match arm keep `#[cfg(linux)]`, so Linux compiles exactly as before).
Verified: clippy -D warnings + host test suite green on both
x86_64-pc-windows-msvc (73/73) and Linux (78/78).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The mgmt API serves HTTPS with the host's self-signed identity cert and requires
mTLS-or-bearer auth (the mTLS work), but the web console's proxy still defaulted to
`http://127.0.0.1:47990` — so a deployment copying .env.example got a plain-HTTP
request to an HTTPS port (→ 502 Bad Gateway, observed live on the Bazzite box).
- .env.example + server/util/auth.ts + vite.config.ts: default PUNKTFUNK_MGMT_URL to
https://127.0.0.1:47990.
- vite dev proxy: `secure: false` (the host cert is self-signed).
- Document that the deployment needs PUNKTFUNK_MGMT_TOKEN (matching the host's) and
NODE_TLS_REJECT_UNAUTHORIZED=0 — the web server's only outbound TLS is the loopback
hop to the host's own self-signed cert, so disabling verify there is scoped + safe.
The running Bazzite box is already fixed live (web.env → https + token + cert-skip,
verified: login 200, /api/v1/status 200). This makes fresh deployments correct.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M4 Android stage 1 (input). Kotlin captures input and forwards it over JNI to
NativeClient::send_input (the connector is linked as a Rust crate).
- crates/punktfunk-android: 4 JNI send fns (pointer move / button / scroll / key)
building InputEvent with the GameStream wire codes — ungated, &self on the Sync
connector (safe from the UI thread).
- clients/android: Keymap.kt (Android KEYCODE_* -> Windows VK, the host's wire
contract, mirroring the Linux/Apple tables); Activity-level dispatchKeyEvent forwards
hardware keys to the active session (above the Compose focus system, so it's reliable);
a Compose touch-trackpad overlay -- 1-finger drag -> relative move, tap -> left click,
2-finger drag -> scroll.
Verified live (emulator -> gamescope host on the LAN box, synthetic `adb input`): host
received 31 input datagrams (input=31) and libei injected KeyDown/KeyUp, MouseButtonDown/Up
and MouseMove all emitted=true. Physical-mouse pointer capture + gamepad are next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Windows host test suite hit two pre-existing portability failures
(the autonomous Windows bring-up never ran `cargo test` on the VM):
- `vdisplay::detect_active_session_*` asserted a non-empty XDG runtime
dir — a Linux concept with no Windows equivalent. Gate just that
assertion to Linux (keep the call so the fn stays used → no dead_code).
- `mgmt::openapi_document_is_complete_and_checked_in` did a byte compare
against the checked-in spec, which git may check out CRLF on Windows
while serde_json emits LF. Compare content with `\r` stripped.
Host suite now 73/73 on x86_64-pc-windows-msvc; Linux unchanged (78 ok).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wire the host→client rumble path on Windows, the analogue of the Linux
uinput EV_FF read loop: a game's force-feedback on the virtual Xbox 360
pad is delivered by ViGEm's notification API (`request_notification` →
`spawn_thread`, gated by the crate's `unstable_xtarget_notification`
feature). A per-pad background thread stores the latest motor levels;
`pump_rumble` relays changes to the client on the universal 0xCA plane
(motors scaled 0..255 → 0..65535). Dropping the target aborts the
notification, so the thread exits with the session. Live verification
still needs a physical pad.
Also fix the Windows backends' clippy debt — these modules are cfg-
excluded from Linux CI, so `clippy -D warnings` never saw them, and the
VM's rustc 1.96 clippy is stricter on shared code than the CI image:
- dxgi: manual checked division → checked_div().map_or
- sendinput: `x = x | y` → `x |= y`
- sudovda: `.then(|| ptr)` → `.then_some(ptr)`
- m3 pick_compositor: drop the needless early return (match form)
- m3 resolve_compositor: Windows arm is a tail expr, not `return`
All Windows backends now build + clippy clean (default and --features
nvenc); Linux unaffected (fmt/clippy/check green).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
"Controller disconnected every few seconds" (Forza Horizon, held steady): the
virtual UHID DualSense emitted HID report 0x01 ONLY on state change, but a real
DualSense streams it continuously (~250 Hz). When the player holds the
wheel/throttle steady the client sends no wire events, so the host wrote nothing
and /dev/uhid went silent for seconds — the kernel hid-playstation driver / Proton
/ SDL treat that as an unplugged controller. (The uinput X-Box pad is immune:
evdev holds last-known state with no periodic-report requirement.)
Add DualSenseManager::heartbeat(max_gap): re-emit each live pad's CURRENT report
when it's been silent for max_gap (idempotent — a stale-but-correct frame, never a
phantom input; write_state bumps seq+timestamp). write() resets the per-pad timer,
so an actively-used pad emits no extra reports — the heartbeat only fills genuine
silence. PadBackend::heartbeat() drives it at an 8 ms gap (~125 Hz) for DualSense
(no-op for X-Box), called every input-thread tick (the loop already runs ≤4 ms).
GET_REPORT feature replies + the pad lifecycle were ruled out by the investigation
(pad is created once, never torn down mid-session). Compiles, clippy/fmt clean, 78
host tests pass. Verify on the box: held-idle DualSense stays present in evtest /
no SDL CONTROLLERDEVICEREMOVED; Forza no longer toasts "controller disconnected".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The DualSense intermittently showed up as an Xbox 360 pad on the host: the
client's `.auto` gamepad-type resolution read `GamepadManager.active`, which is
populated only by the async `.GCControllerDidConnect` notification (or the
init-time snapshot). At connect time `active` could still be nil with a DualSense
attached, so the client sent `.auto` and the host's pick_gamepad mapped that to
Xbox 360. Confirmed live: same box, two connects minutes apart logged
`gamepad="xbox360"` (auto) vs `honoring client gamepad request gamepad="dualsense"`.
resolveType() now calls rebuild() first to re-read GCController.controllers()
synchronously before reading `active`, closing the race for the common case
(controller attached before connecting).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The `m3` audio_thread (desktop capture → Opus 48 kHz stereo 5 ms CBR →
AUDIO_MAGIC datagrams) now runs on Windows, fed by the WASAPI loopback
capturer. The `opus` crate vendors libopus via `audiopus_sys` + cmake
(no system lib / vcpkg), so it builds on MSVC — moved into a
`cfg(any(linux, windows))` deps table and widened the audio_thread cfg
to match (the stub now only covers other targets, e.g. macOS).
Build note: CMake 4 rejects libopus's old `cmake_minimum_required`;
set `CMAKE_POLICY_VERSION_MINIMUM=3.5` when building the host on Windows.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M4 Android stage 1 (audio). An audio thread pulls Opus packets from the connector
(next_audio), decodes to interleaved f32 stereo, and feeds AAudio via its realtime
data callback through a jitter ring ported from the Linux client (prime ~3 quanta,
drop-oldest cap, re-prime on drain). All in Rust on native threads — symmetric with
the video decode path.
- crates/punktfunk-android: audio.rs (Opus decode + jitter ring + AAudio callback);
SessionHandle gains an audio slot; nativeStartAudio/nativeStopAudio JNI; Drop stops it.
Android-only deps: opus 0.3 (libopus via cmake, static) + ndk "audio" (AAudio) — pure
C/NDK, no libc++_shared to bundle.
- clients/android: NativeBridge start/stop audio, called in the SurfaceView lifecycle.
- kit/build.gradle.kts: cargo-ndk env for the libopus cmake build (NDK root, Ninja,
LIBOPUS_STATIC/NO_PKG) + --platform 31 (libaaudio is API 26+).
Verified live (emulator -> gamescope host on the LAN box): AAudio opened 48k/stereo/f32;
a 440 Hz tone played into the host capture sink reached the client decoded -- opus ~200/s,
pcm_frames climbing in lockstep, peak=0.089 (real content, not silence), with video
streaming concurrently. Some underruns under emulator jitter (verify on hardware).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Both landed in 3363576 and validated live on the Bazzite F44 box: a Gaming→Desktop
mid-stream switch shows `settled desktop portal env … compositor=kwin` →
`portal granted devices` → `device RESUMED` (input lands, no reconnect), and
`KWin: streamed output set as the sole desktop also_disabled=["HDMI-A-1"]` (panels
on the streamed screen). Remaining: #1 (F44 gamescope teardown GPU leak) + the
lower-priority polish.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two parked follow-ups from the session-aware host work:
#3 — KWin/Mutter virtual output not set primary. The auto-detected desktop path
*is* "stream this desktop", but the per-session virtual output wasn't promoted to
primary, so KDE/GNOME panels + windows stayed on an unstreamed real output and the
streamed screen showed only wallpaper. apply_session_env now defaults
PUNKTFUNK_KWIN_VIRTUAL_PRIMARY / PUNKTFUNK_MUTTER_VIRTUAL_PRIMARY on for the
auto path (explicit config still wins), so the streamed output becomes the sole
desktop.
#2 — input flaky after a mid-stream Gaming->Desktop switch. The xdg portal
(D-Bus-activated) and the systemd --user env still pointed at the old session, so
the host's RemoteDesktop portal opened against a half-stale env: it accepted
events but they didn't reach the compositor until a reconnect. New
vdisplay::settle_desktop_portal() pushes the live session env into the
systemd/D-Bus activation environment and (for KWin) restarts the portal so it
re-reads it, mirroring a fresh desktop login (and the existing wlroots portal
restart). Called from the mid-stream switch rebuild slot before the injector
reopens. GNOME uses Mutter's direct EIS, so it only gets the env push.
Compiles, clippy/fmt clean, 78 host tests pass. Live validation on the Bazzite
box next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Windows GamepadManager via vigem-client (ViGEmBus) — the uinput-xpad analogue: one virtual Xbox 360 controller per client pad index, created lazily on first State. GameStream/Moonlight already uses the XInput conventions (low-16 button bits, sticks -32768..32767 +Y up, triggers 0..255), so the GamepadFrame->XGamepad mapping is 1:1. Replaces the non-Linux GamepadManager stub (same new/handle/pump_rumble API the m3 PadBackend drives, so no m3 change). Graceful when ViGEmBus is absent (gamepad disabled, session continues). Compiles clean on Windows + Linux; live-test needs the ViGEmBus driver + a physical pad. Rumble back-channel is a TODO (ViGEm notification API).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Zero-copy capture->encode on the GPU via the raw NVENC API (nvidia_video_codec_sdk sys + ENCODE_API; the safe wrapper is CUDA-only). Opens an NV_ENC_DEVICE_TYPE_DIRECTX session on the SAME ID3D11Device as the DXGI capturer (carried on the new FramePayload::D3d11), registers a pool of BGRA textures once, CopyResources each captured texture in and encode_picture; CBR/ULL, infinite GOP, P-only, forced-IDR for RFI. The DXGI capturer gains a D3D11 zero-copy output (selected, like the encoder, by PUNKTFUNK_ENCODER=nvenc) so capture+encode share textures.
OFF by default (the nvenc feature pulls the NVENC SDK + cudarc): the default Windows host links without it (openh264 path). cudarc builds toolkit-less via the SDK ci-check feature (dynamic-loading). At link time --features nvenc needs nvencodeapi.lib (NVENC SDK, or an import lib generated from the driver's nvEncodeAPI64.dll) on PUNKTFUNK_NVENC_LIB_DIR. Both default and --features nvenc builds validated to compile+link GPU-less on the VM (import lib generated from the driver DLL). Runtime needs a real NVIDIA GPU.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The punktfunk/1 control plane already compiled on Windows; these wire the last gaps so the host actually runs: config_dir falls back to %APPDATA% (HOME\.config when set), paired_path uses it, hostname from COMPUTERNAME, and resolve_compositor short-circuits the Linux session-detection on Windows (SudoVDA is the single backend; vdisplay::open ignores the compositor arg). Validated live on the VM: m3-host creates its identity, binds the QUIC endpoint (fingerprint logged), advertises mDNS (_punktfunk._udp, host from COMPUTERNAME), and accepts sessions. GPU-less validations green: m0 synthetic->openh264->core FEC loopback (120/120, 0 mismatches) and the m3 c_abi_connection_roundtrip control-plane test. Full session capture (SudoVDA->DXGI) + NVENC remain GPU-gated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Windows Capturer via DXGI Desktop Duplication: create a D3D11 device on the SudoVDA adapter (by LUID), find the matching output (by GDI name), DuplicateOutput, and per AcquireNextFrame copy the desktop into a CPU-readable staging texture -> tightly-packed BGRA (FramePayload::Cpu, feeds the openh264 software encoder GPU-lessly). Handles WAIT_TIMEOUT (reuse last frame) and ACCESS_LOST (re-duplicate). Adds FramePayload::D3d11(D3d11Frame) for the future NVENC zero-copy path, and a VirtualOutput.win_capture identity (adapter LUID + GDI name) carried out of the SudoVDA backend. Pure helpers (pack_luid/gdi_name_matches/depad_bgra) unit-tested on the VM; the live duplication path needs a real GPU + an activated SudoVDA monitor. Compiles clean on Windows + Linux.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Windows AudioCapturer via the wasapi crate (0.23): loopback the default render endpoint (Render device + Direction::Capture + shared mode => STREAMFLAGS_LOOPBACK) at 48 kHz stereo f32 with autoconvert, feeding the existing Opus path with no resampling. Dedicated COM-MTA thread owns the !Send WASAPI objects; interleaved f32 chunks leave over a bounded lossy channel; RAII Drop stops + joins. Bring-up handshake reports a missing endpoint as Err so a session continues without audio. open_audio_capture Windows factory arm + module. Init chain validated live on the VM (open succeeds; next_chunk waits on a silent system). Virtual mic deferred (no Windows virtual-audio endpoint). m3 audio_thread wiring + opus hoist land with the integration task.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Windows Encoder impl via the openh264 crate (statically-bundled, BSD-2): low-latency screen-content config (Baseline/no-B-frames, bitrate RC, BT.709 limited, near-infinite GOP + forced-IDR recovery via request_keyframe), packed CPU pixels (BGRx/BGRA/RGB/RGBA/RGBx/BGR) -> I420 -> AnnexB with in-band SPS/PPS each IDR. Synchronous: submit encodes immediately, poll hands back the one AU, flush is a no-op. Windows open_video factory selects it (PUNKTFUNK_ENCODER=software|nvenc|auto; NVENC arm lands later), H.264-only with a clear error otherwise, SW bitrate ceiling. Unit-tested live on the VM: synthetic BGRx -> AnnexB IDR + SPS NAL. Unblocks the GPU-less capture->encode->FEC->send pipeline. Compiles clean on Windows + Linux.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Windows InputInjector via SendInput (Win32 KeyboardAndMouse), mirroring the wlroots backend: absolute mouse (MOUSEEVENTF_VIRTUALDESK normalized to the virtual desktop), relative mouse, scancode keyboard (MapVirtualKeyExW + extended-key flagging), scroll (no sign flip — Windows wheel matches GameStream), buttons. Client already sends Windows VK codes (no keycode table). Reattaches the thread to the input desktop (OpenInputDesktop/SetThreadDesktop) to survive UAC/lock switches. New Backend::SendInput, the Windows auto-default in default_backend(), open() arm, windows-crate features. Compiles clean on Windows + Linux. Live injection validates with the in-session host run (SendInput is desktop-isolated from an SSH network logon).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Windows VirtualDisplay backend driving SudoVDA (the Apollo IDD) via its DeviceIoControl IOCTL protocol: open by interface GUID, ADD at the client's exact WxH@Hz (mode baked into the IOCTL, no EDID seeding), mandatory watchdog ping thread, QueryDisplayConfig name resolution, RAII Drop -> REMOVE. Wired behind the existing VirtualDisplay trait (open()/probe() Windows arms). Validated live on the GPU-less VM (standalone + via the trait, env-gated test): version 0.2.1, ADD 1920x1080@60 -> target, watchdog hold, REMOVE. Monitor activation into a WDDM path (-> capturable \\.\DisplayN) needs a real GPU and is deferred with capture/NVENC. docs/windows-host.md updated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M4 Android stage 1 (video). Pull HEVC access units from the connector and render
them to the SurfaceView entirely in Rust (NDK AMediaCodec → ANativeWindow) — no
per-frame JNI, honoring the native-thread hot-path invariant.
- crates/punktfunk-android: decode.rs (one-in/one-out AMediaCodec loop; in-band
VPS/SPS/PPS so no out-of-band csd; dims from NativeClient::mode). SessionHandle
now holds an Arc<NativeClient> + the decode thread; nativeStartVideo/nativeStopVideo.
- clients/android: connect screen (host/port) + full-screen SurfaceView stream
screen — surfaceCreated -> nativeStartVideo, leaving -> stop + close.
Verified live (Android emulator -> m3-host on the LAN box, ABI v2): QUIC handshake,
8-round clock-skew sync, HEVC decoder configured at 1280x720, and the data plane
delivered + fed all 299 access units (the punktfunk/1 NAT hole-punch worked through
the emulator's SLIRP). Real-pixel render is pending a non-synthetic source:
`m3-host --source synthetic` emits dummy transport payloads (not HEVC), so the
decoder correctly produces nothing; `--source virtual` (a compositor on the host)
is needed to verify decode-to-screen.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Capture the deliberately-parked items after live-validating the session-aware
backend selector on the Bazzite F44 box (Desktop KDE + Gaming both at the
client's resolution, warm reuse, Feature B mid-stream switch both directions).
Top follow-ups: (1) F44 gamescope teardown corrupts the GPU context (try SIGKILL
teardown, else keep the managed session warm); (2) mid-stream-switch input is
flaky until a reconnect (portal opens before the systemd/D-Bus activation env
settles — fix: import-environment on switch); (3) the KWin virtual output isn't
set primary. Plus polish: input-loss window on switch, the recovered NVENC
invalid-param log, the 4090 HEVC ~800Mbps cap, restore-guard/keep-warm
interaction, and promoting Feature B from opt-in to default.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
flatpak-cargo-generator.py (master) imports `tomlkit` + `aiohttp`; the workflow
installed `python3-toml`, so the "Generate offline cargo sources" step would fail
with ModuleNotFoundError. Install python3-tomlkit instead, and correct the same
note in build-flatpak.sh.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ship the punktfunk Linux client to the Steam Deck as a Flatpak — the only viable
SteamOS install path, since /usr is read-only and lacks libadwaita/SDL3 — and
publish both it and the Decky plugin through Gitea. Built and validated live on a
Steam Deck (SteamOS 3.7): bundle installs user-scope, all libs resolve, libavcodec
resolves to the codecs-extra HEVC build, devices=all for DualSense hidraw.
packaging/flatpak (new):
- io.unom.Punktfunk.yml on GNOME 50 / freedesktop-sdk 25.08. rust-stable//25.08
(rustc 1.96 — the GTK4 chain needs >=1.92; the EOL GNOME-48/24.08 rust-stable at
1.89 could not build it) + llvm20 (libclang for bindgen in ffmpeg-sys-next/sdl3-sys).
HEVC libavcodec comes from the runtime's auto codecs-extra extension point (no
app-side codec declaration). Bundled SDL3 3.4.10 (matches sdl3-sys 0.6.6+SDL-3.4.10).
finish-args: wayland/fallback-x11, --device=all (GPU/VAAPI + evdev + hidraw — flatpak
cannot bind /dev/hidrawN char devices via --filesystem), pulseaudio, network,
~/.config/punktfunk.
- metainfo.xml, desktop, square SVG icon, build-flatpak.sh (offline cargo-sources;
on-Deck org.flatpak.Builder or CI), README.
clients/decky:
- add LICENSE (MIT), fix package.json license (BSD-3-Clause -> Apache-2.0 OR MIT),
add scripts/{package.sh,deploy.sh} (the plugins dir is root-owned: stage to /tmp,
sudo install, restart plugin_loader), align the launcher fallback to the real
flatpak app id io.unom.Punktfunk, rewrite the install section.
.gitea/workflows:
- flatpak.yml: privileged Fedora container builds the bundle and publishes to the
Gitea generic registry (+ release attachment on tags).
- decky.yml: pnpm build -> store-layout zip -> registry (stable latest/ URL for
Decky "install from URL").
docs: packaging/README + packaging/flatpak/README.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rust-heavy client model (like punktfunk-client-linux): a new cdylib crate
crates/punktfunk-android links punktfunk-core and exposes the JNI seam;
Kotlin (clients/android) owns only the Android-framework surface. Kotlin can't
import the C header the way Swift can, so the bridge is written in Rust to reuse
the Linux client's orchestration rather than re-port it.
- crates/punktfunk-android: JNI bridge — abiVersion/coreVersion native-link
proof + session connect/close handle; plane pumps stubbed for M4 stage 1.
- clients/android: Gradle project — :app (Compose) + :kit (Android library with
a cargo-ndk Exec task -> jniLibs). AGP 9.2 / Gradle 9.4.1 / Kotlin 2.3.21 /
Compose BOM 2026.05.01 / compileSdk 37 / targetSdk 36 / minSdk 31, shipping
arm64-v8a + x86_64. Phone + TV (leanback) installable. README rewritten.
- .gitea/workflows/android.yml: CI mirroring apple.yml on a Linux runner.
- punktfunk-core: switch rcgen to the ring backend so the whole quic tree is
aws-lc-free (smaller client .so, cmake-free cross-compile; a win for all targets).
Validated on this box: :app:assembleDebug -> APK with both ABIs; emulator
first-light renders the bridge linked (core ABI v2) with logcat confirmation;
clippy -D warnings + cargo fmt clean; core tests green on the ring backend.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rewrite the scoping doc into a concrete implementation plan: locked decisions (host-first, SudoVDA virtual display, pure-Rust windows-rs+Reactor client linking core directly, FFmpeg/D3D11VA decode), the SudoVDA IOCTL control protocol, the no-GPU dev strategy, the Windows-specific structural issues (interactive session, clock epoch, no IDD audio), and the phased plan. Step 0 (compile on MSVC) marked done.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Gate the Linux-only bits so the host crate builds on MSVC (it already built on Linux + macOS): drm_sync/dmabuf_fence use DRM ioctls + libc (a linux-only target dep) and have no non-Linux callers; VirtualOutput.remote_fd is a PipeWire concept. The full dep tree (aws-lc-rs, quinn, rusty_enet, axum) builds clean on MSVC and the binary runs (openapi emits the spec) — only these 3 cfg-gates were needed. First step of the Windows host port (docs/windows-host.md).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two fixes from live Bazzite testing of the managed-Gaming + mid-stream work:
1. Input now FOLLOWS the active session. The host-lifetime injector was pinned to
the first backend it opened and only reopened on an inject FAILURE — but with
Feature A keeping the managed gamescope warm, its EIS socket stays alive, so a
switch to the KDE desktop + reconnect kept injecting into the idle gamescope
(input silently dead on KDE). injector_service_thread now compares the
resolved input backend (default_backend() ← PUNKTFUNK_INPUT_BACKEND, set per
connect by apply_input_env, and on a mid-stream switch) each event and reopens
when it changes. Fixes input on a Gaming->Desktop reconnect AND Feature B's
mid-stream input re-route, with no plumbing.
2. Debounced TV-restore no longer yanks you back to gaming. do_restore_tv_session
now checks detect_active_session(): if a desktop session is active (the user
switched), it tears down the idle managed gamescope but does NOT restart the
gaming autologin. Observed live: the restore fired and restarted
gamescope-session-plus@ogui-steam while the client was already on the KDE
desktop.
Also: document PUNKTFUNK_SESSION_WATCH (Feature B opt-in) in the Bazzite host.env
and correct the managed-default description. Compiles, clippy/fmt clean, 78 tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The schedule_restore_tv_session assignment exceeded 100 cols; rustfmt wraps it.
The fix was made post-commit but only m3.rs was staged for 95a820b, so CI's
fmt --check failed on the committed unwrapped line. Stage the wrap.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Feature B: while streaming, follow a Gaming<->Desktop switch on the box without
a reconnect. A ~1s watcher thread (session_watcher_loop) self-baselines on the
live ActiveKind and, when it changes and stays changed for a 3s debounce (the
old/new compositors coexist briefly during a switch), sends a SessionSwitch to
the encode loop. The loop's new rebuild slot — taking precedence over a queued
mode change — retargets the process env (apply_session_env/apply_input_env) and
rebuilds the WHOLE backend in place at the SAME client mode (vdisplay::open +
build_pipeline_with_retry), reusing the proven mode-switch rebuild path: the
Session + send thread (QUIC control + UDP data plane + side planes) stay up, the
client sees a brief freeze then an IDR. Old pipeline kept on a rebuild failure
(transient vs permanent classified via is_permanent_build_error). Input
re-routes via the host-lifetime injector's lazy reopen against the new
PUNKTFUNK_INPUT_BACKEND.
Opt-in via PUNKTFUNK_SESSION_WATCH (off by default; never under an explicit
PUNKTFUNK_COMPOSITOR pin), so it lands inert and is promoted to default only
after live validation on a real Bazzite Gaming<->KDE flip. The watcher snapshots
the SessionEnv so only the encode thread writes process env.
Compiles, clippy/fmt clean, 78 host tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Feature A: in Gaming Mode, default to a host-managed gamescope at the CLIENT's
mode (tear the TV's autologin down on connect) instead of attaching to the
running TV session — so the client receives ITS resolution (capture == encode ==
client mode, fixing the InitializeEncoder size mismatch the attach path hit),
not the TV's 4K.
Reliability is the debounce: restore_managed_session() now SCHEDULES the TV
restore RESTORE_DEBOUNCE (5s) after the last disconnect via a host-lifetime
worker, instead of restoring immediately per-disconnect. A reconnect inside the
window cancels the pending restore and reuses the still-warm managed session
(create_managed_session clears PENDING_RESTORE at the top) — so a quick reconnect
(e.g. a controller hiccup) never triggers a gamescope stop/relaunch, which is the
per-connect churn that leaked NVIDIA GPU context on F44 (the black-screen
reconnect).
- vdisplay/gamescope.rs: PENDING_RESTORE + RESTORE_DEBOUNCE; schedule_restore_tv_session
(debounced), do_restore_tv_session (the actual restore, worker-driven),
start_restore_worker (100ms tick, RAII keepalive handle). create_managed_session
cancels the pending restore + reuse path unchanged.
- vdisplay.rs: apply_input_env flips gamescope to managed-DEFAULT; PUNKTFUNK_GAMESCOPE_ATTACH
(or an explicit _NODE) opts back to attach for couch-on-TV; _MANAGED forces managed.
restore_managed_session schedules; new start_restore_worker wrapper.
- m3.rs serve(): hold the restore worker for the host lifetime.
- bazzite host.env: document managed-default + the ATTACH opt-out.
Compiles, clippy-clean, 78 host tests pass. F44 single stop/start leak to be
verified live on the box.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The session-aware selector drives a KWin virtual output at the client's
resolution when the Bazzite box is in KDE Desktop Mode — validated live. But a
normal KDE login withholds two things the headless host needs:
1. KWIN_WAYLAND_NO_PERMISSION_CHECKS=1 — so KWin exposes the privileged
zkde_screencast virtual-output protocol to an external client.
2. the kde-authorized RemoteDesktop grant — so libei input auto-approves
instead of popping a dialog a headless host can't answer.
Add packaging/bazzite/kde-desktop-setup.sh (idempotent, no root): writes the
environment.d KWIN drop-in and seeds the grant DB (shipped at
/usr/share/punktfunk/headless/kde-authorized) into ~/.local/share/flatpak/db/,
restarting the portal chain. Ship it via the RPM at
/usr/share/punktfunk/bazzite/ and document it in the Bazzite README (new §6.5).
Gaming Mode needs none of this (auto-attach).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bazzite/SteamOS boxes flip between Steam Gaming Mode (gamescope) and a
KDE/GNOME desktop. The host statically read PUNKTFUNK_COMPOSITOR /
XDG_CURRENT_DESKTOP once, so switching to Desktop Mode failed the stream, and
the gamescope managed-session path stopped+relaunched the autologin per connect
— leaking GPU context on F44 (reconnect → black screen).
Replace the static read with a runtime probe of the live session and route each
connect to the right backend, churn-free:
- vdisplay::detect_active_session() probes /proc for the running compositor of
our uid (gamescope|kwin_wayland|gnome-shell|sway, desktop outranks a leftover
gamescope) + scans the runtime dir for the live wayland-* socket. Returns an
ActiveKind + the SessionEnv (WAYLAND_DISPLAY/XDG_RUNTIME_DIR/DBUS/
XDG_CURRENT_DESKTOP) that targets it.
- apply_session_env() writes that into the process env per connect (host serves
one session at a time), so every backend (capture + input) opens against the
live session; apply_input_env() points input at the matching backend and
selects gamescope ATTACH (no managed restart) unless PUNKTFUNK_GAMESCOPE_MANAGED.
- resolve_compositor() (native path) auto-detects + applies; explicit
PUNKTFUNK_COMPOSITOR still wins (legacy/CI/forcing). detect() is now
active-aware for the GameStream/mgmt callers too.
- Bazzite host.env drops the static gamescope force; documents auto-detection
+ the optional overrides.
Result: Desktop Mode → KWin/Mutter virtual output at the client's mode
(churn-free, the reliable path); Gaming Mode → attach to the running gamescope
(no SIGSEGV/GPU leak on reconnect). Compiles + clippy-clean; 78 host tests pass.
Live validation on the Bazzite box pending (box offline).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On a Bazzite host that autologins into gaming mode on a physical display (the F44
default: gamescope-session-plus@ogui-steam on the TV), Steam — single-instance — is
held by that session, which renders to the TV's native mode. The host-managed session
then can't start its own Steam, so it captured the TV's 4K output instead of the
client's mode (stretched). On F43 the box wasn't in gaming mode, so the host's Steam
was the only one.
Fix: on connect, the host-managed gamescope path stops any running autologin
`gamescope-session-plus@*` unit (frees Steam) before launching its own session at the
client's mode; on client disconnect (`restore_tv_session`, called from serve_session
teardown) it stops our session and restarts the autologin one, so the TV returns to
gaming mode by default when no one is streaming. Stopping the `--user` unit sticks
(Relogin only fires on the full logind session ending — verified live), so no sddm
config change is needed. Cost: a Steam cold-start per connect, given single-instance.
No-op on non-Bazzite / headless boxes (nothing to stop → nothing to restore).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The earlier buttonHome handler wasn't enough: on macOS the SYSTEM grabs the DualSense
Home/PS button by default (opens Launchpad's Games folder), so it never reached the app.
The fix is to disable the system gesture on the element —
`physicalInputProfile.buttons[GCInputButtonHome].preferredSystemGestureState = .disabled`
(Apple's documented mechanism) — which hands the button to us.
Then drive `guide` DIRECTLY from that element's pressedChangedHandler instead of via
buttonMask: the legacy `extendedGamepad.buttonHome` is unreliable/often nil even when the
physical element exists, so reading it in the mask dropped presses. `sendGuide` folds the
bit into `buttons` so a held PS button still releases on focus loss. On tvOS the element
is reserved (nil) → the block no-ops.
The host already maps BTN_GUIDE → the DualSense PS bit, so this completes the chain.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Phase 3: the Apple library now talks to the host's HTTPS mgmt API (b4a85a8) over mTLS
using this client's persistent identity — the SAME cert the host paired over QUIC — so
there is NO manual token anymore.
- ClientTLS: builds a SecIdentity from the stored PEM (CryptoKit parses the rcgen P-256
PKCS#8 key → x963 → SecKey; the cert PEM → SecCertificate; SecIdentityCreateWithCertificate
pairs them via the Keychain). macOS-only for now (that API is unavailable on iOS — a
PKCS#12 path would be needed there; the client is macOS-first).
- LibraryTLSDelegate: pins the host's self-signed cert by the fingerprint the client
already trusts, and presents the identity for the client-cert challenge.
- LibraryClient.fetch now does GET https://…/library with the identity + host fingerprint;
the whole connection form (port + token) and StoredHost.mgmtToken/setMgmt are gone — the
library "just works" for a paired host. 401 → "pair with the host first".
Can't compile Swift on the Linux box; CI (apple.yml) compiles the macOS path incl. the
Security/CryptoKit code. Runtime (SecIdentity build + the mTLS handshake) needs Mac
validation. Pairs with the host mTLS already landed + live-tested.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Phase 1 of moving the library off a manual mgmt token: the management API now serves
over HTTPS with the host's persistent identity (the cert clients already pin) and
OPTIONAL client-cert auth. A request is authorized if EITHER the peer presented a
client certificate whose SHA-256 is in the punktfunk/1 paired store (the same trust the
QUIC data plane uses — so a paired native client needs no token), OR it carries the
bearer token (the web console / admin). `/health` stays open.
axum-server can't surface the peer cert to a handler, so `serve_https` runs the rustls
handshake itself (tokio-rustls), reads the verified peer certificate, and serves the
axum Router over hyper with the fingerprint attached to each request; `require_auth`
checks it against `NativePairing::is_paired`. The verifier reuses the GameStream
AcceptAnyClientCert, parameterized to make client auth optional (a browser with no cert
still completes the handshake and falls back to the token).
Validated live: paired cert → 200, unpaired cert / no creds / bad token → 401, bearer
→ 200, /health open. (Note: the API is now HTTPS with a self-signed cert — a browser
shows a one-time trust prompt; native clients pin by fingerprint.)
Next: Apple client presents its identity over mTLS (drops the token field); embed the
web console; enable HTTPS mgmt by default.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The client-side cursor positions the host pointer with ABSOLUTE events, but
gamescope's input socket (EIS) grants only a relative pointer — the host drops the
absolute events (libei.rs: no PointerAbsolute → not emitted), so the pointer never
moves and clicks/scroll land on the stuck position. Auto-mode enabled exactly this on
gamescope, making all input appear dead until toggled off.
Force `cursorVisible = false`, neuter the ⌘⇧C toggle, and hide the now-inert Settings
picker. The resolution logic + handlers are kept (commented) for when per-compositor
gating (KWin/GNOME/Sway have an absolute pointer) or a synthetic-cursor-over-relative
path lands. Relative capture (the working path) is now always used.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two issues from live Mac testing, plus a requested fullscreen option:
- PS button: the Home/PS button (→ guide; the host maps it to the DualSense PS bit)
does not reliably fire GCExtendedGamepad.valueChangedHandler on macOS, so its presses
were dropped. Add a dedicated buttonHome.pressedChangedHandler that re-syncs. The host
already maps BTN_GUIDE→PS, so this is the missing client half.
- Fullscreen: a macOS FullscreenController (NSViewRepresentable) takes the window
fullscreen while a session is up (incl. the trust prompt over the blurred stream) and
restores it on the host list — so only the stream is fullscreen, not the picker. New
`fullscreenWhileStreaming` setting (default on) + a Settings "Window" toggle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two DualSense (UHID) fixes surfaced live on the Bazzite host:
- Battery: serialize_state never set the input report's status byte (struct off 52 →
r[53]), so hid-playstation read battery capacity 0 and SteamOS warned "low battery"
even on a fully-charged pad. Set it to 0x0A (discharging, low nibble 0xA → 100 %) —
a virtual pad has no real cell. (Forwarding the client pad's real charge is a later
feature.) Regression assert added to the layout test.
- Rumble diagnostic: log the silent→active transition when forwarding a buzz on the
0xCA plane, so a live test can tell "host never receives rumble from the game"
(Steam Input / parse) apart from "client doesn't render it". Once per buzz, no spam.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Tapping a game in the (flagged) library now starts a session that asks the host to
launch it — the picked GameEntry id rides the connect down to the host, which resolves
it against its own library (27e5865).
- PunktfunkConnection.init gains `launchID` and calls the new punktfunk_connect_ex4
(wrapping it in withOptionalCString; nil = host default).
- Threaded SessionModel.connect(launchID:) → ContentView.connect(_:launchID:) →
a `launchTitle(host, id)` helper that dismisses the browser and connects.
- LibraryView gains `onLaunch`; cards become buttons that fire it. Wired on every
platform (ContentView sheet on macOS/iOS, HomeView destination on tvOS) via a new
`onLaunchTitle` closure on HomeView. Settings footer updated (launch is live now).
Can't compile Swift on the Linux box; CI (apple.yml) verifies. The host side of this
chain is live-validated on the dev box: a client `--launch custom:<id>` made the host
resolve the id and spawn gamescope running the title (see 27e5865).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Plan step 4 (plumbing + host behavior). A client can ask the host to launch a
library title on connect; the host resolves it against ITS OWN library and runs it
in the session — the client sends only the store-qualified id, never a command, so a
remote peer can't inject one.
- Protocol (quic.rs): `Hello.launch: Option<String>` (the GameEntry id). Appended
after `name`; when launch is present but name absent, a zero-length name placeholder
keeps the offset deterministic — so a Hello with neither field stays byte-identical
to the bitrate-era 26-byte form (test-asserted). Old peers ignore it; new hosts
decode None from old clients. Round-trip + back-compat + truncation tests.
- Host: `library::launch_command(id)` resolves id → command via the host's own library —
`steam_appid` → `steam steam://rungameid/<appid>` (appid validated as digits, the only
client-influenced part), `command` → the host-stored command verbatim (trusted, never
from the client). m3.rs sets PUNKTFUNK_GAMESCOPE_APP from it before bringup, exactly
as the GameStream /launch path does (one session at a time). Unit-tested incl. an
injection-attempt guard. Takes effect on the bare-spawn gamescope path; a no-op on a
shared desktop / attach-to-existing session.
- C ABI: `punktfunk_connect_ex4` adds `launch_id` (NULL = none); `_ex3` now delegates to
it. Threaded through NativeClient::connect → WorkerArgs → Hello.
- client-rs gains `--launch ID` (headless testing); client-linux passes None (no picker
yet). Header regenerated.
Next: the Apple library grid passes the picked id via punktfunk_connect_ex4.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Plan step 3 — the Apple client surfaces the host's game library, behind a feature
flag (`DefaultsKey.libraryEnabled`, default OFF). Browsing only; launching a chosen
title is step 4.
- PunktfunkKit `LibraryClient`: Codable GameEntry/Artwork/LaunchSpec mirroring
crates/punktfunk-host/src/library.rs, and an async fetch of GET /api/v1/library
with a bearer token. Typed LibraryError guides setup (the common case is "needs a
--mgmt-token"). `Artwork.posterCandidates` = portrait → header → hero.
- `LibraryView`: cross-platform poster grid (LazyVGrid, AsyncImage that walks the art
candidates past load failures to a text placeholder), a store badge, and an inline
Connection form (mgmt port + token) that surfaces when the API is unreachable / 401
/ no token set. Read-only.
- StoredHost gains `mgmtPort`/`mgmtToken` (the mgmt API is a distinct port from the
data plane and needs a token off-loopback). Both OPTIONAL — synthesized Decodable
ignores property defaults but treats a missing Optional as nil, so older saved
hosts decode unchanged (a defaulted non-optional would wipe the list). HostStore.setMgmt.
- Entry point: a flag-gated "Browse Library…" host-card context action → LibraryView
(sheet on macOS/iOS, pushed on tvOS), mirroring the pair/speed-test plumbing. Plus a
Settings "Experimental" toggle.
Can't compile Swift on the Linux dev box; CI (apple.yml: swift build + swift test on
the mac mini) verifies the macOS path. Added LibraryClientTests (decode + art order)
for `swift test`. iOS/tvOS-only branches mirror existing patterns. Live-verify on the
Mac pending.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Consumes the new library API (6351d51) via the orval-generated hooks. A poster grid
over GET /api/v1/library (all stores merged), plus create/edit/delete for custom
entries — the admin-UI half of "create custom entries via the web console".
- GameCard: portrait (600×900) art with an onError fallback chain portrait → header
→ text placeholder (many Steam titles lack a 600×900 capsule). A store badge marks
Steam vs Custom; only custom cards expose edit/delete.
- Inline add/edit form (title + portrait/hero/header URLs + optional launch command,
mapped to LaunchSpec{kind:"command"}) wired to useCreateCustomGame /
useUpdateCustomGame / useDeleteCustomGame; the CRUD id strips the `custom:` prefix;
every mutation invalidates the library query. QueryState handles load/empty/error.
- Nav entry (LibraryBig) + en/de i18n strings.
`bun run lint` (tsc) and `bun run build` both green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A new `library` module + four mgmt endpoints surface the host's games to clients
(plan: "surface the user's games"). An adapter layer (`LibraryProvider`) so future
stores (Heroic/Epic, GOG, Lutris) slot in behind one uniform `GameEntry`.
- SteamProvider: reads the LOCAL Steam install — no Steam Web API key, no network.
Installed titles from steamapps/appmanifest_<appid>.acf; extra library folders
(incl. paths with spaces) from libraryfolders.vdf; candidate roots cover classic,
Flatpak and Deck layouts, canonicalized + deduped (the .steam/{steam,root}
symlinks all fold to one). Runtimes/redistributables (Proton, Steam Linux Runtime,
Steamworks Common, SteamVR) filtered out. Artwork = the public Steam CDN by appid
(portrait/hero/logo/header), fetched directly by the client.
- Custom store: ~/.config/punktfunk/library.json, write-then-rename persisted,
CRUD'd via the API — the "create custom entries via the admin web UI" requirement.
- API (under /api/v1, OpenAPI-documented + checked in): GET /library (all stores
merged, sorted), POST /library/custom, PUT/DELETE /library/custom/{id}.
- `punktfunk-host library` subcommand dumps the resolved library as JSON (diagnostic,
mirrors `openapi`).
Validated live against the real Steam library on the Bazzite box: 89 appmanifests →
78 games (11 tools filtered), correct titles/sort, and the CDN art URLs return 200.
5 unit tests for the VDF/ACF parsing, tool filter, art URLs, custom mapping.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Decky plugin (b3f98a5) launches `punktfunk-client`, but the Arch package only
shipped the host, so the Deck had nothing to launch. Convert the PKGBUILD to a
split package (pkgbase=punktfunk → punktfunk-host + punktfunk-client), mirroring the
rpm subpackages and the two deb build scripts:
- punktfunk-host: unchanged artifact set + NVENC/compositor optdepends.
- punktfunk-client: the GTK4 binary + io.unom.Punktfunk.desktop + the hidraw udev
rule + the 32MB recv-buffer sysctl; depends gtk4/libadwaita/sdl3/ffmpeg/pipewire/
opus; optdepends libva-mesa-driver (VAAPI decode on the Deck's AMD APU, software
fallback otherwise). New punktfunk-client.install scriptlet.
- build-sysext.sh now derives the package name from the file, so it wraps either the
host OR the client into a systemd-sysext .raw — on a Deck you wrap the client.
- README: split-package usage + a "Steam Deck (the client)" section tying the sysext
to the Decky plugin (client is on PATH → plugin launches `punktfunk-client
--connect host:port`). Clarified the VAAPI gap is host-ENCODE only; the client
DECODES via VAAPI on the Deck today, so streaming to a Deck works now.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A Decky Loader plugin so a Steam Deck / SteamOS box can launch the punktfunk
client from Gaming Mode using REAL Steam UI components (it runs inside Steam's
CEF, so the panel is built from @decky/ui — the literal Big Picture primitives,
not a replica).
- Frontend (src/index.tsx, @decky/api + @decky/ui): a Quick Access Menu panel —
Refresh → discover hosts, a native list (name, ip:port, pairing flag), tap to
connect with a status toast, Disconnect.
- Backend (main.py): discover() shells `avahi-browse -rpt _punktfunk._udp` and
parses the host's advertised TXT keys (proto/fp/pair/id from discovery.rs),
dedup by id preferring IPv4; connect() resolves + spawns
`punktfunk-client --connect host:port` (gamescope composites its video like a
game), tracking the child; disconnect() terminates it.
- Mirrors the current official Decky template (the API moved to @decky/ui +
@decky/api). Frontend builds clean (pnpm build → dist/index.js); main.py
py_compiles. dist/ + node_modules gitignored — build on the Deck per README.
Spike scope: launcher only, runtime untested (no Deck here). Next on this track:
the in-stream Quick-Access overlay (volume/disconnect/stats over the running
stream) and a fuller real-components UI. Client decode on the AMD Deck is the
existing VAAPI path; the host-encode VAAPI gap is separate (NVIDIA host = NVENC).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
gamescope's PipeWire capture carries no cursor (verified upstream — it never
composites the cursor or adds SPA_META_Cursor), so the cursor must be drawn on the
client. New macOS "cursor-visible" capture mode: instead of disassociating+hiding
the system cursor and sending relative deltas (the game path, unchanged), it keeps
the system cursor visible over the stream and sends ABSOLUTE positions
(MouseMoveAbs), mapped through the video's aspect-fit (AVMakeRect) to host pixels
with the letterbox bars dropped. The visible system cursor IS the client cursor —
zero added latency, no double cursor (gamescope draws none), accurate (the client
drives the host's absolute mouse).
- Default: on iff the session's resolved compositor is gamescope (via the new
punktfunk_connection_compositor getter, fc30307).
- Settings: "Cursor in stream" → Auto (gamescope) / Always / Never.
- Shortcut: ⌘⇧C toggles it live mid-session (re-engages capture so disassociation
+ abs/rel forwarding swap atomically); shown in the HUD.
macOS-only (the visible-cursor mode lives in the macOS StreamView). Verified to
compile + link via xcodebuild Release on the Mac; runtime behavior (cursor landing,
hover forwarding) to be confirmed live. Rust ABI side committed separately.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add punktfunk_connection_compositor() (mirrors punktfunk_connection_gamepad): a
client getter for the compositor the host actually resolved for the session, read
from Welcome.compositor and threaded through NativeClient.resolved_compositor. The
Apple/Linux clients use it to enable the client-side cursor by default on gamescope
sessions, whose PipeWire capture carries no cursor (verified upstream). Header
regenerated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add packaging/arch: a PKGBUILD mirroring the rpm/deb artifact set (binary, udev
rule, 32MB sysctl, systemd USER units with ExecStart rewritten, headless helpers,
env templates, openapi), a pacman .install scriptlet, a systemd-sysext builder for
immutable SteamOS, and a README. Builds the working tree via PF_SRCDIR (CI/dev) or
a git tag (AUR). Arch's stock ffmpeg already ships NVENC, so deps collapse to ~10
packages with nvidia-utils/compositors as optdepends (never hard-depend on the
driver, same invariant as rpm/deb).
SteamOS delivery is a **systemd-sysext** (overlays /usr read-only from writable
/var/lib/extensions/, survives A/B OS updates, no steamos-readonly disable) —
pacman/distrobox/flatpak are all unsuitable for a host that needs uinput/uhid, the
host PipeWire socket, the GPU node, and to spawn a compositor.
KNOWN GAP, documented prominently: encode is NVENC-only (src/encode/linux.rs has no
VAAPI backend), so this works on Arch+NVIDIA (and bazzite-deck-nvidia) but an AMD
Steam Deck installs yet cannot encode until a hevc_vaapi backend is written — a code
change, not packaging.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A global PROVISIONING_PROFILE_SPECIFIER on the xcodebuild command line is
applied to every target in the graph, including the shared SwiftPM compiler-
plugin macros (OnceMacro/SwizzlingMacro/AssociationMacro). Those build for the
macOS host and reject a provisioning profile, so the iOS/tvOS device archives
failed at build-description time with "<macro> does not support provisioning
profiles". (The macOS archive is immune: its host-SDK macros carry
CODE_SIGNING_ALLOWED=NO, so the global specifier is silently ignored there.)
Move the signing settings into a generated -xcconfig and condition the profile
+ identity on the device SDK ([sdk=iphoneos*] / [sdk=appletvos*]). xcconfig
conditionals are honored and a command-line -xcconfig outranks target settings,
whereas a CLI "SETTING[sdk=..]=val" is mis-parsed — both verified via
xcodebuild -showBuildSettings against the real project. The profile now lands on
the app/framework slices only; the macosx-host macros get nothing.
macOS App Store archive is unchanged (already green; installer cert now present
on the runner). tvOS upload may still need tvOS on the App Store Connect record,
but that step is continue-on-error.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of the Mac "session ended" at 880 Mbps / 1.3 Gbps: the host requests a
bitrate NVENC can't express at any codec level and `avcodec_open2` returns EINVAL
("Invalid argument"), so the pipeline build fails after 4 identical retries and the
session dies at encoder init — before a single video packet (which is why the
client's UDP counters never moved). The ceiling is GPU/driver-specific: an RTX 4090
caps HEVC at ~800 Mbps (Level 6.2 High tier) and rejects above it, while an RTX
5070 Ti accepts 1.3 Gbps.
Rather than hard-cap every build to a conservative guess (which would needlessly
throttle capable cards), open_video now PROBES: open at the requested bitrate, and
step down (codec spec ceiling, then 0.75x to a 50 Mbps floor) ONLY when this GPU
returns EINVAL. Each GPU runs at its own real maximum — the 5070 Ti keeps 1.3 Gbps,
the 4090 lands at 800 Mbps and streams instead of dying. Non-EINVAL failures (no
GPU, bad mode, OOM) still surface immediately rather than being masked by retries.
Codec::max_bitrate_bps is now just the first step-down candidate, not a clamp.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The client asks the kernel for a 32 MB SO_RCVBUF, but the kernel silently clamps
it to net.core.rmem_max — whose default is far too small. A too-small recv buffer
is the dominant client-side wall above ~1 Gbps. Measured live (Fedora host -> two
clients, real 2.5G LAN, GSO off): a client capped at 4 MB rmem_max dropped 31.6%
of a 2 Gbps stream at the receiver, while a 32 MB client delivered the same
2 Gbps at 0.0% loss. The host already shipped this tuning; the client packages
didn't (the RPM's %post even referenced the host-only file), so a client-only
install streamed lossy at high bitrate.
Add scripts/99-punktfunk-client-net.conf (rmem/wmem = 32 MB, distinct filename so
host+client can coexist) and ship+apply it from both the .deb (build-client-deb.sh)
and the RPM client subpackage (install, %files client, %post client).
For reference the full ladder (punktfunk speed-test): 0% loss to 1.5 Gbps on a
4 MB client; 31.6% at 2 Gbps on 4 MB vs 0% at 2 Gbps on 32 MB. iperf3 put the raw
link at ~2.35 Gbps TCP / ~2.4 Gbps UDP, so the stack now tracks the wire given a
big enough recv buffer.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Enabling PUNKTFUNK_GSO on a host whose egress MTU is below our UDP segment size
made every GSO send return EMSGSIZE (code 90, "Message too long") — the kernel
validates each GSO segment against the device MTU at send time, which plain
sendmmsg does not. EMSGSIZE wasn't in gso_unsupported() (nor is_transient_io), so
it propagated as a fatal "send failed — stopping stream" and instantly killed
every session the moment GSO was on (observed live: connection fails instantly /
speed-test 0 Mbps).
Add EMSGSIZE to gso_unsupported() so it latches GSO off for the process and
finishes via sendmmsg — the standard "GSO not usable on this path" fallback.
Measured after: the same host+path does 1 Gbps at 0.0% loss over the real LAN via
sendmmsg (and the host send path sustains a 2 Gbps probe with send_dropped=0), so
GSO is a >2 Gbps optimization, not required for 1 Gbps.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of the Mac "session ended" at higher bitrates. The video data plane is
a *connected* UDP socket; with data-plane hole-punching the path can blip and the
kernel surfaces an asynchronous ICMP port-unreachable/reset as ECONNREFUSED /
ECONNRESET on a later send or recv. Both the host send loop and the client
poll_frame treated that as fatal and tore the session down:
ERROR punktfunk_host::m3: send failed — stopping stream
error=send_sealed: Io(ConnectionRefused, code 111) <-- observed live
That also cascades: a transient ICMP makes the client's poll_frame bail and close
its data socket, which makes the host's next send get a *real* ECONNREFUSED, which
tears the host side down too — exactly the "broke at 500 Mbps+" report.
Fix: classify ECONNREFUSED/ECONNRESET alongside WouldBlock as transient (a lossy
drop / "no data this poll"), never a teardown, at every data-path send/recv site
(send, send_batch, send_gso, recv, recv_batch x2, recv_batch_x). FEC + the next
frame/RFI recover; if the peer is genuinely gone the QUIC control plane's
conn.closed() ends the session cleanly (no infinite "stream into the void").
This is the standard connected-UDP rule that ICMP errors are advisory — doubly
true with hole-punching. Adds is_transient_io() + a unit test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Mac App Store requires App Sandbox, which the macOS app didn't declare.
App Sandbox is macOS-only (invalid on iOS/tvOS, fails upload validation), so
the macOS target now uses a dedicated Config/Punktfunk-macOS.entitlements while
iOS/tvOS keep the shared Config/Punktfunk.entitlements (unchanged). The single
macOS app is sandboxed for BOTH channels — the Developer ID DMG is codesigned
with the same file — so the local build equals what App Store users get.
Entitlement set (verified against the code + Apple docs):
- app-sandbox, network.client.
- network.server: NOT optional despite the client being outbound-only — the
sandbox gates the bind() syscall as network-bind, and quinn (quic.rs) + the
raw-UDP plane (transport/udp.rs) both bind explicitly, so host->client
datagrams never arrive without it (the classic QUIC-under-sandbox trap).
- device.audio-input (mic uplink), device.bluetooth + device.usb (Xbox/DualSense
controllers over BT/USB via GameController), keychain-access-groups (existing).
Omitted: device.hid (undocumented), files.user-selected.* (no pickers),
networking.multicast (Bonjour browse is exempt; requesting it breaks signing).
CI (release.yml): add a macOS App Store archive+upload-to-TestFlight step
mirroring the iOS lane (manual Apple Distribution signing + the 'Punktfunk macOS
App Store Distribution' profile, app-store-connect/upload, installer-signed pkg),
continue-on-error until the portal prereqs exist; point the Developer ID DMG
codesign at the sandboxed entitlements. Docs (ci.md) + clients/apple README
updated; the runner additionally needs the macOS platform on the App Store
Connect record + the '3rd Party Mac Developer Installer' cert.
Verified: signed Debug build embeds exactly the intended entitlements
(codesign -d --entitlements), swift build green against the rebuilt xcframework.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Mac/iOS client's wall around ~380 Mbps on a 2.5 G path is the receive
drain, not the transport: a loopback speed-test pushes 380/600/1000 Mbps at
0.0% loss, but Darwin has no recvmmsg(2), so the macOS client was doing one
recv() syscall per packet — ~40-90k syscalls/s on one core. When the recv loop
can't drain fast enough the kernel socket buffer backs up and drops, which the
client sees as a sustained stream stalling/freezing in the 300-400 Mbps range
(and an immediate "session ended" when a 500 Mbps+ first keyframe bursts in).
- core/transport: flip recvmsg_x (the batched Darwin recv, ~30x fewer syscalls)
from opt-in to default ON, opt-out via PUNKTFUNK_RECVMSG_X=0. Keeps the
auto-fallback to the scalar loop on any unexpected syscall error. The Apple CI
swift-test loopback now exercises this path by default.
- packaging/kde host.env: enable PUNKTFUNK_GSO=1 — UDP segmentation offload on
the host send path (one sendmsg per ~64 packets), the dominant lever above
~1 Gbps. Already wired (send_sealed -> send_gso) with sendmmsg auto-fallback.
- apple SpeedTestSheet: lengthen the bandwidth probe 2 s -> 5 s so the measured
number stops swinging wildly (50 vs 900 Mbps on the same link) — long enough
for steady-state send + recv drain to settle. Matches host MAX_PROBE_MS.
- host capture: PUNKTFUNK_SYNTH_NOISE synthetic high-entropy source for
reproducible throughput testing of the encode->FEC->send->recv path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
tvOS archive failed 'Macro AssociationMacro/SwizzlingMacro/OnceMacro must be
enabled before it can be used' — Xcode 15+ requires interactive trust for SPM
Swift macros (objc-runtime-tools, swift-once-macro via swiftui-navigation-
transitions), which a headless build can't grant. Add -skipMacroValidation
-skipPackagePluginValidation to all three archive commands so CI never hits the
trust prompt.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Audio: a headless host has no speakers, and on a LAN with AirPlay devices PipeWire picks a random
HomePod as default — so desktop audio (which the host captures from the default sink's monitor)
went to a HomePod over AirPlay instead of to the client, and there was no "Punktfunk" output to
select. Ship a `punktfunk-sink.conf` (a `support.null-audio-sink` adapter — NOT the non-existent
module-null-sink, which makes pipewire refuse to start) with high priority.session so it's the
default; run-headless-kde.sh installs it and restarts pipewire once on first install. The host then
captures its monitor and streams it. (Disable AirPlay sinks out of band: `dnf remove
pipewire-config-raop`.)
Input: the host's libei portal D-Bus connection goes stale when the compositor session restarts the
portal under it, and the in-process reopen loop can't recover it (EIS setup keeps timing out) — only
a full restart does. Add PartOf=punktfunk-kde-session.service so the host restarts with the session.
Both verified live on the Fedora 44 KDE box.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
tvOS is scaffolded (Punktfunk-tvOS target/scheme + build-xcframework BUILD_TVOS).
Wire it: install nightly + rust-src (tier-3 -Zbuild-std), build the xcframework
with BUILD_TVOS=1, and add a tvOS archive+export+upload step mirroring iOS
(manual signing with the 'Punktfunk tvOS App Store Distribution' profile, since
the App-Manager ASC key can't cloud-sign). Also point iOS at the renamed
'Punktfunk iOS App Store Distribution' profile. macOS App Store/TestFlight still
pending (needs App Sandbox entitlements). Needs tvOS on the App Store Connect
app record + the tvOS platform installed on the runner.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
X11/Electron apps (Discord — "Missing X Server or $DISPLAY", Steam, many launchers) failed in the
headless KWin session: `kwin_wayland --virtual` starts NO X server unless asked, and even with one
KWin reserves the X11 display + starts Xwayland *on demand* (no Xwayland process or "Using public
X11 display" log line until the first client connects) — so the old detection (pgrep the Xwayland
process) found nothing and never exported DISPLAY. Two fixes: pass `--xwayland`, and detect the
display from the reserved /tmp/.X11-unix/X<N> socket (with the log + process checks as fallbacks).
Verified live on the Fedora 44 KDE box: DISPLAY=:0 lands in plasmashell + the activation env and
xdpyinfo responds, so menu-launched X11 apps open a display.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A running Xcode.app manages ~/Library/Developer/Xcode/UserData/Provisioning
Profiles/ and deletes manually-installed (unrecognized) distribution profiles —
which is why the App Store profile vanishes. Quit Xcode at the start of the iOS
step so the manually-installed 'Punktfunk App Store Distribution' profile
survives for manual signing; headless xcodebuild doesn't need the GUI app.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
macOS Developer ID + notarize + DMG now works with the clean login-keychain
workflow. iOS export failed with 'Cloud signing permission error' — with
-allowProvisioningUpdates Xcode forces cloud-managed signing, which the
App-Manager-role ASC key can't authorize. Switch iOS to MANUAL signing with the
local (valid) Apple Distribution identity + the 'Punktfunk App Store
Distribution' provisioning profile; ASC key stays only for the upload. Profile
must be installed via Xcode -> Accounts -> Download Manual Profiles.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The runner now runs as a user LaunchAgent in the logged-in Aqua session, so it
uses the login keychain directly, where Developer ID Application + Apple
Distribution are installed and VALID (the missing WWDR intermediate — the real
root cause of the whole iOS saga — is now present). Delete all the throwaway-
keychain / secret-cert-import / raw-keychain-plumbing / Xcode-quit / diagnostic
machinery: macOS = archive-unsigned + a single Developer ID codesign + notarize/
DMG; iOS = standard xcodebuild archive + export with -allowProvisioningUpdates
(automatic signing manages the App Store cert + profile). Only ASC_API_KEY_*
secrets remain; DEVID_CERT_*/IOS_DIST_CERT_*/IOS_PROFILE_B64 no longer needed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On a headless KDE appliance, libei input injection silently failed: the EIS socket comes from the
xdg RemoteDesktop portal, which never came up, and even up it would pop an unanswerable "Allow
remote control?" dialog. Three fixes in run-headless-kde.sh, all idempotent + safe on the dev box:
- Reach graphical-session.target: xdg-desktop-portal is ordered behind it and its start job fails
without it, but a headless linger session never gets there and Fedora's target has
RefuseManualStart=yes — drop that in once, then start the target.
- Start the portal with `start` (the old `try-restart` is a no-op when inactive — the first-boot
case), so it actually comes up.
- Pre-seed the RemoteDesktop grant: vendor the `kde-authorized` permission-store GVariant DB and
copy it to ~/.local/share/flatpak/db/ (never clobbering an existing one), so the portal grants
RemoteDesktop without a dialog. Shipped by the RPM + .deb.
Diagnosed + fixed live on the Fedora 44 KDE box: libei devices RESUME and emit (MouseMove/keys).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On a headless KDE appliance the session has two outputs — run-headless-kde.sh's `kwin --virtual`
bootstrap (where plasmashell draws by default) and our per-session streamed output — so the client
saw only the wallpaper of an empty extended output (the KWin analogue of the GNOME/Mutter
VIRTUAL_PRIMARY issue). New opt-in PUNKTFUNK_KWIN_VIRTUAL_PRIMARY: after creating the virtual
output, set it primary via kscreen-doctor (KWin then re-homes the desktop onto it and disables the
bootstrap), then belt-and-suspenders disable anything still enabled. The keepalive re-enables the
bootstrap on teardown — though KWin also auto-re-enables it when our output is reclaimed, so there's
never a zero-output window. Set in packaging/kde/host.env. Verified live on the Fedora 44 KDE box:
mid-session the streamed output is the sole desktop at 0,0; post-session the bootstrap is back.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The throwaway-keychain codesign still fails 'unable to build chain to self-signed
root / errSecInternalComponent' despite cert/chain/key all verifying. Sign by the
Apple Distribution identity's SHA-1 hash (eliminates name-matching ambiguity, a
known cause) and run codesign --verbose=4 + print valid/matching identities at
sign time, to surface the exact failure on the next run.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
iOS codesign still failed with 'unable to build chain to self-signed root /
errSecInternalComponent' after the keychain re-assert. verify-cert proves the
chain is trusted, so this is the private-key ACL (errSecInternalComponent is
classically that) and/or codesign not finding the chain certs in the identity's
keychain. Right before the iOS codesign: re-run set-key-partition-list (re-grant
codesign access to the key) and import the WWDR G3 intermediate + Apple Root CA
into the throwaway keychain so the full leaf->WWDR->root chain is present there.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The iOS archive SUCCEEDS now (raw-codesign path), but codesign failed with
'unable to build chain to self-signed root / errSecInternalComponent'. Cause:
xcodebuild archive (run in the same step, just before codesign) resets the user
keychain search list, so codesign can no longer find the WWDR intermediate that
lives only in the throwaway keychain. The macOS sign avoids this by running in a
separate step after its re-assert. Re-assert the search list + default keychain
(and unlock, via KEYCHAIN_PASS now exported to GITHUB_ENV, masked) immediately
before the iOS codesign.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
cargo fmt --all --check failed CI on the long match-arm guard in UdpTransport::connect_via_punch;
apply the formatter's wrapping. No behavior change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
xcodebuild's signing-identity selection enforces an online revocation/OCSP check
that excludes the freshly-minted Apple Distribution cert (find-identity -v drops
it) even though verify-cert confirms it's valid and codesign signs with it fine.
So sign iOS the same way as the macOS DMG: archive CODE_SIGNING_ALLOWED=NO, embed
the profile, raw 'codesign --keychain' with the profile's entitlements (extracted
via plutil), package the .ipa, and upload with 'xcrun altool --upload-app'. Drops
the xcodebuild manual-signing path entirely — no profile-dir install, no
Xcode-quit, no provisioning-profile discovery.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of 'No profile matching Punktfunk App Store Distribution': the GUI
Xcode.app was running on the runner and actively manages
~/Library/Developer/Xcode/UserData/Provisioning Profiles, pruning our
manually-installed App Store profile from the exact dir xcodebuild reads, right
before signing (the legacy ~/Library/MobileDevice copy survives but Xcode 26's
xcodebuild doesn't read it). Quit Xcode.app at the top of the iOS signing block;
xcodebuild runs independently and headless CI doesn't need the GUI app.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- scripts/bench/compare.py: diff criterion medians (target/criterion/**/estimates.json) vs a
committed baseline, print a markdown table to the job summary, flag >threshold regressions, always
exit 0 (shared CI hardware is too noisy to gate on). --update rewrites the baseline.
- ci.yml `bench` job: runs Tier-1 (criterion) + Tier-2 (loss-harness FEC recovery) GPU-free in the
rust-ci container, then compare.py — report-only visibility per push/PR.
- scripts/bench/gpu-stream.sh + bench-gpu.yml: Tier-3 real pipeline (virtual output → zero-copy →
NVENC → punktfunk/1 → reassemble) on a self-hosted GPU runner; captures encode_us/tx_mbps/
send_dropped + client capture→reassembled latency, compares to gpu-baseline.json (20% threshold).
Needs the dev box registered as a `[self-hosted, gpu]` act_runner (one-time, see the workflow
header) — the dedicated hardware makes its absolute baseline meaningful, unlike shared CI.
- baseline.json: dev-box Tier-1 numbers.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
iOS manual signing fails 'No profile matching Punktfunk App Store Distribution'
despite the profile being installed (content verified: right name/team/iOS/app-id).
The profile is in ~/Library/MobileDevice but Xcode 26 reads
~/Library/Developer/Xcode/UserData/Provisioning Profiles, which is empty. Print
both dirs before the archive and again at failure to confirm whether Xcode
regenerates/prunes the UserData copy during the build.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GPU-free, so they run in normal CI. Two layers: crypto/{seal,seal_in_place,open} on one MTU shard,
and pipeline/{gf8,gf16}/{64KB,1MB} — a whole frame through the real per-frame path end to end over
the loopback transport (FEC encode → AES-GCM seal → packetize → reassemble → FEC decode → open).
Baselines on the dev box (RTX 5070 Ti VM): AES-GCM ~1.57 GiB/s/shard; gf16 ~418 MiB/s at 1 MB vs
gf8 ~23 MiB/s (the GF(2^8) O(n^2) ceiling the GF(2^16) Leopard wall-breaker removes — exactly the
kind of regression this should catch). The GPU capture/NVENC path is out of scope here (Tier 3).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The profile-name/UUID read used 'security cms -D ... || true' which masked a
failed decode, then PlistBuddy printed 'Error Reading File' to stdout and that
got captured as the UUID, producing a garbage cp path. Now: check the extracted
plist is non-empty, fall back to 'openssl smime' if 'security cms' fails,
validate the UUID is actually hex+dashes, and print the decoded byte count +
decoder stderr + first bytes so a bad IOS_PROFILE_B64 is obvious in-log. Still
non-fatal (skips iOS, never blocks the macOS release).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The video data plane is a raw UDP socket separate from the QUIC control connection. On a flat LAN
the host can send straight to the client, but across NAT or a stateful inter-VLAN firewall the
unsolicited host→client video is rejected (ICMP port-unreachable → the session dies immediately,
while control/audio/input keep working since they ride the client-initiated QUIC). Observed live:
a client on 192.168.6.2 streaming from a host on 192.168.1.48.
Fix: client-initiated hole-punching. The client sends PUNCH_MAGIC datagrams from its data socket
to the host's advertised data port (Welcome.udp_port); that opens the firewall/NAT return path and
lets the host learn the client's OBSERVED source (the NAT-translated address, not the client's
reported private one). The host (UdpTransport::connect_via_punch) waits ≤2.5s for the first punch
and streams there, falling back to the client-reported address for clients that don't punch
(flat-LAN behaviour unchanged). The client keeps a low-rate keepalive so a stateful firewall's idle
timeout can't close the path during a static, low-bitrate scene. Wired into client-rs and the
NativeClient connector (covers the Linux + Apple clients; the Apple app needs an xcframework rebuild
to pick up the new core).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
run-headless-kde.sh gated KWin readiness on `$ROOT/target/release/punktfunk-host
probe-compositor`, else `cargo run`. On an RPM/.deb install ROOT resolves to /usr/share (no
target/ tree) and there's no Cargo.toml either, so the probe could never succeed: the session
unit hit its 30s readiness timeout, exited, and systemd restart-looped it forever — KWin never
reached the plasmashell step, so the streamed virtual output was an empty black desktop.
Add a `command -v punktfunk-host` branch (the packaged /usr/bin binary) between the source-tree
and cargo-run fallbacks. Verified live on the Fedora 44 KDE host: session goes stable
(NRestarts 0), plasmashell comes up, and a client streams the real desktop.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Automatic signing during the iOS archive resolved to App *Development* (wanted
an Apple Development cert + tried to revoke the account's orphaned one, and no
dev profile) — wrong for App Store. Switch to MANUAL distribution signing:
import an App Store provisioning profile from IOS_PROFILE_B64, read its
Name/UUID, install it, and archive with CODE_SIGN_STYLE=Manual + Apple
Distribution + that profile; export with manual signingStyle +
provisioningProfiles map. Step self-skips until IOS_PROFILE_B64 is set.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Apple Distribution identity has its key + intermediate + valid dates (it's
in 'Matching identities') but stayed out of 'Valid identities only' — a trust
strictness (most likely a pending online revocation check on an hour-old cert)
that codesign/xcodebuild do NOT enforce. Gate the iOS step on the MATCHING list
so the archive actually attempts signing, and print 'security verify-cert -p
codeSign' in the import step so the exact trust verdict shows if it still balks.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The iOS Apple Distribution identity imported WITH its private key (it's a
'Matching identity') but was dropped from find-identity -v — i.e. an untrusted
chain: the WWDR G3 intermediate it chains through didn't land, while Developer
ID's DeveloperIDG2CA did. The fetch was a single 'curl || warn' with no retry, so
a transient miss silently breaks iOS only. Retry each intermediate 3x, and print
the runner UTC date + whether the WWDR intermediate is present, to separate a
chain miss from the cert's notBefore being ahead of the runner clock.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- docker.yml: build the punktfunk-fedora44-rpm builder image (parameterized Dockerfile,
FEDORA_VERSION=44) alongside the F43/Bazzite one.
- rpm.yml: matrix the build/publish over both channels — fedora-fedora-rpm→bazzite (F43,
libavcodec.so.61) and fedora44-rpm→fedora-44 (F44, libavcodec.so.62). fail-fast:false so one
channel's break doesn't sink the other. (Bootstrap: the F44 builder image must be pushed by
docker.yml once before rpm.yml's fedora-44 job can pull it — same dance as the other images.)
- fedora-kde.md: rewrite as the reproducible RPM-install guide validated live on a Fedora 44
KDE box (RTX 4090): RPM Fusion + akmod-nvidia + the ffmpeg-free→RPM-Fusion swap for NVENC +
Secure Boot MOK enroll; the fedora-44 dnf repo + `dnf install punktfunk`; and the headless
punktfunk-kde-session.service (kwin --virtual with NO_PERMISSION_CHECKS — an interactive
Plasma session won't hand its privileged zkde_screencast protocol to an external client).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The iOS Apple Distribution cert imported (1 identity imported) but never
appeared in find-identity -v, and the iOS step then silently skipped. Make the
import step explain itself without exposing secrets or blocking the macOS
release: print secret byte-lengths + decoded p12 size + import rc, strip
stray whitespace/newlines before base64 -d, and after the partition-list warn
(not fail) with the likely cause + an incl-invalid identity list when the iOS
secret is set but yields no valid Apple Distribution identity. The shared import
step must not hard-fail on an iOS-cert problem — that would also block the
proven macOS DMG path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Three changes to make a reproducible Fedora KDE host install:
- ci/fedora-rpm.Dockerfile: parameterize the Fedora base (ARG FEDORA_VERSION, default 43) so the
same builder produces the Bazzite (F43, libavcodec.so.61) and Fedora 44 (libavcodec.so.62) RPMs.
A binary RPM is soname-coupled to its base, so each target Fedora needs its own build/channel.
- spec: install punktfunk-kde-session.service (was in the tree but never packaged) with its
ExecStart repointed from the dev source tree to the installed run-headless-kde.sh. This is the
headless `kwin --virtual` session (KWIN_WAYLAND_NO_PERMISSION_CHECKS=1) the kwin backend needs —
an interactive Plasma session refuses to hand its privileged zkde_screencast protocol to an
external client, so a dedicated session is required. Not enabled by default (kwin hosts opt in).
- ship packaging/kde/host.env as host.env.kde — the ready KWin appliance config (wayland-kde).
Validated live on a Fedora 44 KDE box (RTX 4090): KWin virtual output + zero-copy dmabuf->CUDA->NVENC.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Prepares the iOS/TestFlight path. The runner has the iOS 26.5 SDK but no
signing identities, so import an Apple Distribution cert+key from
IOS_DIST_CERT_P12_B64 / IOS_DIST_CERT_PASSWORD into the same throwaway keychain
(the WWDR intermediates already fetched chain it). The iOS archive uses
automatic signing (-allowProvisioningUpdates + the ASC key creates/downloads the
App Store profile against the present cert, so no keychain-write that would hit
the macOS -61). Re-assert the keychain on the search list like the macOS sign
step. Until the secret is set the step self-skips with a warning, so it stays
green. Still needs an App Store Connect app record for io.unom.punktfunk to
upload.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The macOS Developer ID DMG path is green (signed + notarized + stapled). The
iOS/TestFlight step (already best-effort + continue-on-error) was failing on
this runner with 'iOS 26.5 is not installed' — the iOS platform SDK is a
separate Xcode component that isn't installed. Guard the step on
`xcodebuild -showsdks | grep iphoneos` and exit 0 with a warning when it's
missing, so runs are unambiguously green. Install on the runner with
`xcodebuild -downloadPlatform iOS` when iOS goes live.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
codesign --sign 'Developer ID Application' reported 'no identity found' even
though the import step's find-identity saw it: the bare lookup relies on the
default keychain search list, which doesn't reliably carry the throwaway
keychain across steps on this runner. Re-assert the search list + default
keychain in the signing step and pass --keychain "$KEYCHAIN" so the identity
search is scoped to it (it stays unlocked with a codesign-allowed partition
list from the import step, so no password is needed). Adds a find-identity
diagnostic right before signing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
xcodebuild's archive gate demands a provisioning profile for the app's
keychain-access-groups entitlement (the 'Keychain Sharing' capability) under
both automatic AND manual signing — even though a Developer ID app honours that
team-prefixed entitlement at runtime with no profile. So manual signing just
traded the -61 keychain error for 'requires a provisioning profile'.
Sidestep the gate: archive with CODE_SIGNING_ALLOWED=NO, then codesign the app
bundle directly with the Developer ID identity, hardened runtime and a secure
timestamp, applying the entitlements via --entitlements (with $(AppIdentifierPrefix)
resolved to the team prefix, which codesign won't expand). Safe because the
bundle is a single statically-linked binary — static PunktfunkCore.xcframework,
SPM static products, macOS 14 target, no Embed-Frameworks phase — so there is no
nested code to sign inside-out. No Apple Developer portal profile or new secret
needed. iOS App Store path unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The cert import now yields a valid 'Developer ID Application' identity, but
the macOS `xcodebuild archive` step still inherited the project's automatic
'Apple Development' signing via -allowProvisioningUpdates. That made Xcode try
to mint an Apple Development cert (install fails in the CI keychain,
DVTSecErrorDomain -61 'Write permissions error') and locate a 'Mac App
Development' provisioning profile for io.unom.punktfunk (none exists) —
** ARCHIVE FAILED ** before signing even happened.
A Developer ID DMG needs neither: pin CODE_SIGN_STYLE=Manual + the Developer ID
identity + no profile, mirroring what the export step already does. The app is
non-sandboxed and its only entitlement (keychain-access-groups, team-prefixed)
is authorized by the Developer ID team, so no provisioning profile is required.
ENABLE_HARDENED_RUNTIME=YES is already set, so notarization stays happy.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The oven/bun:1 image is Debian-slim and ships no CA bundle, so
actions/checkout's git-over-HTTPS fetch died with 'Problem with the SSL
CA cert (path? access rights?)' — curl error 77 (no CA bundle file),
not an untrusted cert; git.unom.io serves a public Let's Encrypt cert.
The rust/deb/rpm builder images already install ca-certificates; do the
same in the two slim bun jobs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The stray blank line after build_primary_config tripped cargo fmt --all
--check in CI. Formatting only, no code change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The pointer/keyboard injector is host-lifetime (one EIS connection for every punktfunk/1
session), so its existing release_all only fires on EIS disconnect — never when a *client*
session ends. A button still down at an abrupt client disconnect therefore stayed latched in
the compositor: Mutter keeps the destroyed press's implicit pointer grab, so after reconnect a
stuck left-button-down turns every motion into a drag (windows move, text selects) while a
fresh click's press is swallowed — clicking buttons and text inputs does nothing. Only the one
held button is affected; keyboard and the other buttons are fine, exactly as reported.
Fix: input_thread now tracks the buttons/keys the client holds and, when the session ends,
synthesizes the matching up-events through the host-lifetime injector (whose EIS connection —
and the dangling grab — outlive the session). Backend-agnostic (normal inject path), so it
covers libei/EIS, wlr and uinput alike.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The libei backend forwarded mouse wheel only via scroll_discrete (120-per-detent).
Mutter floors a sub-detent delta — a trackpad, a precise/high-res wheel, or a
fractional smooth-scroll event — to zero whole clicks, so small scrolls never land and
you have to spin the wheel a lot before anything moves. Emit the continuous `scroll`
axis (logical px, ~15 px/detent) alongside the discrete steps, matching the wlroots
backend's 15-px/notch behaviour, so every delta moves proportionally while full
detents still drive line/page scrolling.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
After a session ends, the Mutter backend (with PUNKTFUNK_MUTTER_VIRTUAL_PRIMARY=1)
re-asserted the physical monitor layout with an explicit ApplyMonitorsConfig. On
Mutter 50 + NVIDIA that monitor reconfig — issued while the just-removed high-refresh
virtual output is still tearing down — SIGSEGVs gnome-shell. Observed live on
home-worker-3: the teardown ApplyMonitorsConfig returns "recipient disconnected from
message bus" (the shell died mid-call), GDM's crash-loop guard then drops to the
greeter and STAYS there, so org.gnome.Mutter.RemoteDesktop/DisplayConfig vanish and
every subsequent reconnect fails with RemoteDesktop.CreateSession ServiceUnknown —
i.e. "after a disconnect I can't reconnect anymore."
make_virtual_primary applies an APPLY_TEMPORARY config, which Mutter reverts on its
own once the virtual output disappears and our DisplayConfig connection closes. So the
explicit restore was both redundant and the crash trigger: drop it, drop the dc_pre
connection at teardown, and let Mutter revert the temporary config itself. Setup is
unchanged (the virtual output is still made primary so the desktop lands on the
streamed surface). Removes the now-unused to_apply_logicals/apply_config helpers.
Verified live on home-worker-3 (5120x1440@240, VIRTUAL_PRIMARY=1): 6/6 back-to-back
connect/disconnect cycles streamed cleanly with gnome-shell holding the same PID
throughout (previously it crashed within the first few disconnects).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Saved host cards now show a presence dot — green when the host is advertising on
the LAN right now, grey when not seen. Cross-references each StoredHost against the
live mDNS discovery set (HostDiscovery). No host changes: the host already
advertises _punktfunk._udp with a stable id + cert fingerprint, which the client
already browses.
- StoredHost.matches(DiscoveredHost): fingerprint-first (survives a DHCP address
change), address:port fallback. The discovered-section dedup now uses the same
match, so a saved host whose IP changed no longer also shows up as a stranger.
- HostCardView gains an isOnline presence dot (accessibility-labelled).
- HomeView.isOnline recomputes on every @Published discovery change, so the dot
tracks hosts joining/leaving the network live.
Online detection is LAN-scoped by design: a remote/cross-subnet host that doesn't
advertise here shows grey ("not seen"), not a false "offline". Swift-only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Live confirmation on worker-3: the flash was Mutter's CORRUPTED, size-0
cursor-update buffers (chunk_flags=CORRUPTED) carrying recycled old frames —
drained=1 always, so latest-frame-only draining wasn't the lever, the CORRUPTED
skip was (OBS issue 8630). Demote the verbose drain diagnostic to a rate-limited
debug line and document the root cause inline. Validated: zero-copy back on GNOME
(dmabuf->CUDA, 5120x1440) AND flash-free with FORCE_SHM off.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Deep research (OBS Studio's linux-pipewire, Mutter bug tracker) found the GNOME
stale-frame flash is a buffer-RECYCLING race, not damage (Mutter sends whole
frames, no SPA_META_VideoDamage) and not buffer count. OBS's proven fix is
latest-frame-only dequeue: each process callback, drain ALL queued PipeWire
buffers, requeue the older ones, and consume only the NEWEST — plus skip
CORRUPTED buffers. Our code dequeued one buffer per callback (oldest-first) and
the bounded channel dropped the NEWEST when full, so during Mutter's bursty
delivery the encoder got stale frames → the flash.
Switch the process callback to raw dequeue_raw_buffer + drain-to-newest (requeue
older), extract the consume logic into consume_frame(spa_buf) sourcing datas via
the transparent Data cast, skip SPA_META_HEADER_FLAG_CORRUPTED / CORRUPTED-chunk
buffers (size-0 skip kept SHM-only so dmabuf isn't regressed), and remove the
earlier content-hash drop heuristic (it couldn't tell stale re-deliveries from
legit repeating content). Diagnostic logs drain depth + chunk/header flags.
Reverts none of the FORCE_SHM / dmabuf_fence work.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Instrumented worker-3: even on the ordered FORCE_SHM download path, Mutter
re-delivers COMPLETE OLD pool buffers — 655 frames in a 15 s session whose content
exactly matched an earlier frame (not damage-incremental; full old frames, in
runs, ~45% during motion). NVIDIA gives no fence to prevent it, so the producer
delivery can't be made clean from our side.
Detect it and drop it: hash a spatial sample of each captured frame; a frame whose
content equals an EARLIER distinct frame (vs the current one, whose duplicates pass
through) is a stale re-delivery — skip it so the encoder never emits the flash
(try_latest re-sends the last good frame; brief hold instead of a backward jump).
Runs on the CPU/SHM path (where Mutter+NVIDIA capture lives); never triggers on
static content or non-Mutter compositors (no reverts). PUNKTFUNK_KEEP_STALE=1
disables it for A/B.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
scripts/punktfunk-host.service is dev-oriented — its ExecStart references the
source tree (%h/punktfunk/target/release/punktfunk-host). When the deb/rpm ship
it to /usr/lib/systemd/user, a fresh install with no hand-rolled unit would try
to run a binary that isn't there. Rewrite the ExecStart to the installed
/usr/bin/punktfunk-host during packaging (sed in build-deb.sh + the spec); the
source unit stays as-is for from-source dev. Hosts with a custom ~/.config unit
(which shadows the packaged one) are unaffected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
cbindgen swept transport/udp.rs's `recvmsg_x` foreign import and its `MsghdrX`
#[repr(C)] struct into the generated C header — they're internal Apple-only FFI,
not part of the public C ABI, and reference socklen_t/ssize_t/iovec which the C
ABI harness doesn't include, so c_abi_harness_round_trips failed to compile.
Add them to cbindgen.toml export.exclude and regenerate the header.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Empirically, Mutter+NVIDIA dmabuf capture has NO working GPU sync — confirmed on
worker-3: explicit sync fails buffer alloc (EINVAL, no cogl sync_fd), and the
dmabuf carries no implicit fence (EXPORT_SYNC_FILE waited=false). So any dmabuf
read — zero-copy import OR mmap — races Mutter's render and flashes the buffer's
previous frame. The prior "CPU fallback" still listed DmaBuf in its buffer types,
so Mutter kept handing dmabufs and it never fixed anything (got worse).
PUNKTFUNK_FORCE_SHM=1 offers MemPtr+MemFd ONLY (no DmaBuf), forcing Mutter to
glReadPixels-download into mappable memory — which orders against its render, so
the frame is complete + current by construction (race-free). Costs the download
(~3 ms) + zero-copy; correct at 1080p/4K60. KWin/gamescope are unaffected (they
blit into the buffer, no read-before-render race) and keep zero-copy.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The previous attempt (8531135) dropped zero-copy on Mutter+NVIDIA for a sticky
CPU/SHM fallback that (a) still listed SPA_DATA_DmaBuf in its buffer types, so
Mutter kept handing dmabufs that got mmap-read UNsynced — making the flashing
worse, not better — and (b) hinged on producer explicit sync, which Mutter+NVIDIA
cannot do (`error alloc buffers` / no cogl sync_fd, confirmed in worker-3 logs).
Revert the capture restructure to the original zero-copy dmabuf path, and fix the
NVIDIA stale-frame race the RIGHT way for a producer that can't do explicit sync:
the consumer snapshots the dmabuf's implicit fence (DMA_BUF_IOCTL_EXPORT_SYNC_FILE)
and waits the producer's render before sampling (new dmabuf_fence module, ioctl
number unit-tested). Covers the GPU import and the CPU mmap read. Logs once whether
a render was actually in flight (waited=true → the driver fences and the race is
closed; false → no implicit fence, so we learn zero-copy still needs SHM here).
drm_sync (the explicit-sync primitive) is kept and verified but marked unused —
no targeted compositor produces a usable sync_fd today; ready to wire in when one
does. The Bug-2 input fix (held-key release on disconnect) from 8531135 is kept.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Deep dive into the two GNOME-only host bugs (KWin/gamescope clean):
1. Stale-frame flashes (windows at old positions, typed text reverting):
Mutter renders its virtual monitors DIRECTLY into the PipeWire buffer
pool, and NVIDIA has no implicit dmabuf fencing — our zero-copy
import raced the render and encoded each pool buffer's PREVIOUS
contents. Fix, in order of preference:
- Consumer-side PipeWire explicit sync (SPA_META_SyncTimeline): new
drm_sync module (DRM timeline-syncobj wait/signal via raw ioctls,
unit-tested incl. a live signal->wait round trip); announced
post-format via update_params (the OBS pattern — at connect time
the meta makes producers fail allocation, observed on KWin), with
a blocks=3 Buffers filter so the producer's sync pod wins; acquire
point awaited before any read (GPU import or CPU mmap), release
point signaled on every path.
- Where the producer can't do explicit sync (Mutter on NVIDIA today:
no cogl sync_fd, "error alloc buffers"), a sticky fallback flips
the capture to the synchronous CPU/shm path — Mutter's glReadPixels
download orders against its render, so frames are correct by
construction. First session pays one ~10 s probe+retry; later
sessions go straight there. Validated live on home-worker-3
(GNOME 50 + RTX 4090): clean fallback, 30 MB HEVC streamed.
- Sync is only announced on Mutter sessions (new VirtualOutput.mutter
tag): KWin+NVIDIA fails allocation when merely asked, and doesn't
need it (verified unchanged: zero-copy CUDA import + 1.1 MB/10 s).
PUNKTFUNK_EXPLICIT_SYNC=0 disables the probe outright.
2. Clicks wedged in the focused app after disconnect+reconnect: a client
vanishing mid-press left keys/buttons latched in the compositor —
Mutter keeps the destroyed EIS device's implicit grab and the focused
app stops taking clicks until restarted. EiState now tracks held
keys/buttons/touches (wire codes) and synthesizes releases through
the normal inject path before the EIS connection goes away.
GNOME hosts on NVIDIA temporarily lose zero-copy (correctness over
throughput); the moment Mutter+driver gain working explicit sync, the
sync path engages automatically and zero-copy returns.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
macOS/iOS have no recvmmsg(2), so the Mac client did one recv() syscall per
packet (non-allocating after the earlier fix, but still a syscall each — a
single-core wall at line rate that Moonlight avoids). Add the Darwin recvmsg_x(2)
batched-receive path (the recv counterpart of Linux recvmmsg): one syscall drains
up to RECV_BATCH datagrams into the reused ring. struct msghdr_x + the extern
aren't in the libc crate, so declared here (cfg target_vendor=apple).
Opt-in via PUNKTFUNK_RECVMSG_X (it's FFI we can't exercise off-Apple) with
auto-fallback to the tested scalar recv-loop on any unexpected error. Linux
recvmmsg + the non-Apple scalar loop are unchanged; apple.yml compiles the path.
Re GRO: Linux recv already batches via recvmmsg (32/syscall), so UDP GRO is only a
marginal add there and needs a recv-path redesign to split coalesced buffers —
deferred as low-ROI vs the Mac, which had no batching at all.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The host sealed every packet with ~3 heap allocations: aes-gcm's convenience
encrypt() allocates the ciphertext Vec, seal_for_wire allocates the seq||ct||tag
wire Vec, and seal_frame allocated a fresh Vec<Vec<u8>> per frame. At line rate
(~250k–500k pkt/s for 2.5–5 Gbps) that's the single-core allocator wall.
- SessionCrypto::seal_in_place uses AeadInPlace::encrypt_in_place_detached to
encrypt into the caller's buffer and write the detached tag at the end —
byte-identical to seal's ciphertext||tag, no allocation (unit-tested for byte
equality + decrypt).
- Session keeps a wire_pool the caller returns via reclaim_wires; seal_frame
seals each packet in place into the reused buffers (clear() keeps capacity), so
after warmup there's no per-packet ciphertext/wire allocation. paced_submit and
submit_frame reclaim the pool after sending.
End-to-end encrypted/lossless multi-frame tests stay green (validates the pool
reuse doesn't corrupt across frames). Next: write packetize directly into a
contiguous send buffer (kills the remaining shard allocs + GSO's coalescing copy).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
sendmmsg already batches syscalls but still builds one sk_buff per datagram —
the kernel-side wall above ~1 Gbps. UDP Generic Segmentation Offload hands the
kernel one big buffer it splits into gso_size datagrams, building ~1 GSO skb per
≤64 segments. Research (LWN/Cloudflare/Tailscale) measures ~2.4x throughput at
equal CPU and 17-44x fewer syscalls, and that sendmmsg batching alone is
insufficient — you need true segmentation offload.
Adds Transport::send_gso (default = send_batch) + a UdpTransport Linux override:
coalesces a frame's equal-size wire packets (shards are zero-padded to a constant
size, so a whole frame is one gso_size) into ≤64-segment sendmsg(UDP_SEGMENT)
calls. seal/send routes through it. Opt-in via PUNKTFUNK_GSO (new unsafe hot-path
code) with automatic fallback to sendmmsg on any GSO error (unsupported kernel/
path), latched per process. Loopback unit test validates the cmsg segmentation;
full session over loopback streams clean (0% loss). Linux-only; loopback/non-Linux
keep sendmmsg/scalar.
Next levers: in-place AES-GCM seal (kill per-packet allocs), UDP GRO on recv,
drop the sleep-pacing in favor of the kernel qdisc, jumbo MTU.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Stage 1.5: on Intel/AMD clients libavcodec's VAAPI hwaccel decodes on
the GPU; frames map to DRM-PRIME dmabufs (av_hwframe_map, zero copy)
and reach GTK as GdkDmabufTexture (BT.709 limited CICP color state —
GDK's dmabuf default is BT.601). Inside GtkGraphicsOffload that is the
decoder-to-subsurface path, direct-scanout eligible when fullscreen.
Fallback ladder, live-verified on the NVIDIA dev box: no VAAPI device
-> software decode at session start (logged reason); a mid-session
VAAPI error (e.g. broken nvidia-vaapi-driver) demotes to software and
the host's IDR/RFI recovery resynchronizes; a rejected dmabuf import
logs and the stream continues. PUNKTFUNK_DECODER=software|vaapi
overrides; the first-frame log now names the active path.
The hwaccel path is raw ffmpeg-sys FFI (ffmpeg-next wraps none of it):
hw device ctx + get_format pinned to AV_PIX_FMT_VAAPI (NONE on
mismatch so cpu-fallback never silently engages inside libavcodec),
thread_count=1, LOW_DELAY. Surface lifetime rides DrmFrameGuard into
the texture's release func — GDK runs it on both success and failure.
Needs an Intel/AMD client box (Steam Deck/Bazzite) to live-verify the
hardware path; the software path is unchanged and revalidated.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Groundwork for multi-Gbps (2.5G link here, 5G to the Mac Studio). The encoder is
pixel-rate bound, not bitrate bound, so these unblock the transport:
- MAX_BITRATE_KBPS 2G -> 8G, MAX_PROBE_KBPS 3G -> 10G (the cap was policy, not a
hardware limit — NVENC emits multi-Gbps trivially with the 2-way split).
- Welcome shard_payload 1200 -> 1452: fills a 1500 MTU, ~17% fewer packets for
free (even size, FEC-safe; negotiated so the client follows).
- PUNKTFUNK_FEC_PCT env overrides the 20% FEC default — a clean wired LAN can drop
it (every recovery shard is wire bytes+packets); 0 disables FEC.
Next: UDP GSO (the dominant lever — research shows ~2.4x throughput / ~40x fewer
syscalls; sendmmsg batching alone is insufficient) + in-place AES-GCM seal.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Hook the Linux client into the existing packaging CI:
- deb.yml builds both binaries and publishes punktfunk-host AND
punktfunk-client to the Gitea apt registry; new
packaging/debian/build-client-deb.sh mirrors the host script
(shlibdeps auto-Depends — GTK4/libadwaita/SDL3/FFmpeg/PipeWire
sonames; no NVIDIA filter, the client links no CUDA). Built and
inspected locally on Ubuntu 26.04.
- punktfunk.spec gains a "client" subpackage (binary + desktop entry +
udev rule); rpm.yml's publish loop picks it up unchanged.
- New shared assets: packaging/linux/io.unom.Punktfunk.desktop and
scripts/70-punktfunk-client.rules — DualSense hidraw uaccess (USB +
Bluetooth, steam-devices style) so SDL's HIDAPI driver gets
touchpad/motion/lightbar/triggers instead of degrading to evdev.
- Builder images learn the client link deps (rust-ci already had
them; fedora-rpm adds gtk4/libadwaita/SDL3-devel) with idempotent
install steps in deb.yml/rpm.yml since jobs run against the
previous push's image.
Workspace check CI (build/clippy/test) already covers the crate since
f09def4.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
6b5ee9f added a libc-based batched recv_batch for the Apple/BSD targets
(cfg(all(unix, not(target_os = "linux")))) but left libc declared only under
cfg(target_os = "linux"). The macOS host build pulls libc in transitively so it
compiled, but the iOS/tvOS cross-compiles (no transitive libc, dev-deps off) failed
with E0433 "cannot find crate libc", breaking the full xcframework build. Widen the
gate to cfg(unix): libc is now used by sendmmsg/recvmmsg on Linux AND recv() on the
other unix (Apple/BSD) targets.
Verified: cargo build --release -p punktfunk-core --features quic for
aarch64-apple-ios, x86_64-apple-ios, and aarch64-apple-tvos (-Z build-std) all link.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Applied via Xcode's recommended-settings upgrade and distribution prep:
- LastUpgradeCheck / scheme LastUpgradeVersion 2650 -> 2700.
- DEVELOPMENT_TEAM (F4H37KF6WC) hoisted to the project-level build configs; the
now-redundant per-target copies are dropped (all targets inherit it).
- ENABLE_HARDENED_RUNTIME = YES on the macOS app target (required for Developer ID
notarization). Signing stays Apple Development + Config/Punktfunk.entitlements.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The experimental stage-2 presenter (CAMetalLayer + display link) stuttered badly
in fullscreen but ran fine windowed. render() runs on the display-link / MAIN
thread and calls layer.nextDrawable(), which blocks that thread until a drawable
frees. With the layer's own displaySyncEnabled left on (default), present also
waits for the hardware vsync, so the block serializes the main thread to the
display — windowed, the WindowServer's looser compositing hides it; fullscreen's
tighter, more-direct path exposes it as judder. (Apple dev-forum guidance:
displaySync off measurably reduces nextDrawable() blocking.)
- displaySyncEnabled = false (macOS-only): the display link is already the per-
vsync pacing source, so the layer's redundant vsync wait only adds the stall.
- maximumDrawableCount = 3 (explicit): more in-flight headroom before
nextDrawable() has to block on the main thread.
Swift-only (no core/ABI change → no xcframework rebuild). Validated: swift build;
swift test (39 passed, 0 failures).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The batched `recvmmsg` recv path was Linux-only; macOS fell back to the trait
default, which calls the scalar `recv` — a fresh `vec![0u8; 2049]` allocation
(plus zeroing and a copy) PER PACKET on the single receive thread. At line rate
that alloc/free churn, not the syscall, was the single-core wall: measured the
real Mac client topping out ~315 Mbps and dropping the session at 800, while a
Linux client (recvmmsg) held a clean 1 Gbps against the same host, and Moonlight
(batched recv) does 900 on the same Mac.
Add a `cfg(all(unix, not(linux)))` `recv_batch` that drains up to RECV_BATCH
datagrams per call with `libc::recv(MSG_DONTWAIT)` straight into the caller's
reused ring buffers — no per-packet allocation or copy. Still one syscall per
datagram (a future `recvmsg_x` batch would cut that too), but it removes the
dominant cost. Linux recvmmsg path and the Windows/loopback default unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Fixes the intermittent first-connect freeze. The host streams infinite GOP — one
opening IDR, then P-frames only (recovery keyframes just on loss) — so when the
client's decoder wedges on the cold first session (a lost/corrupt opening IDR, a
bad early P-frame) the picture stays frozen until the far-off next keyframe. The
client had no way to ask for one; now it does.
Add a RequestKeyframe control message (client -> host, reliable control stream),
mirroring Reconfigure:
- core: quic.rs RequestKeyframe (type 0x03) + roundtrip test; client.rs
CtrlRequest::Keyframe + NativeClient::request_keyframe; abi.rs
punktfunk_connection_request_keyframe (header regenerated).
- host: m3.rs decodes it in the control loop and signals the encode loop, which
coalesces a burst and calls enc.request_keyframe() — wiring the existing
NvencEncoder hook (force_kf -> next frame pict_type=I), the same recovery the
GameStream path already had via force_idr.
- apple: PunktfunkConnection.requestKeyframe(); StreamPump (stage-1) requests on
layer.status==.failed; Stage2Pipeline (stage-2) on a sync submit failure and on
the async decode-error callback via a thread-safe KeyframeRecovery. All
throttled to <=1/250ms (the decode stays wedged for several frames until the IDR
lands, so per-frame requests would flood the control stream).
Self-healing: a lost recovery IDR is re-requested after the throttle; the host
coalesces bursts into a single IDR.
Validated: cargo fmt + clippy clean; core + host test suites green (incl. new
request_keyframe_roundtrip); swift build + test (39 passed); xcframework rebuilt
(all 5 slices), header regenerated with no unrelated drift.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
deb.yml runs in the Ubuntu rust-ci image whose /bin/sh is dash, where the bash
substring `${GITHUB_SHA::8}` is a "Bad substitution" — the deb build failed at the
Version step every run. Compute the short SHA with `cut` instead. (rpm.yml ran fine
because the Fedora image's /bin/sh is bash, but fix it the same way for robustness.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The host requests a 32 MB SO_SNDBUF, but the kernel clamps it to net.core.wmem_max
(~416 KB on a stock box) — so high-bitrate frames overflow the socket buffer and
the host drops a large fraction of packets on send (measured 28.5% loss / 54k
dropped at 1 Gbps to a clean LAN client on a fresh Bazzite box). scripts/99-punktfunk-net.conf
fixes it (32 MB caps) but the packages never installed it. Ship it to
/usr/lib/sysctl.d/ (auto-applied at boot by systemd-sysctl) and apply it in the
deb/rpm postinst. This is the dominant cause of the sub-Gbps ceiling on an
untuned host.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Gitea GPG-signs the repo metadata but not the individual packages, while its
auto-served bazzite.repo sets gpgcheck=1 — so `rpm-ostree install` fails with
"could not be verified" on our unsigned RPMs. Document writing the repo
explicitly with gpgcheck=0 + repo_gpgcheck=1 (verify the signed metadata, which
carries each package checksum) instead of curling the served .repo. Note the
TLS-only fallback and that per-package signing is future hardening.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mirrors the apt pipeline for Fedora Atomic / Bazzite. New `rpm` workflow builds
the host RPM in a Fedora 43 builder image (ci/fedora-rpm.Dockerfile — matches
Bazzite's libavcodec.so.61, with a self-contained 16-symbol libcuda link stub so
no NVIDIA packages are needed in CI) and uploads to Gitea's public RPM registry
(group "bazzite") on every main push (rolling 0.0.1-0.ciN.<sha>) and v* tag
(clean X.Y.Z-1). Bazzite hosts then track it with `rpm-ostree upgrade`.
- packaging/rpm/build-rpm.sh: git-archive tarball + rpmbuild (--nodeps, since the
toolchain is rustup + dnf, not RPMs); copies to dist/, asserts no cuda/nvidia leak.
- punktfunk.spec: overridable pf_version/pf_release for CI snapshots; exclude
libcuda.so from auto-Requires (NVENC/EGL come from the driver, out of band) —
same NVIDIA filter as the .deb; fix a bogus changelog weekday.
- docker.yml builds+pushes the new fedora-rpm image; packaging README + rpm/README
document the rpm-ostree install/update path (recommended option).
Builder image seeded to the registry so rpm.yml's first run finds it. RPM build +
clean-Requires verified locally in the image (libavcodec.so.61 / libavutil.so.59,
no cuda).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The client identity prompted for Keychain access on every launch/rebuild. Root
cause: the macOS app target was ad-hoc signed (CODE_SIGN_IDENTITY = "-"), and
the identity lived in the file keychain whose "Always Allow" ACL is bound to the
app's exact code signature (cdhash for ad-hoc). Every rebuild changed the binary
-> changed the cdhash -> the ACL no longer matched -> re-prompt.
- Sign the macOS target with Apple Development (team already set) instead of
ad-hoc, so the designated requirement is identity-based and stable across
rebuilds.
- Move the identity to the data-protection keychain (kSecUseDataProtectionKeychain)
gated by a team-scoped keychain-access-group entitlement — access is granted by
the app's entitlement, not a per-binary ACL, so it's prompt-free and survives
rebuilds. Add Config/Punktfunk.entitlements and wire CODE_SIGN_ENTITLEMENTS into
all six app configs (macOS/iOS/tvOS).
- Unsigned / ad-hoc builds (e.g. `swift run`) lack the entitlement
(errSecMissingEntitlement) — fall back to the legacy file keychain so they still
work (with the old prompt), no hard failure.
macOS re-mints the identity on first run (the old file-keychain copy isn't in the
data-protection keychain) -> one re-pair, which is acceptable. iOS keeps its
identity (the explicit access group equals the prior default).
Validated: swift build; swift test (39 passed, 0 failures); xcodebuild
-showBuildSettings confirms Apple Development + Config/Punktfunk.entitlements.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wires up the half-built Debian packaging: build-deb.sh existed but nothing
invoked or published it. Adds a `deb` workflow that builds the release host in
the Ubuntu 26.04 rust-ci image, packages it (dpkg-shlibdeps-resolved Depends,
NVIDIA driver filtered out), and uploads to Gitea's public Debian registry on
every main push (rolling 0.0.1~ciN.<sha>) and v* tag (clean X.Y.Z). Ubuntu hosts
then track it with `apt update && apt upgrade`.
Also: box-setup docs (packaging/debian/README.md), a pointer from the packaging
README, ignore dist/, and drop backticks from the package Description (the
unquoted control heredoc ran them as a command substitution).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Everything the macOS app does that stage 1 lacked, before any new
feature work (user directive):
- Input capture is now a deliberate, reversible STATE (Moonlight-
style): engaged on stream start and click-into-video (the engaging
click is suppressed), released by Ctrl+Alt+Shift+Q (toggles) or
focus loss; held keys/buttons are flushed host-side on release;
cursor hiding + shortcut inhibition follow the state; HUD hint when
released. Per-session window handlers disconnect with the page.
- Gamepads: app-lifetime SDL service (GamepadManager parity) — pad
list + "Forwarded controller" pin in Settings (auto = most recent),
"Automatic" pad TYPE resolves from the physical pad at connect;
DualSense touchpad contacts + ~250 Hz motion samples on the 0xCC
plane (Swift GamepadWire scale constants); feedback grows adaptive-
trigger replay and player LEDs via raw DS5 effects packets (the
wire's 11-byte blocks drop into SDL_SendGamepadEffect verbatim);
held pad state zeroed on pad switch/detach. sdl3 "hidapi" feature.
- Microphone uplink: PipeWire capture -> Opus 20 ms -> 0xCB datagrams
(validated live: host received 711 mic packets), Settings toggle.
- Speed test per saved host (Swift's "Test Network Speed…"): 2 s
probe burst, goodput/loss + recommended ~70 % bitrate, one-tap apply.
- Settings: host compositor preference (sent in the Hello), native-
display resolution/refresh resolved from the window's monitor at
connect (new default), bitrate ceiling to 3 Gbit/s.
- Hosts page: saved/trusted hosts section for direct pinned reconnect
(mDNS not required), rebuilt on every page return.
Deliberately not ported: audio device pickers (PipeWire routing owns
this on Linux), resize-to-request_mode (not wired in Swift either),
pointer-lock relative mouse (stage-2 presenter, needs raw Wayland).
DualSense fidelity needs a physical pad to live-verify.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three native-client bugs isolated against a stock Moonlight client (which
stays connected / keeps input working under the same actions):
- Connection drops mid-stream: the quinn endpoints (host + client) ran with
default transport config, so keep_alive_interval was OFF. Any quiet stretch
(no input, audio muted/stalled, a capture hiccup, a mode change) let the
idle timer expire and quinn closed the session -> next_au=Closed -> "Session
ended". Moonlight's ENet sends keepalive pings; we sent nothing. Add a shared
TransportConfig (keep-alive 4s under an explicit 20s idle timeout) to both
endpoint::server_from_der and endpoint::client_pinned_with_identity.
- Reconnect input dead (macOS): the session-start auto-capture one-shot was
consumed even when engageCapture(fromClick:false) was refused (window not key
yet at the instant of reconnect), with no retry -> capture stayed off and
input never forwarded. Clear the one-shot only on a successful engage, and
retry on NSWindow.didBecomeKey. Stays scoped to session start, so it does not
resurrect the rejected auto-grab-on-activation behavior.
- Reconnect input dead (iOS): wasCapturedOnResign leaked stale state across
sessions and the foreground-restore could fire before this session's
InputCapture was wired (setForwarding no-ops on nil). Reset it per session in
start() and guard the didBecomeActive restore on inputCapture != nil.
Validated: cargo build -p punktfunk-core --features quic; swift build;
swift test (39 passed, 0 failures); xcframework rebuilt (all 5 slices), no
ABI/header drift.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
NVENC ran CBR (bit_rate == max_bit_rate, rc=cbr) but never set rc_buffer_size,
so it used a loose default VBV. A high-motion P-frame was then allowed to spike
to many times the average frame size; the extra packets overflow the depth-2
send queue (newest frame dropped) and the kernel UDP buffer (WouldBlock drops),
which the client sees as framedrops/jitter — and on the infinite-GOP GameStream
path as old/stale frames flashing until the next RFI.
Set a tight ~1-frame VBV (rc_buffer_size = bitrate/fps) so the encoder holds
frame size roughly constant and absorbs motion as a momentary QP/quality dip
instead — the Sunshine/Moonlight low-latency model. Tunable via
PUNKTFUNK_VBV_FRAMES (default 1.0); larger trades burst tolerance for motion
quality. Fixes both the punktfunk/1 and GameStream paths (shared encoder).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The MouseMoveAbs wire contract packs the client coordinate-space size
as (width << 16) | height in `flags` (same as touch); injectors
normalize against it and drop the event when it is zero. The GTK
client sent flags=0, so KWin's libei path refused every motion
(`emitted=false`) — found via the first real user test from
home-worker-3.
- ui_stream: send_abs() packs the negotiated mode into flags for
motion + click-position events.
- core input.rs: document the contract on MouseMoveAbs itself (it was
only implied by TouchDown's doc).
- client-rs --input-test: add a MouseMoveAbs sweep so the absolute
path stays covered — Moonlight and the Mac client only send relative
motion, which is why this gap survived every prior live test.
Validated live against serve --native: kind=MouseMoveAbs emitted=true.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Baked into the rust-ci image, plus an idempotent apt step in the rust
job itself — ci.yml runs against the previous push's image (docker.yml
bootstrap note), so the image change alone would leave this push and
the next one red.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New crate crates/punktfunk-client-linux (binary punktfunk-client), the
native Linux client on the Option A architecture (2026-06-12 research):
- GTK4/libadwaita shell linking punktfunk-core directly (no C ABI):
mDNS host list, TOFU fingerprint prompt, SPAKE2 PIN pairing dialog,
preferences (mode/bitrate/gamepad/shortcut capture), stats overlay,
--connect host[:port] for scripting.
- Video: FFmpeg software HEVC decode (LOW_DELAY, slice threads) ->
RGBA -> GdkMemoryTexture inside GtkGraphicsOffload (the dmabuf
subsurface path lights up when VAAPI lands; black-background keeps
fullscreen scanout-eligible).
- Audio: Opus -> PipeWire playback stream, the host virtual-mic's
adaptive jitter ring inverted.
- Input: keyboard as the exact inverse of the host VK table (evdev
keycodes, layout-independent; unit-tested), absolute mouse through
the Contain-fit transform, WHEEL_DELTA(120) scroll, compositor
shortcut inhibition while streaming, Ctrl+Alt+Shift+Q release chord,
F11 fullscreen. SDL3 gamepad capture (single pad-0 model) + rumble
and DualSense lightbar feedback on the same thread.
- Session pump owns video+audio pulls; the gamepad thread owns
rumble+hidout — possible because NativeClient's plane receivers are
now mutexed, making it Sync (Arc-shared, compiler-verified per-plane
contract instead of the ABI's manual assertion).
- Linux-gated deps + a stub main keep cargo build --workspace green on
macOS.
Validated live against serve --native on this box: 1920x1080@60,
locked 60 fps, capture->decoded p50 ~6.4 ms (software decode, debug
build). Teardown keys off AdwNavigationPage::hidden — NavigationView
push fires a transient unmap/map cycle that must not end the session.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
An identified-but-unpaired device that knocks on a pairing-required host is now
held as a pending request the operator approves from the web console — pairing it
with no PIN fetched out of band — instead of a flat reject.
- core: Hello gains an optional trailing device name (len u8 || UTF-8, ≤64,
same trailing-back-compat pattern as compositor/gamepad/bitrate). client-rs
--name sends it; the connector sends None (fingerprint-derived label).
- native_pairing: in-memory pending queue (note_pending dedups by fingerprint,
evicts the least-recently-active past a 32 cap, 10-min TTL); approve_pending
pins the fingerprint, deny drops it. Names are sanitized (strip control/ANSI/
bidi — untrusted wire input); add()/remove() roll back in-memory on a persist
failure; pairing clears any stale pending knock.
- m3: the require_pairing gate records the knock (sanitized label) before
rejecting; anonymous (certless) clients record nothing.
- mgmt: GET /native/pending, POST /native/pending/{id}/approve (optional {name})
and /deny; OpenAPI + tests; docs/api/openapi.json regenerated.
- web: a "Waiting for approval" section on the Pairing page (live-poll, Approve/
Deny, error-surfaced via QueryState); en+de strings.
- Also completes an in-progress NativeClient Sync refactor (receivers behind
per-plane mutexes) that was left half-applied in the tree.
Adversarially reviewed (4 lenses + 3-vote verify); the confirmed findings are
fixed here. Validated live on the GNOME box: knock (with a wire name, and a
malicious ANSI/bidi name that got neutralized) → pending → approve → the same
identity streams real video. Full workspace tests + clippy + fmt green; web tsc
clean. Roadmap §8b-1 marked done; §8b-2 (peer-push approval) is the client
follow-up. See docs-site pairing page.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
exportArchive's signing lookup consults the default keychain; search
list membership alone leaves the (valid) identity invisible to it.
Restored to login.keychain in cleanup.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
With -allowProvisioningUpdates, exportArchive prefers cloud-managed
Developer ID signing; the App-Manager API key can't ("Cloud signing
permission error") and the valid local identity is never tried.
signingStyle=manual + explicit signingCertificate, cloud flags off
this step (archive keeps them for profile fetch).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fresh boxes lack the Developer ID / WWDR intermediates; without the
issuing chain the imported identity is invalid and xcodebuild says
"No signing certificate Developer ID Application found".
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The app shell's only navigation was the desktop sidebar (`hidden … sm:flex`), so
on phones (< sm) it was hidden with no replacement — you couldn't navigate at all.
Add a responsive mobile layout shown only below `sm`: a top bar (brand + language
switcher) and a fixed bottom tab bar with the five nav items (icon + label). The
desktop sidebar is unchanged. Page content gets bottom padding so the fixed bar
doesn't cover it, and the bar respects the iOS `safe-area-inset-bottom`.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
release.yml (v* tags / dispatch, macos-arm64 runner): universal mac +
iOS xcframework -> xcodebuild archive -> Developer ID export ->
notarytool + staple -> dmg on the Gitea release; iOS archive uploads
to TestFlight (app-store-connect/upload). Per-run throwaway keychain;
ASC API key authenticates notarization, upload, and automatic-signing
profile fetch. macOS App Store lane deferred (needs App Sandbox);
tvOS deferred (tier-3 Rust targets).
All app targets now share bundle ID io.unom.punktfunk — ONE App Store
listing with universal purchase (decided pre-submission; effectively
unchangeable after). ITSAppUsesNonExemptEncryption=false declared
(standard-algorithm AES-GCM, exempt).
build-xcframework.sh resolves Apple toolchains itself: cargo's HOST
artifacts (proc-macros, build scripts) are loaded by the running OS,
and a newer-than-OS beta Xcode ld emits LINKEDIT layouts dyld rejects
("mis-aligned LINKEDIT string pool" -> misleading E0463) — so prefer
a non-beta Xcode for everything, fall back to CLT for mac-only slices
(env untouched: an explicit DEVELOPER_DIR=<CLT> trips xcrun's license
check), refuse iOS/tvOS without a real Xcode (CLT has no iOS SDK).
The runner plist no longer injects DEVELOPER_DIR for the same reason.
punktfunk_Logo.icon: dropped the Xcode-27-beta-only Icon Composer
features (refractivity, specular-location) — 26.5's actool crashes on
them, and store builds must use release Xcode. Visual delta is the
refraction/specular nuance only; re-author when 27 ships.
Validated on home-mac-mini-1 with Xcode 26.5: mac+iOS xcframework
slices, unified bundle IDs, signing-free app build.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Split the monolithic ContentView into focused view files — a pure structural refactor
with no behavior change (verified: builds macOS/iOS/tvOS, the test suite is green, and a
fidelity review against the original found no discrepancies):
- ContentView (272): the coordinator — owns the session model / host store / discovery,
switches home<->session, holds the connect logic (it reads @AppStorage) + the dev
hooks, and the stream builder (whose stable identity across awaiting-trust->streaming
must NOT move — it stays here).
- HomeView (251): the hosts grid + navigation + toolbar + sheets + "On this network"
discovery section + empty state.
- HostCards (158): HostCardView + DiscoveredCardView, sharing a CardMetrics struct
(dedupes the platform-tuned sizing the two cards had copy-pasted).
- TrustCardView (80): the TOFU prompt + fingerprint formatting.
- StreamHUDView (67): the streaming overlay HUD.
State flows idiomatically: @StateObject (ContentView) -> @ObservedObject in subviews,
@State -> @Binding; the connect logic is passed as closures. Sheet placement is
preserved — the pairing/speed-test sheets stay on the outer body so they survive the
trust->home transition.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A 6-agent adversarial audit of the client (11 confirmed of 39 findings, the rest
filtered) drove these:
- fix: SessionAudio ring buffer — guard a write larger than the ring (would push
readIdx past writeIdx and corrupt the buffer; never happens, but guard not corrupt).
- fix: CADisplayLink retain cycle (stage-2 presenter) — a weak-target DisplayLinkProxy
so the view can deallocate (the link retains its target); stage-2 teardown added to
both StreamView/StreamViewController deinits as a safety net.
- fix: GamepadFeedback deinit { flag.stop() } — the drain thread holds the connection
strongly and self weakly, so an abrupt teardown without stop() would leak it.
- refactor: centralize the 12 UserDefaults/@AppStorage key literals (scattered across
8 files) into one DefaultsKey enum — a typo silently splits a setting's reader from
its writer.
- docs: RumbleRenderer @unchecked Sendable invariant; the HID digit-row table; the
stage-2 layer compositing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The high-refresh teardown SIGSEGV was caused by ApplyMonitorsConfig disabling the
still-actively-captured high-refresh virtual output. Reorder teardown: Stop the screencast
FIRST (Mutter removes the virtual + auto-reverts the temporary config), then re-assert the
physical layout once the virtual is gone. Never reconfigure a live virtual CRTC.
With this, PUNKTFUNK_MUTTER_VIRTUAL_REFRESH=1 is stable: validated at 5120x1440@240 on
Mutter 50 + NVIDIA — virtual output Meta-0@240, real 240fps, gnome-shell survives back-to-back
sessions + teardowns, physical (HDMI-1) restored each time.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace the dev/agent-log pages with a proper user-facing doc set:
- Getting Started: Introduction (rewritten), How It Works, Quick Start.
- Host Setup: Requirements, then clean per-platform guides — Ubuntu GNOME,
Ubuntu KDE, Fedora KDE (new), Bazzite (rewritten) — plus Running as a Service
(desktop / headless GNOME / headless KDE).
- Connecting: Clients overview, Moonlight, Pairing & Trust.
- Configuration: host.env reference, Host CLI, Troubleshooting.
- The dev/design notes (architecture, roadmap, the deferred design specs, CI)
move to a clearly-separated "Project & Internals" nav section.
Removes the superseded box-specific pages (gnome-box, headless-box, linux-setup,
overview). status.md (the internal progress tracker, with box IPs) is kept as a
file but dropped from the public nav. Site builds clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Pinning the virtual output to a high client refresh via RecordVirtual "modes" works
mid-stream, but a high-refresh virtual CRTC SIGSEGVs gnome-shell on session TEARDOWN
(observed at 5120x1440@240) — taking down the whole GNOME session, so subsequent connects
fail with RemoteDesktop ServiceUnknown.
Gate it behind PUNKTFUNK_MUTTER_VIRTUAL_REFRESH, default OFF — Mutter then derives the
virtual monitor's refresh from the PipeWire framerate (60Hz, stable). The >60Hz path stays
in-tree for investigation; re-enable once the teardown crash is understood.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
RecordVirtual without a "modes" property makes Mutter derive the virtual monitor's refresh
from the PipeWire stream framerate and default to 60 Hz — so a 240 Hz client mode rendered
at 60 (the encoder just padded to 240 with duplicate frames). Pass an explicit "modes" entry
(size + refresh-rate + is-preferred) so Mutter creates the virtual monitor at the client's
exact WxH@Hz. Mutter >= 47; older Mutter ignores the unknown key (60 Hz fallback, no regression).
Confirmed first via raw D-Bus on the box, then validated end-to-end: the virtual output
Meta-0 reports 1920x1080@240.00 and the host encodes 480 *immediate* (real, not paced)
frames per 2 s.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Document the gamescope multi-user (independent-desktops) research and defer it:
the current shared host-lifetime input/audio/mic vs the per-session plumbing it
would need — per-instance EIS sockets + a per-session injector + per-session
null-sink audio routing + per-session mic — and why it's not worth it now (a large
multi-file refactor for the niche multi-user-on-one-box case, while the common
multi-device scenario is already covered by the shared-desktop multi-view
concurrency that landed). New gamescope-multiuser.md + roadmap section 14
(concurrent sessions: multi-view done, multi-user deferred).
Also park render->capture in section 12: pipewire-rs 0.9.2 exposes no
buffer-meta / raw-pointer / stream-timing API, so reading SPA_META_Header.pts
would need raw spa_sys FFI into the working capture hot path — disproportionate
for the smallest glass-to-glass term; g2g is effectively complete as
capture->present (the stage-2 presenter measures it).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Opt-in (Settings -> Presenter; `punktfunk.presenter`, default stage-1). Stage-1's
AVSampleBufferDisplayLayer decodes AND presents internally with no per-frame
callback, so neither decode nor present can be stamped or hand-paced. Stage-2
takes explicit control:
- VideoDecoder: VTDecompressionSession, async output callback stamps
decode-completion, session rebuilt on every IDR / format change. Unit-tested
(testVideoDecoderAsyncCallbackDeliversPixels).
- MetalVideoPresenter: CAMetalLayer + CVMetalTextureCache + a runtime-compiled
BT.709 limited-range NV12->RGB shader, present at the next vsync. The
CVMetalTextures + pixel buffer are held until the GPU completes.
- Stage2Pipeline: pump thread -> decoder -> newest-ready 1-slot ring; the hosting
view's display link drains it once per vsync and stamps capture->present
(the display-link target time projected into CLOCK_REALTIME).
- LatencyMeter gains record(ptsNs:atNs:offsetNs:); the HUD shows a capture->present
(glass-to-glass, modulo host render->capture) line, skew-corrected via
clockOffsetNs. Measured live ~11 ms p50 vs ~2.2 ms capture->client.
- StreamView / StreamViewIOS host the CAMetalLayer as a sublayer + a CADisplayLink
(NSView.displayLink on macOS) when stage-2; input capture + HUD unchanged. The
session-active gates switch from `pump != nil` to `connection != nil` so capture
engages without a StreamPump.
Validated: builds macOS/iOS/tvOS; the decode half is unit-tested; the Metal
present is live-validated on glass (correct image + the capture->present number).
Colorspace is BT.709 SDR for now; 10-bit/HDR + a pacing policy are later.
Plan: docs-site/content/docs/apple-stage2-presenter.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
website/cms deploy to the unom-1 DMZ VM (192.168.50.50) — the
website README's home-main-2 mention is stale. Caddy upstream fixed
in unom/reverse-proxy 6ae79b8, firewall port in unom/infra 9670aa8.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Keeping the physical monitor enabled as a secondary let the cursor, windows, and keyboard
focus land on it — relative pointer motion wandered off the streamed surface, so on the
client the cursor "disappeared" and clicks/keys went nowhere visible. Omit the physical
outputs from ApplyMonitorsConfig so Mutter disables them for the session; everything is
confined to the streamed virtual output. Restored on teardown.
Validated on-box: mid-session DisplayConfig shows only the virtual output (Meta-0) as the
sole primary; the physical (HDMI-1) is restored after the session ends.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
docker.yml gains a deploy-docs job after the image pushes: scp
compose.production.yml to ~/punktfunk-docs on home-main-2, then
docker compose pull + up over SSH — the unom/website / unom/cms
deploy pattern, same DEPLOY_* secret set (unom-ci-deploy key). Docs
bind host port 3220; the docs.punktfunk.unom.io vhost lives in
unom/reverse-proxy (306d9c0).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
On a headless GNOME host the xdg-desktop-portal RemoteDesktop Start() blocks on an
interactive "Allow remote control?" approval nobody can click, so libei input timed out
("EIS setup timed out") and neither mouse nor keyboard worked — even though video worked
(it uses Mutter's direct RemoteDesktop API).
Add EiSource::MutterEis: obtain the EIS fd from
org.gnome.Mutter.RemoteDesktop.Session.ConnectToEIS (CreateSession → Start → ConnectToEIS),
no portal and no approval. Selected for GNOME/Mutter; KWin keeps the RemoteDesktop portal,
gamescope keeps its own EIS socket.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The accept loop no longer awaits each session inline — it spawns each onto a
JoinSet, bounded by a semaphore (--max-concurrent, default 4: a NVENC session
bound; overflow clients wait in QUIC's accept backlog until a slot frees). The
QUIC handshake stays in the accept loop so a failed handshake (e.g. a pin
mismatch where the client aborts) doesn't consume a session slot or block
accepting the next client; the slow part (control handshake, pairing, the
capture/encode pipeline) runs in the spawned task.
Each session already had its own virtual output + NVENC encoder; the
host-lifetime input/audio/mic services stay shared — the natural "multiple
devices viewing/controlling the same desktop" semantic on kwin/mutter/wlroots.
gamescope's independent-desktops (per-session input/audio) isolation is a
follow-up. New M3Options.max_concurrent + the `--max-concurrent` CLI flag.
Validated live (GNOME box): two clients connected at once -> two independent
Mutter virtual outputs (720p60 + 1080p60) streaming simultaneously (39 MB +
48 MB). All 61 host tests green (the c_abi/pairing tests exercise the new loop +
the failed-handshake-doesn't-count semantics).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
runs-on: ubuntu-24.04 (the label the existing Linux runner actually
advertises — ubuntu-latest queued forever). Mac runner: strip the
docker:// default labels generate-config seeds (they override the
host-mode registration labels and make the daemon demand a Docker
engine), and ship the service as a root LaunchDaemon — macOS Local
Network privacy silently blocks LAN dials from unbundled CLI binaries
in gui/user launchd domains ("no route to host"), system daemons are
exempt. Without sudo the script leaves an interim nohup daemon. CI
surface documented in CLAUDE.md + docs-site ci.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three workflows: ci.yml (Rust workspace inside the punktfunk-rust-ci
builder image + web/docs-site build+typecheck), docker.yml (build+push
punktfunk-web, punktfunk-docs, punktfunk-rust-ci to git.unom.io — host
and native clients stay un-dockerized by design), apple.yml (host-mode
macos-arm64 runner: Rust core -> PunktfunkCore.xcframework ->
swift build + swift test).
ci/rust-ci.Dockerfile: Ubuntu 26.04 with the workspace's link deps
(FFmpeg 8, PipeWire, Opus, GL/EGL/GBM, xkbcommon, libcuda via the
580-server userspace as a link stub) + pinned rustup + node for the JS
actions. Verified end to end in-container: build, 141/141 tests, C ABI
harness; all three images seeded to the registry manually.
scripts/ci/setup-macos-runner.sh provisions the Mac (rustup + darwin
targets, Node tarball, gitea-runner 1.0.8 host mode, LaunchAgent with
DEVELOPER_DIR auto-detect for sudo-free Xcode selection). Docs in
docs-site/content/docs/ci.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Comment reflow only — the pinned "stable" channel moved and CI checks
formatting with the current toolchain.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Stage-2 was a one-line "next" in the README. Add a full, actionable spec
(docs-site apple-stage2-presenter.md) a Mac agent can execute: VTDecompressionSession
decode (with decode-completion stamping) -> CAMetalLayer + display-link present, the
exact integration points against the existing StreamPump/StreamView/AnnexB/LatencyMeter,
the three-stage measurement wiring (capture->decoded / decode->present / capture->present
= glass-to-glass, using the already-wired PunktfunkConnection.clockOffsetNs), a cheaper
decode-only intermediate, validation, and gotchas. Link it from the Apple README's
Stage 2 item. (meta.json nav entry left in the working tree to land with the CI docs WIP.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
PUNKTFUNK_MUTTER_VIRTUAL_PRIMARY=1: after RecordVirtual, promote the per-session
virtual output to the primary monitor (physical kept on, secondary) via
org.gnome.Mutter.DisplayConfig.ApplyMonitorsConfig, restoring on teardown.
Without it, a GNOME host that also has a physical monitor attached keeps the physical
primary, so the virtual output is an empty extended desktop — the client streams only
the wallpaper. (The backend was validated on headless GNOME, where the virtual output
is the only display.)
Best-effort + opt-in: default behavior is unchanged; any DisplayConfig failure just
logs and streaming continues. method=temporary, so nothing is written to monitors.xml
and Mutter auto-reverts the layout when the virtual output is torn down.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The capture->client latency line concatenated a String onto a LocalizedStringKey
(Text("...\(x, specifier:)..." + (cond ? "" : "...")), which doesn't type-check:
the specifier: interpolation makes the literal a LocalizedStringKey, which has no
'+'. Fold the conditional suffix into the interpolation instead — the Apple
client didn't build on the latency-HUD commit (e04328f).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Input handling, building on macOS/iOS/tvOS:
- macOS recapture after navigating out: engageCapture no longer latches
captured=true when the cursor grab is refused mid app-activation (which left
a free cursor that no later click could re-grab); cursorCapture.capture() now
reports success. + canBecomeKeyView.
- iOS/iPadOS recapture: restore the prior capture on didBecomeActive (nothing
re-grabbed mouse/keyboard on return before).
- iPad indirect pointer (no lock) is forwarded as an absolute MOUSE (move +
buttons + scroll via hover / UITouch.indirectPointer), not as touch, with the
local cursor visible; GCMouse owns the locked regime, gated so the two never
double-send. Adds the MouseMoveAbs wire helper.
- Trackpad scroll on iOS (was entirely missing): GCMouse scroll dpad when
locked + a scroll-only UIPanGestureRecognizer otherwise.
- tvOS: no focusable control during play (a focusable Disconnect button ate the
controller's A in the focus engine); Siri Remote Menu disconnects.
- Don't leak touch to the host under the TOFU trust prompt (gate on
captureEnabled).
LAN discovery: HostDiscovery (NWBrowser over _punktfunk._udp, the host's
crate::discovery advert) resolves each service to IP:port and parses the TXT
(fp advisory, pair, id); an "On this network" section in the grid (tap to save
+ connect, or pair if required). iOS/tvOS get NSBonjourServices via a merged
Config/Info.plist. Integration-tested end to end against a fake NWListener advert.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Apple speed test asked for only 400 Mbps, capping the measured throughput
there and hiding the link's real headroom. Request the host's full
MAX_PROBE_KBPS (3 Gbps) instead, and raise the recommended-bitrate clamp from
500 Mbps to the host's 2 Gbps session ceiling so a fast measurement yields a
usable recommendation.
Also fix the stale caps left when the host clamps were raised (b8a33e2): the
resolved-bitrate range and the probe doc comments (abi.rs, client.rs,
regenerated header), plus the section 9 roadmap copy, now read 3 Gbps probe /
2 Gbps session.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The clock_offset test's assert_eq! carried an inline message that newer rustfmt
wants to wrap while the repo's committed style keeps such asserts on one line.
Move the message to a comment and use bare assert_eq! so it formats identically
under any rustfmt version — no new fmt-check ambiguity from this addition.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Apple client now consumes the connector's clock offset. PunktfunkConnection
reads punktfunk_connection_clock_offset_ns into clockOffsetNs at connect; a new
LatencyMeter (PunktfunkKit, NSLock + percentiles, mirrors FrameMeter) records each
AU's capture->client-receipt latency = now(CLOCK_REALTIME) + offset - pts_ns, and
SessionModel drains p50/p95 into the macOS HUD ("capture->client N/N ms p50/p95",
"(same-host)" when the host didn't answer the skew handshake). Wired at the
existing onFrame hook in ContentView — additive, no change to the decode/present
path. Unit test for the meter (percentiles, skew flag, absurd-value guard).
This is the first cross-machine latency the real Apple client reports. SCOPE:
stage-1 AVSampleBufferDisplayLayer decodes+presents compressed samples internally
with no per-frame callback, so this excludes decode+present; true decode->present
needs the stage-2 presenter (VTDecompressionSession + CAMetalLayer). Rebuild
PunktfunkCore.xcframework (for the new C getter) before swift build/test on a Mac.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Factor the client-side skew handshake into a shared core helper (quic::clock_sync
-> ClockSkew) so both the reference client and the embeddable connector use one
implementation. NativeClient now runs the handshake at connect (right after Start,
before the control task takes the stream) and stores the host-client offset; it's
read over the C ABI via punktfunk_connection_clock_offset_ns (i64 ns, host minus
client; 0 = no correction / old host).
This is the substrate the Apple client needs for the decode->present (glass-to-
glass) term: stamp present time, add the offset to express it in the host's
capture clock, subtract the AU pts_ns. client-rs drops its local clock_sync copy
and uses the shared helper (behavior unchanged; validated locally).
Regenerates include/punktfunk_core.h. Roadmap section 12 + status updated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per the new docs workflow (docs-site = KB layer; repo docs/ keeps design notes):
- Add a canonical Status & Progress tracker (status.md): milestones, per-box live
state, and a dated progress log — the go-forward place to track progress.
- Add setup guides: GNOME/Mutter host (gnome-box — Secure Boot MOK enroll, the
libnvidia-gl EGL fix, autologin, screen-lock disable, appliance unit), headless
KDE box, and Bazzite host (ujust input group, gamescope session, gotchas).
- Roadmap is now canonical in docs-site (synced the skew-handshake section 12
update); removed the repo docs/roadmap.md copy and repointed README to docs-site.
- Nav (meta.json) + landing cards updated; site builds (bun run build).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ClockProbe/ClockEcho on the QUIC control stream — 8 NTP-style rounds right after
Start; the min-RTT sample gives the host-client clock offset (clock_offset_ns
estimator in punktfunk-core). The client adds the offset to its receive instant
before differencing against the AU pts_ns, so the capture->reassembled latency
percentiles are valid across machines (skew_corrected=true), not just same-host.
Back-compat: an old host that doesn't answer the probe times out and the client
falls back to a shared-clock assumption (skew_corrected=false).
Host adds one ClockProbe dispatch arm in the control task; the client runs
clock_sync after Start, before the --remode/--speed-test tasks take the stream.
Validated cross-LAN (GNOME box -> dev box): offset ~ -1.57 ms (reproducible),
rtt ~140 us, p50 1.30 ms skew-corrected capture->reassembled — the offset is
exactly the systematic error the handshake removes. Unit tests for the message
codecs and the min-RTT offset estimator.
Roadmap §12: skew handshake done; remaining for true glass-to-glass is the Apple
client present-stamp (decode->present) plus the host render->capture term.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Both the unified host (serve --native) and standalone m3-host now advertise the
native punktfunk/1 service over mDNS (_punktfunk._udp) — the analogue of the
GameStream _nvstream._tcp advert. TXT records carry proto, the host cert
fingerprint (fp, the value clients pin), the pairing requirement
(pair=required|optional), and the host id. New crate::discovery module, wired
into m3::serve so both host entry points get it; best-effort, never blocks
streaming (--connect always works).
Client gains `punktfunk-client-rs --discover [SECS]`: browses the LAN and prints
each host (name, addr:port, pairing, fingerprint), then exits. Apple clients
browse the same service natively via NWBrowser (service type + TXT keys are the
contract).
Validated cross-LAN: the dev box discovered the GNOME-box appliance
(pair=required) and a standalone synthetic host (pair=optional); fingerprint and
pairing state correct in both.
Also refresh the now-stale sendmmsg caveat in the bitrate doc (batched/paced send
landed + validated to 1 Gbps) and mark the encode|send thread split done in §12.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Prep for a third (Ubuntu) test host: document the Mutter backend env — wayland-0
(not wayland-kde), XDG_CURRENT_DESKTOP=GNOME, PUNKTFUNK_COMPOSITOR=mutter, virtual
source via RecordVirtual, libei input via the RemoteDesktop portal.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bigger-bet #1 from the latency plan. virtual_stream ran capture+encode+seal+
paced-send on ONE thread, so frame N+1's capture/encode couldn't start until
frame N's entire paced tail had left the wire — the pacing budget (~0.9×interval)
was serialized in front of the next encode. Port GameStream's spawn_sender model
to the native path:
- A dedicated send thread (`send_loop`) owns the WHOLE Session (so no socket
clone or shared/Arc stats needed — `seal_frame` mutates the nonce, `send_sealed`
+ the probe bursts all live there) and does FEC+seal + microburst-paced send.
- The encode thread captures+encodes + handles reconfig and hands each AU over a
bounded sync_channel(3) as a FrameMsg (data, capture_ns, flags, deadline,
encode_us). It BLOCKS on backpressure if the send falls behind — frames slow
down rather than a dropped frame freezing the infinite-GOP stream (we don't
drop). Clean shutdown: drop the channel → send thread drains/exits → join.
- Probes (run_probe_burst) move to the send thread since they need the Session; a
burst naturally pauses video (the encode thread blocks on the full channel).
- Per-frame encode_us/pace_us histogram moved to the send thread (carries
encode_us in the FrameMsg) and now reflects the overlap.
Removes the encode↔paced-tail serialization (~2-8 ms @60-120 fps), independent of
the pacing policy, no quality cost. Substrate for the future NVENC slice wrapper.
Verified live on this box (appliance restarted onto it): a client streamed the
KWin desktop (1.49 MB H.265, clean, no panic) and a 200 Mbps speed-test probe
completed through the send thread (0 drops). Build + clippy + fmt green.
Real-NIC sustained soak (reconfig under load, line-rate, mode switches) pending
the Ubuntu third host.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
From a bug-hunt + unsafe-audit pass (4 reviewers + adversarial verify). It
confirmed ZERO real bugs in the recent batched/paced data-plane work — these are
the surfaced cleanups + one genuine soundness fix:
- SOUNDNESS (reduce unsafe): inject/gamepad.rs::pump_ff did `ptr::read` of an
InputEventRaw (align 8, holds a timeval) out of a 1-aligned [u8; N] buffer — UB
per the reference (x86_64 tolerates it, but it can miscompile under LTO). Use
ptr::read_unaligned + a SAFETY note. Zero behavior change.
- recv parity: recv_batch (recvmmsg) didn't drop an oversized/truncated datagram
the way scalar recv does — poll_frame now skips a message whose len fills the
buffer (> MAX_DATAGRAM_BYTES), matching recv's `n >= RECV_BUF` drop. (AEAD
already rejected these on encrypted sessions; this restores the documented
invariant on the batched path.)
- dedup unsafe FFI: factor the identical mmsghdr-from-iovec construction out of
send_batch + recv_batch into one `mmsghdrs()` helper — the raw-pointer
scaffolding + its lifetime SAFETY note now live in one place.
- docs: TARGET_SOCKBUF no longer calls paced sending future work (it landed,
m3.rs::paced_submit); gamescope.rs input is no longer "(TODO)" (wired +
live-validated); the PUNKTFUNK_PERF `wire_mbps` field is renamed `tx_mbps` and
noted as attempted/sealed bytes (send_dropped shows what didn't reach the wire).
Full suite (35 + loopback round-trip + 6) + clippy + fmt green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
From the latency investigation: the freeze-fix pacing (paced_submit) was the
single biggest software-controllable latency term — it unconditionally spread
EVERY multi-chunk frame over ~90% of the frame interval, adding up to ~7.5 ms
@120 / ~15 ms @60 to a frame's last packet even when the frame was small or the
link idle. Recover that on the common case while keeping the freeze fix:
- Microburst-cap pacing: a frame whose sealed size is <= a cap (default 128 KB,
PUNKTFUNK_PACE_BURST_KB) goes out in ONE immediate burst — no pacing latency.
Only the OVERFLOW of a bigger frame (IDR / sustained high bitrate, the bursts
that actually overran the tx buffer and froze) is spread. 128 KB is well under
the ~150 Mbps@60 frame size where drops began, so the default is safe; raise it
after confirming send_dropped stays 0 on a given link. Still never slower than
unpaced (budget collapses to 0 with no slack). seal-once/in-order nonce
preserved — chunks are split, never reordered or re-sealed.
- Per-frame instrumentation (PUNKTFUNK_PERF, zero-cost off): encode_us +
pace_us (the pacing tail) p50/p99/max histograms + immediate-vs-paced frame
counts in the periodic perf line, so the pacing tail is finally visible and the
cap is tunable against real numbers.
Host builds + clippy + fmt green. NOT yet deployed to the running hosts (still on
the safe full-pacing A+B build) — needs the user's LAN soak to validate the cap
doesn't reintroduce send_dropped before raising it. Deferred bigger bets (need
real-NIC/GPU/Mac validation): encode|send thread split on the native path,
CUDA stream+event (one redundant sync), NVENC slice wrapper, stage-2 Apple
presenter, glass-to-glass probe — see docs/roadmap.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Final increment of the 1 Gbps data-plane rework — the recv counterpart of the
sendmmsg work. The client recv path did one recvfrom + one Vec allocation per
packet (and the pump's 300µs idle sleep could let packets pile up at line rate).
- Transport gains recv_batch(&mut [Vec<u8>], &mut [usize]) -> count; default is
a single scalar recv into out[0] (loopback + non-Linux).
- UdpTransport overrides it on Linux with recvmmsg (MSG_DONTWAIT) draining up to
N datagrams per syscall into the caller's reused buffers — no per-packet alloc.
- Session::poll_frame owns a lazily-allocated recv ring (RECV_BATCH=32) and
consumes it one packet at a time across calls, refilling with one recvmmsg when
drained. Encapsulated: the punktfunk-client-rs + NativeClient pumps are
unchanged, and draining a batch per syscall means the 300µs sleep no longer
underdrains. Added UdpTransport::local_addr (used by the test, generally handy).
~125k → ~4k recv syscalls/sec at line rate, zero per-packet recv allocation.
Verified: new recv_batch_drains_over_loopback test (50 datagrams drained intact
via recvmmsg) + the existing loopback round-trip now runs through the batched
poll_frame; full suite (35 + round-trip + 6) + clippy + fmt green.
Decode-in-place (kill the per-packet open_from_wire alloc) is a separate later
optimization. With A (sendmmsg) + B (paced send) + C (recvmmsg), the native data
plane is batched + paced end to end.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Increment B of the send-path rework — the actual fix for "freezes get more
common over ~150 Mbps, no image at all at 400 Mbps" on the native path. Cause:
the encoder emits a frame and submit_frame blasted ALL its packets at once into
the NIC; a real link drops the line-rate burst (host send buffer EAGAINs), and
under infinite GOP one dropped frame freezes the decode until the next keyframe.
(The speed-test probe showed 0 drops at 400 Mbps because the probe is self-paced;
real video wasn't.)
Adaptive pacing, no extra thread, no regression:
- Session splits into seal_frame (FEC + packetize + seal → wire packets, no
send) and send_sealed (one batched sendmmsg of a chunk, counts drops);
submit_frame is now their composition (synthetic + probe paths unchanged).
- virtual_stream's paced_submit seals a frame then sends it in 16-packet chunks
spread over ~90% of the time until the next frame is due. At 60 fps desktop
(fast encode → lots of slack) the frame spreads across the interval → no NIC
burst → no freeze. At 240 fps@5K (encode ≈ interval → ~0 slack) the budget
collapses and every chunk goes out immediately → never slower than before.
Core suite (34 + loopback round-trip + 6) + clippy + fmt green. The seal/send
split is covered by the existing loopback tests; the pacing is host timing,
verified by review (live-test needs a real NIC — your Mac at a raised bitrate).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Cargo.lock update for the Linux-only `libc` dependency added in c24b571
(batched sendmmsg send). Keeps the lockfile in sync with Cargo.toml.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First increment of the 1 Gbps send-path rework (the measured bottleneck): the
native data plane did one send() syscall per packet — at ~125k pkt/s (1 Gbps
wire) that burns a core on syscalls. Port the proven GameStream sendmmsg path
into the core Transport seam.
- Transport gains `send_batch(&[&[u8]]) -> usize` (count handed to the kernel;
caller counts the rest as send-buffer drops). Default = the scalar send loop
(loopback transport + non-Linux).
- UdpTransport overrides it on Linux with `sendmmsg` (64 datagrams/syscall);
the connected socket needs no per-message address. Non-blocking-aware: a full
send buffer yields a short count / EAGAIN, and we stop + report what went out
rather than block or retry (same lossy, FEC-protected contract as send()).
- Session::submit_frame seals every shard then hands the whole frame to
send_batch in ONE call instead of looping send() — ~64x fewer syscalls per
frame on the native + GameStream-over-core paths; send_dropped accounting
preserved (total - sent).
~125k → ~2k syscalls/sec at 1 Gbps line rate. Verified: new loopback-UDP test
send_batch_delivers_over_loopback (100 batched packets arrive intact, datagram
boundaries preserved); full core suite + clippy + fmt green.
Next increments: a paced send thread (microburst shaping so a real NIC doesn't
drop line-rate bursts) and recvmmsg on the client.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Xwayland-DISPLAY poll did `d=$(pgrep -a Xwayland | grep … | head -1)`, but
under `set -euo pipefail` pgrep/grep exit non-zero when Xwayland isn't running,
so the command substitution failed and `set -e` aborted the WHOLE script —
killing KWin with it — on the loop's first iteration instead of polling.
It only ever worked when launched from an interactive shell where Xwayland
happened to already be up (so pgrep matched on try 1). Under the systemd boot
appliance (punktfunk-kde-session.service) Xwayland isn't up that early, so the
session crash-looped (restart counter climbing, KWin never staying), the host
had no compositor, and clients couldn't connect.
Append `|| true` to the substitution so the loop polls as intended and a session
with no Xwayland at all still proceeds (DISPLAY just stays unset → warn).
Verified live: the unit now stays active (0 restarts), KWin + the wayland-kde
socket persist, probe-compositor reports ready, and a real client session
captured 4.8 MB of H.265 off the running serve --native host.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Make a headless box a self-contained streaming appliance: after boot, with no
display manager / login / manual script, the headless KWin Plasma session and
the punktfunk host both come up so a client can just connect and stream the
desktop.
- New scripts/punktfunk-kde-session.service: a Type=simple user unit that runs
run-headless-kde.sh (kwin --virtual on wayland-kde + Plasma + portals + a
supervised plasmashell). The script foregrounds on `wait $KWIN_PID`, so
Restart=always keeps the desktop alive across a KWin crash.
- scripts/punktfunk-host.service: ExecStart now `serve --native` (the unified
GameStream + punktfunk/1 host, matching how it's actually run), After= the
kde-session unit (soft ordering — the host listens immediately and only needs
the compositor per session, so a missing unit on the gamescope backend is
harmless), and appliance install docs (kwin vs gamescope backend).
Boot still requires `sudo loginctl enable-linger $USER` (the one thing that
starts user units without a login) — documented in both unit headers.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First step of 1 Gbps+ readiness (the whole point of the GF(2^16) Leopard FEC):
make 1 Gbps configurable and its dominant failure mode observable, before the
real transport work (sendmmsg + paced encode|send split) lands.
Investigation (6-way) verdict: we're ~halfway, and it's mostly clamps plus one
real piece of work. The integer/type path, FEC (a 1 Gbps frame is only a few
hundred shards in one GF(2^16) block, far under the 65535 ceiling), AES-GCM
(AES-NI, ~10-25x headroom), and the M1 reassembler bounds (fully derived from
the negotiated FecConfig) are ALL already 1 Gbps-ready and untouched.
This commit (the configurable + observable foundation):
- m3.rs: MAX_BITRATE_KBPS 500_000 -> 2_000_000 (2 Gbps headroom over the 1 Gbps+
target); MAX_PROBE_KBPS 1_000_000 -> 3_000_000 (probe can demonstrate headroom
ABOVE the session cap so a client can confidently pick a 1 Gbps+ bitrate).
- transport/udp.rs: TARGET_SOCKBUF 8 MB -> 32 MB (a multi-MB IDR keyframe burst
no longer fills the buffer); scripts/99-punktfunk-net.conf bumped to match.
- Observability: Transport::send now returns Ok(true|false) (false = WouldBlock
send-buffer drop, previously a silent Ok(())). Session counts these as a new
`packets_send_dropped` stat (distinct from recv-side packets_dropped) — in
Stats, the C ABI PunktfunkStats (header regenerated), a PUNKTFUNK_PERF periodic
wire-Mbps + drop dump in virtual_stream, and the speed-test probe completion
log. This is the dominant 1 Gbps+ loss mode and was invisible.
Loopback-verified: a probe now runs at 1.2 Gbps target (no longer truncated to
1 Gbps) with the drop counter live. NOT yet a sustained-1-Gbps proof — the
single-send()-per-packet native path is the next, real piece of work (port the
proven GameStream sendmmsg + paced send thread into the core Transport).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On Bazzite (atomic rpm-ostree) `sudo usermod -aG input $USER` doesn't stick —
/etc/group is managed declaratively, so the change is dropped or reverted on
the next update. The supported path is the `ujust add-user-to-input-group`
recipe, which edits the group the immutable-OS-correct way. Update the bazzite
README + the packaging quickstart + the troubleshooting note (which also now
points at the host's "virtual gamepad/DualSense created" vs "creation failed"
log as the unambiguous signal).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The uinput X-Box 360 backend logs "virtual gamepad created" on success, but
the UHID DualSense backend logged only on failure — so a working DualSense
session was silent and indistinguishable in the logs from one where no pad
was ever created. Add the matching success log.
This makes a DualSense-not-working report self-diagnosing: the host now logs
either "virtual DualSense created (UHID hid-playstation)" or the existing
"virtual DualSense creation failed — controller input disabled" (which fires
when /dev/uhid isn't writable — i.e. the 60-punktfunk.rules uhid rule isn't
installed or the user isn't in the 'input' group).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two related additions to the native protocol, host-side (the client side of
each is exposed over the C ABI so the platform clients can wire it up).
Bitrate negotiation
- Hello/Welcome carry `bitrate_kbps` (appended trailing-byte field, back-compat:
old peers decode 0 = host default). The client requests a rate; the host
clamps it to [500 kbps, 500 Mbps] (or its 20 Mbps default when 0) and echoes
the resolved value in Welcome. Replaces the hardcoded 20 Mbps NVENC bitrate in
m3.rs — threaded through virtual_stream → build_pipeline → open_video, applied
on the initial mode and every reconfigure rebuild.
- C ABI: punktfunk_connect_ex3(..., bitrate_kbps, ...) (ex2 delegates with 0);
punktfunk_connection_bitrate() reads the resolved value.
Speed test (bandwidth probe)
- New typed control messages ProbeRequest{target_kbps,duration_ms} (0x20) /
ProbeResult{bytes_sent,packets_sent,duration_ms} (0x21), plus a FLAG_PROBE
packet flag. The client asks the host to burst zero-filled, FLAG_PROBE-tagged
access units over the data plane at a target goodput for a duration (clamped
≤ 1 Gbps / ≤ 5 s), pacing by a bytes-allowed budget; video pauses for the
burst. The host reports what it actually sent; the client measures received
bytes + window → goodput and loss. Probe filler is never fed to the decoder
(diverted in the connector pump and the reference client's poll loop).
- The host control task now multiplexes Reconfigure + ProbeRequest (inbound)
and ProbeResult (outbound) over select!; a probe channel reaches the
data-plane thread (both virtual and synthetic sources).
- Connector: NativeClient::request_probe()/probe_result() with an internal
accumulator; C ABI punktfunk_connection_speed_test() +
punktfunk_connection_probe_result() → PunktfunkProbeResult.
- punktfunk-client-rs gains `--bitrate KBPS` and `--speed-test KBPS:MS` (its own
loop measures + logs goodput/loss) for loopback verification.
Validated on loopback (synthetic source): a 20 Mbps / 2 s probe measured
20050 kbps at 0% loss, bitrate negotiated (0→20000 and 50000→50000), and the
interleaved probe AUs were correctly excluded from frame verification
(mismatched=0). Wire codecs + trailing-byte back-compat have unit tests. C
header regenerated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
macOS GCKeyboard delivery is flaky — the same GameController quirk that
killed GCMouse motion (e414ec0). Keyboard input intermittently failed to
reach the host (e.g. typing in a gamescope game). Switch the macOS key
source to NSEvent, mirroring the mouse fix:
- StreamLayerView.keyDown/keyUp map NSEvent.keyCode (Carbon virtual
keycode) → Windows VK via the new InputCapture.keyCodeToVK table and
forward through InputCapture.sendKey, then consume the event (no beep).
- flagsChanged drives InputCapture.handleFlagsChanged, which diffs the raw
modifier flags to recover each L/R modifier down/up (modifiers never fire
keyDown/keyUp on macOS) and emits the same L/R VKs hidToVK already does.
- The macOS GCKeyboard keyChangedHandler is disabled (#if !os(macOS)) so it
can't double-send; iOS keeps the GCKeyboard path unchanged.
sendKey honors the ⌘⎋ capture-toggle suppressedVK latch and tracks into
pressedVKs so releaseAll()/blur flushes anything still held. The emitted
VKs are identical to the existing HID path, so the host (vk_to_evdev)
needs no change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
§10: HDR/10-bit is blocked at the capture source, not our stack — gamescope's PipeWire
node is hardcoded 8-bit (BGRx/NV12; confirmed in src/pipewire.cpp on the box's c31743d
build), issue #2126 open+unstarted; PipeWire >=1.6 needs Fedora 44 (Bazzite F44 is
testing-only with a confirmed NVIDIA Game-Mode crash, so a rebase clears only the
PipeWire wall). The realistic route is KWin MR !8293 (HDR PipeWire capture, draft),
i.e. the desktop path. Records the settled constraints (NVENC 10-bit max, HDR⟹HEVC
Main10, AV1 can't HDR on VideoToolbox) + the ready-to-build downstream design.
Also notes the landed Bazzite dynamic-resolution work (host-managed gamescope-session
at the client's exact res+refresh, c894c6f) + the macOS/iPad input and 4K/5K UDP-buffer
freeze fixes in the Done summary.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Nested games on the Bazzite host saw the wrong display: refresh capped at 60 Hz,
the box's connected TV's EDID modes leaking in (DOOM landed on 2560×1440@60), and
the resolution fixed at whatever the always-on session was launched at — the
client's requested mode never reached the game. Root causes: the session-plus
gamescope command has no --nested-refresh (Xwayland advertises 59.96 Hz for every
mode), --prefer-output HDMI-A-1 makes gamescope read the TV EDID, and the ATTACH
model launches one fixed-resolution session.
New vdisplay path: PUNKTFUNK_GAMESCOPE_SESSION=<client> — the host LAUNCHES
gamescope-session-plus headless AT THE CLIENT'S mode and relaunches it when the
mode changes. Injected via a host-written GAMESCOPE_BIN wrapper (--nested-refresh
$PF_HZ, the flag session-plus doesn't expose) + DRM_MODE=cvt (gamescope generates
clean CVT modes at that refresh instead of the TV's EDID). The session runs as a
transient `systemd-run --user` unit (clean cgroup teardown of the Steam tree);
state lives in a host-lifetime static (MANAGED_SESSION), NOT in GamescopeDisplay
(which is per-client-session) — so a same-mode reconnect REUSES the running
session instantly (no Steam restart) while a different mode RELAUNCHES it (games
can't change output mode live; a game/Steam restart on a mode change is
unavoidable and acceptable). Reuses the existing node + EIS auto-discovery
(find_gamescope_node / find_gamescope_eis_socket, factored into
point_injector_at_eis) and the existing mid-stream Reconfigure → vd.create(mode)
machinery — no protocol or m3 control-flow change.
Validated live on bazzite (RTX 4090): games' Xwayland now advertises 5120×1440 @
239.90 Hz as the preferred mode (was 59.96), the TV's 3840×2160/4096×2160@60 modes
are gone, frames stream; reconnect at 1920×1080@120 relaunches and games see that;
same-mode reconnect reuses with no restart and frames flow instantly.
scripts: host.env.example documents PUNKTFUNK_GAMESCOPE_SESSION (mutually exclusive
with the legacy NODE=auto attach); punktfunk-steam-session.service marked
deprecated (superseded — must not run alongside the host-managed path).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Apple client grows full gamepad support and punktfunk/1 learns to negotiate
the virtual pad type:
- Protocol: Hello carries a GamepadPref byte (offset 21, the same trailing-byte
back-compat pattern as the compositor; echoed resolved in Welcome at 54).
Host precedence: explicit client choice > PUNKTFUNK_GAMEPAD env > Xbox 360,
DualSense (UHID) only where available. ABI: punktfunk_connect_ex2 +
punktfunk_connection_gamepad (connect_ex delegates; ABI_VERSION stays 2 — the
trailing byte IS the compat mechanism). punktfunk-client-rs gets --gamepad.
- Swift client: GamepadManager (app-lifetime discovery + selection — Settings
lists every controller with capabilities/battery/"In use"; exactly ONE pad
forwards as pad 0, auto = most recently connected, or pinned), GamepadCapture
(snapshot-diff button/axis events, DualSense touchpad + ~250 Hz motion on the
rich-input plane, held state released on switch/deactivate/stop),
GamepadFeedback (rumble → CoreHaptics per-handle engines; lightbar →
GCDeviceLight; player LEDs → playerIndex; adaptive-trigger blocks → the
table-driven DualSenseTriggerEffect parser → GCDualSenseAdaptiveTrigger,
exact for the 10-zone positional modes). The pad type auto-resolves from the
physical controller at connect time, user-overridable in Settings.
- Host DualSense fixes surfaced by adversarial review against hid-playstation /
SDL / Nielk1 ground truth: input-report sensor/touch offsets were off by one
(the kernel read garbage motion + phantom touches), the L2/R2 trigger blocks
were swapped (the report is right-trigger-first), feedback now gates on the
report's valid-flags (a plain rumble write no longer blanks lightbar/
triggers), and the touchpad rescale clamps to the advertised ABS_MT extents.
- Tests: Hello/Welcome trailing-byte back-compat, pick_gamepad precedence,
byte-exact input-report layout, valid-flag gating, per-mode trigger-parser
table (incl. packed 3-bit zones), wire conversions, and a scripted loopback
feedback burst (PUNKTFUNK_TEST_FEEDBACK=1) asserted through the xcframework
on the rumble + HID-output planes.
Validated: cargo test/clippy/fmt green on macOS + Linux (61 host tests), swift
build/test green, test-loopback.sh green, tvOS/iOS targets compile. DualSense
motion sign/scale is derived from the calibration blob, not yet live-verified
(constants isolated in GamepadWire).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The host warns when its UDP socket-buffer grant is small (Linux caps SO_SNDBUF at
net.core.wmem_max, ~208 KB by default). Validated zero-loss at 5K even at that cap,
but raising it gives send-side headroom for higher bitrates / concurrent sessions.
Referenced from the headless-Steam appliance setup. macOS clients need no tuning.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The data-plane UDP sockets used the OS default buffer (~208 KB on Linux, similar
on macOS), which is smaller than a single high-resolution frame burst: a
5120×1440 keyframe is ~130 packets the encode|send thread hands to sendmmsg at
once. The burst overflows the buffer — EAGAIN on the host send (now dropped, was
fatal) or a silent drop on the client recv — and because the data plane runs
infinite-GOP, one lost frame breaks every subsequent reference and the decode
freezes on the last good frame until an RFI refresh that may never catch up.
Symptom: connect at 5120×1440, see ONE frame, then a frozen image (audio + input
keep working — those ride QUIC, not this socket).
Set SO_SNDBUF/SO_RCVBUF to 8 MB (clamped by the OS to net.core.{w,r}mem_max on
Linux / kern.ipc.maxsockbuf on macOS); warn if the grant lands far below target so
an undersized host is diagnosable. The client side matters most — the SAME
UdpTransport backs the Apple client's data plane via the C ABI, and macOS grants
multi-MB buffers without any sysctl, so a rebuilt client stops losing frames.
Validated live, bazzite→client at 5120×1440: was 1319/1500 frames (12% loss →
freeze), now 1500/1500 @60 and 5279/5279 @240 (split-encode active), zero
mismatches, p50 1.9–3.4 ms. Host send buffer was still capped at 416 KB and lost
nothing — the loss was purely the client recv buffer.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bazzite (and SteamOS-like hosts) run Steam Big Picture inside their OWN
gamescope-session-plus session. Nesting a second gamescope+Steam can't work — the
second Steam sees the first and exits, taking the nested gamescope down with it
(crash in its exit handlers), killing both video and input. The robust model is to
let punktfunk OWN that session: run gamescope-session-plus headless at the client's
resolution (full Steam Deck UI polish: MangoApp, VRR, controller config) and have
the host ATTACH to it rather than spawn its own.
The video half already existed (PUNKTFUNK_GAMESCOPE_NODE=<id> attaches to a
PipeWire node). This finishes it:
- PUNKTFUNK_GAMESCOPE_NODE=auto discovers the gamescope Video/Source node, so the
(dynamic) node id needn't be hand-wired.
- The attach path now also points the libei injector at the running session's EIS
socket: find_gamescope_eis_socket() scans XDG_RUNTIME_DIR for gamescope-<N>-ei,
connect()-probes each (stale dead-session sockets refuse), and writes the newest
live one to the relay file the injector reads. So input reaches the attached
session with zero manual config.
scripts/punktfunk-steam-session.service: a systemd --user unit that runs
gamescope-session-plus headless at a configured resolution, with the one-time
headless-appliance setup (linger + multi-user.target) documented inline.
Validated live on bazzite (RTX 4090): the full Steam Big Picture session streams
(1499 frames, p50 ~1ms) with mouse/keyboard injected into it (device resumed, all
caps, emitted=true), node + EIS socket both auto-detected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
UdpTransport sockets are non-blocking, so a momentarily-full kernel send buffer
makes socket.send return WouldBlock (EAGAIN). submit_frame propagated that as a
fatal error, tearing the whole punktfunk/1 session down — observed when attaching
to an already-running source (a headless Steam session) that emits frames at full
rate the instant capture connects: the first burst saturates the tx queue and the
session dies before a single frame reaches the client.
The data plane is lossy + Leopard-FEC-protected and runs infinite-GOP with RFI
keyframes, so the real-time-correct response to a full tx queue is to DROP the
packet (the next frame / FEC recovers) — exactly what the recv path already does
for WouldBlock. Blocking would queue stale frames and add latency. Loopback/M1
paths are unaffected (LoopbackTransport never blocks; M1 tests stay green).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The PIN now surfaces in the host's web admin UI (port 3000 → Pairing), which is where
users will actually read it — the pairing sheet's footer, field prompts, the tvOS
keyboard title, and the wrong-PIN/failure errors all reference the console instead of
the host log / --allow-pairing flag (the log mention stays in the README as the
secondary path).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Host-side logs proved the macOS client sent keyboard + scroll but ZERO relative
mouse-motion and ZERO button events for an entire session — the user was moving
the mouse the whole time. Root cause is client-side: GCMouse's
mouseMovedHandler/pressedChangedHandler silently never fired on the live Mac
(a documented GameController quirk) while GCKeyboard worked and scroll already
rode NSEvent. So motion/buttons were the only input on a GCMouse-only path, and
that path was dead.
macOS: stop relying on GCMouse for motion/buttons (compiled out with
#if !os(macOS)); drive them from a local NSEvent monitor installed only while
captured — the same channel scrollWheel already uses successfully. Under
CGAssociateMouseAndMouseCursorPosition(false) the mouseMoved/dragged deltaX/deltaY
ARE the relative motion (OS-acceleration-applied, exactly what Moonlight's macOS
client ships). All four motion event types are covered so motion keeps flowing
during a button-held drag; buttons map left/right/middle/X1/X2 through the
existing engage-click-suppression + release-on-blur logic. NSEvent deltaY is
already screen-space (+y down) so, unlike the GCMouse path, it is NOT negated.
iPad: the input failure there was a different cause — GCMouse only delivers
relative deltas while the scene holds a true pointer LOCK, which the system grants
only to a full-screen, frontmost iPad scene and which UIHostingController doesn't
consult for children. Gate prefersPointerLocked to iPad + captured, add
childViewControllerForPointerLock so a reparenting container forwards the lock
decision to this VC, and log the resolved lock state. Touch remains the
unconditional fallback.
Adds a PUNKTFUNK_INPUT_DEBUG=1 switch (os.Logger, throttled) so motion/buttons
being SENT is verifiable on-device without host-side logs. iOS GCMouse path
otherwise unchanged; GCKeyboard unchanged on both. Researched + adversarially
reviewed; Swift builds only on a Mac, so this is unverified-compiled here.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The slide now runs on UISpringTimingParameters (stiffness 300, damping 30 — a ~0.87
damping ratio: settles quickly with a hint of life, no overshoot ping-pong) via the
transition library's .interpolatingSpring animation.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The navigationLink Picker's INTERNAL destination list renders its rows in the focused
(dark-text) style while the push animates — black text over the dark backdrop until
focus settles (present under the old fade too; a SwiftUI-on-tvOS quirk we don't
control). Settings now uses its own primitives instead:
- TVSelectionRow: label + current value, pushes…
- TVSelectionList: a Settings-app-style option list (plain button rows + checkmark,
selecting pops back) — ordinary button chrome, no focused-style pre-rendering.
The stream-mode and compositor pickers are gone on tvOS; the Settings screen itself is
a plain scroll of rows + footer (no Form), matching the rest of the tv UI.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
SwiftUI's NavigationStack on tvOS animates pushes as a bare crossfade with no public
customization — the system Settings app slides. The home stack now applies
.customNavigationTransition(.slide) on tvOS via davdroman/swiftui-navigation-transitions
(MIT, tvOS 13+), covering the top-level routes AND the settings pickers' drill-ins.
The dependency is referenced by the Xcode PROJECT only and linked solely by the
Punktfunk-tvOS target: its manifest (no macOS platform declared vs 10.15 deps) breaks
SwiftPM whole-graph validation for plain `swift build`, and the #if os(tvOS) import
never compiles in the macOS-only SwiftPM dev shell anyway. Headless builds need
xcodebuild -skipMacroValidation (the lib pulls Swift macro packages; in the Xcode UI
it's a one-time Trust & Enable prompt).
iOS/macOS keep their untouched system navigation animations.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Add Host, Settings and PIN pairing were fullScreenCover overlays, which is why
navigating felt unlike the system Settings app (no push animation, no Menu-pops-a-level
semantics). They are now navigationDestination ROUTES pushed inside the home
NavigationStack:
- the system push/pop animation and Menu-button back navigation come for free;
- the Settings pickers' navigationLink pushes reuse the same stack (its inner
NavigationStack wrapper is gone, as is the tvOS Done row — Menu pops, like Settings);
- Add Host is a real full-screen page (system navigation title, Settings-style rows on
the standard backdrop) instead of a floating dialog, same for the pairing page;
- the thickMaterial cover backdrops became unnecessary and are gone. The system
keyboard entries stay as covers — that presentation is system-owned either way.
iOS/macOS keep their sheets. Verified by screenshot: Add Host renders as a pushed
full-screen route with the title top-center.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The host-lifetime libei injector could connect to a gamescope EIS socket whose
listen socket exists but whose server never drives the EI handshake — a stale
socket left by a SIGKILLed prior session, or one created early in a new
gamescope's startup before its libei server is ready. `UnixStream::connect` to a
socket *file* succeeds the moment the path exists, so the worker sailed past the
connect and then hung forever in `handshake_tokio` (or sat connected with no
device ever resumed). Because `LibeiInjector::inject` only enqueues onto a
channel (the !Send worker owns the connection), the send never errors, so
InjectorService never noticed the dead worker and never reopened — every input
event for the whole session was silently swallowed. The 30s setup timeout didn't
help: a typical session ends first, so input just died with no error logged.
Reconnecting made it worse (more stale sockets to land on).
Two self-heal bounds, both paths (gamescope socket + KWin/GNOME portal):
- Bound the EI handshake at 8s — a non-responding EIS server now errors instead
of hanging, so the worker exits and the next inject() reopens.
- Watchdog: if no input device resumes within 5s of connecting, treat the
connection as dead-on-arrival and exit (same reopen path). Healthy servers
add+resume a device within a beat of the handshake.
Verified on-box: clean gamescope + KWin paths connect/resume/emit unchanged; a
stale listener that accepts-but-never-handshakes now errors in 8s; two
back-to-back gamescope sessions both inject (session 2 reopens against the fresh
socket). Independently confirmed end-to-end delivery on KWin — a focused wev got
the injected motions/keys/buttons — i.e. injection itself was never broken, only
its recovery from a bad connection.
Also adds permanent low-volume diagnostics so the next "input dead" report is
instantly triageable: log each EIS device's capabilities on resume, the first of
each InputKind a client sends + whether it emitted, and no-resumed-device drops.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
SwiftUI's inline TextField on tvOS is structurally wrong for television: it grows when
activated, shows a full-width editing surface behind the pill, and floats labels
off-center — none of it stylable into the Settings-app look. Per Apple's tvOS text
input guidance, real tvOS apps never edit inline: a field is a value ROW, and pressing
it raises the SYSTEM fullscreen keyboard.
- TVTextEntry (UIViewControllerRepresentable): a UITextField that becomesFirstResponder
on appear, presenting the standard tvOS fullscreen keyboard with the field's prompt;
done/dismiss commits the text. TVFieldRow is the Settings-style label+value lozenge.
- Add Host and PIN pairing on tvOS now use rows + keyboard covers exclusively (the
port row also fixes the off-center value text for good — it's a Text, not a field);
the port input validates 1...65535.
- No SwiftUI TextField remains in any tvOS code path.
Verified by screenshot: the dialog rows render exactly like the Settings app, and the
address row raises the system linear keyboard with prompt + done.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three more tvOS-isms, all the same lesson — let the focus engine own the chrome:
- Host cards drew their own material platter + accent ring INSIDE the .card button
style, muting the native grow/tilt focus motion. On tvOS the card style now owns the
platter outright (material/ring stay on the pointer platforms), and the grid gets
48 pt spacing so the focused card swells without overlapping siblings.
- Add Host and Settings no longer sit in the hosts row: they're a compact button row
below the grid (and the empty state gains a Settings button, since tvOS has no
toolbar).
- The Add Host and pairing dialogs drop Form entirely on tvOS — list rows added a
full-width focus fill plus a row platter behind every field's own pill (the
"second outer pill"). As standalone fields in a centered dialog over the dimmed
home, each input is exactly one pill with vertically centered text.
Verified by screenshot in the Apple TV simulator (home grid + Add Host dialog).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The inline iOS form widgets fought the tvOS focus system at every turn: focused fields
showed nested pills, rows darkened oddly and grew on activation, the Compositor picker
was not even focusable, and prefilled fields (port, client name) floated their label
inside the pill, shoving the value off-center.
- Settings is now a fully tv-native screen: NO inline text entry — the stream mode is
a preset picker (This TV native / 720p / 1080p / 4K, plus a Custom entry preserving
a mode set on another platform) and both pickers use .navigationLink style (pushed
selection lists, exactly like the system Settings app — and properly focusable; the
cover wraps in a NavigationStack for the pushes).
- Where text entry is unavoidable (Add Host, PIN pairing), the fields keep their stock
single-pill chrome (the grouped form style stays off tvOS — its row platters were
one of the nested pills) and prefilled fields hide their floating label so values
center vertically.
- All earlier row-clearing experiments reverted.
Verified by screenshot in the Apple TV simulator: Settings rows render as single
focus lozenges with chevrons; the Add Host pills are uniform with centered text.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
tvOS forms/lists have CLEAR backgrounds and a fullScreenCover only shows what the
presented view paints, so Settings/Add Host/pairing rendered transparently over the
hosts grid. All three covers now sit on .thickMaterial edge to edge — the standard
tvOS blur-over-content panel look (verified in the Apple TV simulator).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The iOS chrome half-worked on tvOS: toolbar items rendered tiny with clipped labels
and could not even be focused (which is why "+" never opened the add-host form), and
sheet presentations are not a tvOS idiom (the Settings form looked broken).
- The toolbar is gone on tvOS. Add Host and Settings live IN the hosts grid as
full-size, focus-native tiles (.card style, same geometry as the host cards) — the
natural way actions work on television.
- Every modal (Add Host, Settings, PIN pairing) presents as a fullScreenCover on tvOS;
Settings gains a tvOS-only Done button (covers don't dismiss themselves).
- iOS/macOS keep their existing toolbar + sheets untouched.
Verified in the Apple TV simulator: title, host card and both action tiles render
full-size and focusable.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Icon Composer doesn't cover tvOS — tvOS app icons are the older parallax format:
flat layers in an asset-catalog "App Icon & Top Shelf Image" brand asset. Generated
from the same Affinity layer exports the Icon Composer .icon uses, mirroring its
composition (violet automatic-gradient background → light circle → dark circle →
blob in front), via scripts/render-tvos-icon.swift (checked in for regeneration):
- App Icon.imagestack 400×240 @1x/@2x + App Icon - App Store.imagestack 1280×768,
four layers each so the focus engine gets real parallax depth.
- Top Shelf Image (1920×720) + Wide (2320×720) @1x/@2x as flat composites.
- ASSETCATALOG_COMPILER_APPICON_NAME = "App Icon & Top Shelf Image" on the tvOS
configs; verified on the Apple TV simulator home screen.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A bandwidth probe over the punktfunk/1 data path so a user-settable bitrate can be
informed by what the network actually sustains (throughput/loss/jitter), surfaced in
the client UI + web console. Reuses the existing Session/FEC plumbing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of "input doesn't work" on the unified host: a single fresh session
injects fine (EIS connects, "Gamescope Virtual Input" device added), but the
host-lifetime injector reused a STALE per-session EIS socket across sessions →
"connect EIS socket …: Connection refused". (Headless gamescope is EIS-only — it
ignores uinput — so libei/EIS is the one input path for both gamescope and KWin;
no second path needed.)
- connect_socket_file: re-READ the relay file and RETRY the connect on
refused/missing (the live gamescope's EIS appears shortly), bounded at 15s,
instead of connecting once and bubbling ECONNREFUSED.
- GamescopeProc::drop: clear the relayed EIS socket name on teardown so a dead
session can't hand a stale path to the next reconnect.
Validated: two sessions back-to-back each reconnect (EIS connected + device added).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The same app now runs on tvOS (target Punktfunk-tvOS, bundle io.unom.punktfunk.tvos),
validated live against the box: vkcube at 1280x720@60, 60 fps in the Apple TV 4K
simulator, glass HUD with a focusable Disconnect button.
- PunktfunkCore.xcframework grows tvOS device + universal-simulator slices. These are
TIER-3 Rust targets (no prebuilt std): BUILD_TVOS=1 builds them with nightly and
-Zbuild-std from rust-src — the full quic stack (quinn/rustls-ring/tokio) compiles
for tvOS unchanged.
- The UIKit stream view covers iOS AND tvOS, with pointer interaction, pointer lock,
touch forwarding and InputCapture gated to iOS — tvOS is view-only until gamepad
capture lands (the natural tvOS input).
- SessionAudio on tvOS: .playback session, no mic (no app-accessible microphone).
- App chrome gates: keyboardShortcut/textSelection/controlSize/statusBarHidden are
iOS/macOS-only; host cards use the focus-native .card button style on tvOS; the
Audio settings section hides (system-routed); mode seeding works from the TV screen
(1920x1080@60).
- Package platforms += .tvOS(.v17); new Xcode target + shared scheme
(TARGETED_DEVICE_FAMILY 3, local-network usage description included).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Working through the brand-color follow-ups:
- AccentColor gains a dark-appearance variant (#8678F5 — the brand violet lifted one
step toward the icon's light periwinkle) so tinted controls keep contrast on dark.
- Host cards remember sessions: StoredHost.lastConnected (set when a session reaches
streaming) renders as a "Connected … ago" relative-time line, and the most recent
host's card carries a subtle accent ring — the grid finally has hierarchy.
- The HUD swaps the pre-glass black-50% rectangle for .regularMaterial with an accent
live-dot; hint lines use semantic .secondary instead of opacity.
- Security moments: the trust card's lock.shield and the pairing sheet's header take
the brand tint; the PIN field is larger monospaced and uses the number pad on iOS.
Icon ↔ accent decision: the accent stays the exact brand #6656F2; the Icon Composer
layers keep their adjacent palette (#6C5BF3 family) — close enough to read as one
brand, and the icon remains the design-tool source of truth.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
AccentColor color set + ASSETCATALOG_COMPILER_GLOBAL_ACCENT_COLOR_NAME on all four app
configurations — the platform-sanctioned global tint, so the host-card icons, prominent
buttons, toggles, pickers and links all carry the brand violet on macOS and iOS without
any per-view styling.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The 160 pt grid minimum packed five small cards per iPad row. iOS columns now use a
280 pt minimum (one full-width card on iPhone portrait, 3–4 generous cards on iPad)
and the card content scales with it: 56 pt icon, title3 name, taller padding. macOS
keeps its compact 180–240 pt cards.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Matches the bundle display name; was the lowercase project name "punktfunk" in the
home navigation title (iOS large title / macOS titlebar) and the WindowGroup title.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Streaming on iPad left the status bar up and the video boxed inside the safe areas, on
top of a 16:9 default mode letterboxing on the 4:3 screen, with the iPadOS cursor
hovering over the video. The session view is now immersive on iOS:
- .ignoresSafeArea + .statusBarHidden + .persistentSystemOverlays(.hidden) for the
session only (home gets its chrome back on disconnect).
- First run seeds the stream mode from the device's native screen
(UIScreen.nativeBounds + maximumFramesPerSecond) instead of 1920×1080 — verified
live: a fresh install negotiated the iPad's 2752×2064 with the host. macOS keeps the
1080p default (a desktop window is not the screen).
- The iPadOS cursor hides while over the video (UIPointerInteraction .hidden(),
re-resolved on capture toggles) — the host renders its own cursor from our deltas;
true pointer lock through UIHostingController remains the documented gap.
Found along the way (host-side, not fixed here): at very high modes a keyframe burst
can fill the UDP send buffer and m3 treats the sendmmsg WouldBlock as fatal
("session ended with error: submit_frame: WouldBlock") instead of backpressuring.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
§8a (require native pairing by default, serve --open) shipped + deployed. §8b
(delegated approval) refined into §8b-1 (host pending-requests + mgmt endpoints +
web Approve/Deny — achievable now) and §8b-2 (peer push to a paired Device A —
needs the native/Apple client UI).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
An open punktfunk/1 host any LAN device can trust-on-first-use and stream from is
insecure. The unified host now gates native sessions on pairing by DEFAULT: a client
must complete the SPAKE2 PIN ceremony (armed from the web console) before it's
admitted; paired devices persist. `serve --open` keeps the old TOFU behavior for
trusted single-user setups.
native_serve_opts now takes a NativeServe { port, require_pairing }; parse_serve
builds it with require_pairing = !--open. GameStream pairing (separate) is unchanged.
The require_pairing gate + ceremony are already covered by m3::pairing_ceremony_and_gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Icon Composer re-import after stripping the Affinity artboard rects (full-canvas
fill:none rects the exporter adds per layer) that caused rendering artifacts.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Completes the web-UI native (punktfunk/1) pairing flow the unified host backs.
The Pairing page now leads with a native card that arms a window via the mgmt API
and DISPLAYS the host PIN (the SPAKE2 ceremony is host-mints / client-enters) with
a live countdown + Cancel, plus a paired-devices list with unpair — no journalctl.
The existing Moonlight PIN-submit moves into its own section below.
Uses the orval-generated `native` hooks (regenerated from the committed OpenAPI on
build) + en/de strings. Validated end-to-end through the web server's proxy + cookie
auth: login → status → arm (PIN shown) → clients. tsc + production build clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The "title looks off" report traced to the GRID, not the title: the Mac-tuned
adaptive(180–240) columns yielded a single max-width card, centered, so nothing aligned
with the leading large title. The header is now entirely stock primitives — default
.navigationTitle large-title behavior (the inlineLarge experiment is gone), default
.padding() so content sits on the system 16 pt margins — and the grid columns are
platform-tuned: iOS drops the max so columns FILL the width and the cards stay
edge-aligned with the title; macOS keeps the 180–240 cap (huge windows shouldn't grow
huge cards).
Verified in the iPhone 17 simulator with seeded hosts: pill top-right, large title at
system metrics, two full-width-filling cards flush with the title's leading edge.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- README: replace the stale M0/M2-in-flight status with reality — M1 hardened, M2
GameStream host live to stock Moonlight, M3 punktfunk/1 validated, M4 Apple first
light, web console + unified host; FFmpeg 7/8; Bazzite-deployed. Layout adds
web/, packaging/, native_pairing, dualsense.
- CLAUDE: protocol-growth item now reflects the unified host + web-console native
pairing (done) and flags the next steps; layout updated.
- roadmap §7 Windows: de-risked via SudoVDA (the Sunshine Virtual Display Adapter) —
no self-signed kernel IDD needed; the virtual-display backend drops XL→M.
- roadmap §8 (new) Pairing & trust hardening: mandatory PIN pairing by default
(TOFU-open is insecure on a LAN) + delegated pairing approval (an already-paired
device approves a new one, no out-of-band PIN).
- windows-host.md: SudoVDA path throughout (status, table, phasing, effort M not L).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`serve --native` now runs the GameStream host AND the native punktfunk/1 (QUIC)
host in ONE process, sharing a single NativePairing handle with the management API
— so native pairing is operable from the web console instead of journalctl.
- gamestream::serve gains a native_port: spawns crate::m3::serve in the same
runtime and passes the shared NativePairing to mgmt::run. Validated live: one
process binds both RTSP 48010 and QUIC 9777.
- mgmt API: new `native` endpoints — GET /native/pair (status), POST
/native/pair/arm (mint a fresh, time-limited PIN to DISPLAY), DELETE /native/pair
(disarm), GET/DELETE /native/clients (list/unpair). GameStream-only hosts report
enabled:false. OpenAPI regenerated (checked-in doc + drift test).
- main.rs: serve --native / --native-port flags.
The native host arms pairing on demand (the operator reads the PIN from the
console; the SPAKE2 ceremony is host-shows-PIN). New mgmt + native_pairing tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Groundwork for web-UI-driven native (punktfunk/1) pairing. Replaces m3's fixed
startup PIN + local paired store with a shared `NativePairing` (new module):
arm-on-demand with a fresh, time-limited PIN (`arm(ttl)`), `current_pin()` read
per ceremony so a lapsed window stops pairing, plus the trust store (list/add/
remove/is_paired) and a `status()` snapshot. The management API (next commit) and
the QUIC accept loop share one handle. CLI `--allow-pairing`/`--require-pairing`
still arm at startup (no expiry, PIN logged) — back-compat. m3 pairing ceremony +
gate and the C-ABI roundtrip stay green; new unit tests for arm/expire/pair.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The ToolbarSpacer split into separate circles was the wrong read — with the
inline-large title row in place, the expected header is the single grouped pill
(the system default for adjacent trailing items). Dropped the spacer and the
availability fork; the two trailing items now share one pill next to the title.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The home screen stacked the toolbar row above the large title; the modern (iOS 26
Liquid Glass) header puts the large title leading and the glass action circles trailing
on the SAME row. That's exactly .toolbarTitleDisplayMode(.inlineLarge) — applied on iOS
only, macOS keeps its window chrome untouched.
Verified in the iPhone 17 simulator: "punktfunk" large title left, gear/+ circles
right, one row.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The iOS chrome inherited macOS dialog sizing and read as undersized on a phone:
- Toolbar: the two trailing actions shared one compact glass pill; on iOS 26+ each now
gets its own full-size circle (explicit .topBarTrailing placements split by a fixed
ToolbarSpacer — the system-app look, e.g. Files), with the grouped-pill fallback on
iOS 17–18. The buttons are extracted so macOS keeps SettingsLink + .help untouched.
- Sheets and CTAs (AddHostSheet, PairSheet, trust card, empty-state Add Host) get
.controlSize(.large) on iOS — proper touch targets instead of macOS dialog buttons.
Verified in the iPhone 17 simulator: two ~44 pt glass circles matching the Files app's
toolbar sizing; macOS suite and app build unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The whole client now runs on iPadOS/iOS from the same sources, first-lit live in the
iPad simulator against the real host at 1280x720@60 (60 fps on the HUD, capture state
machine active, mic permission flow shown).
- PunktfunkCore.xcframework grows iOS device + universal-simulator slices
(BUILD_IOS=1; rustup targets aarch64-apple-ios{,-sim} + x86_64-apple-ios).
- The decode pump is extracted into a shared StreamPump (identical IDR re-gate logic on
both platforms); the iOS StreamView (StreamViewIOS.swift) has the same name/signature
as the macOS one, so ContentView & co. are byte-identical across platforms — hosted
in a UIViewController for prefersPointerLocked (the iPadOS cursor capture; see README
note 9 for the UIHostingController forwarding caveat).
- Touch is always forwarded: per-finger wire ids, coordinates mapped through the
aspect-fit letterbox into LIVE host-mode pixels (surface == host mode, identity
rescale host-side; follows mid-stream requestMode switches).
- InputCapture is cross-platform: GC works the same on iPadOS, ⌘⎋ is detected from the
HID stream there; stale-⌘ tracking after focus loss fixed on both platforms
(releaseAll now drops the modifier/latch state — a ⌘ released in another app
otherwise hijacked Esc forever).
- SessionAudio: AVAudioSession on iOS (.playAndRecord + .defaultToSpeaker — without it
iPhones route host audio to the EARPIECE; deactivated with
notifyOthersOnDeactivation on stop so interrupted background audio resumes); HAL
device pinning + the Settings pickers stay macOS-only.
- New Punktfunk-iOS app target (shared synchronized sources, generated Info.plist with
mic + local-network usage descriptions — QUIC to a LAN host trips local network
privacy on real devices — scene manifest + indirect input events for Stage Manager /
external displays), shared scheme, macOS min-window frames gated off iOS.
For the iPad-on-an-external-screen idea: with multiple scenes + indirect input enabled,
Stage Manager iPads can drag the punktfunk window onto the external display and drive
the PC with keyboard/mouse/touch. Known gaps (README note 9): the pointer-lock
preference isn't consulted through UIHostingController (relative mouse works, the local
cursor just stays visible) and AVAudioSession interruptions don't auto-restart audio.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
punktfunk-host builds unchanged against either FFmpeg 7.x (libavcodec 61) or 8.x
(libavcodec 62) — ffmpeg-sys-next auto-detects the system version, and the host's
ffmpeg FFI only touches long-stable APIs. Confirmed by building + running live on a
Bazzite F43 box (FFmpeg 7.1.3): full gamescope capture → zero-copy dmabuf→CUDA →
NVENC H.265 at 1280x720x60, p50 ~0.96 ms. Just doc/spec accuracy, no code change:
- encode/linux.rs + CLAUDE.md: drop the "FFmpeg 8 only" claim; note 7.x/8.x both work.
- rpm spec: add the missing zero-copy GPU build deps the link actually needs —
pkgconfig(gl) + pkgconfig(gbm) (mesa) — and document that -lcuda needs libcuda.so at
link time (NVIDIA host, or the CUDA toolkit stub on a headless COPR/koji builder).
Tracked for a proper fix: make the cuda/gbm/GL FFI dlopen-based like khronos-egl so
the RPM builds on a GPU-less host.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A step-by-step walkthrough for running the host on Bazzite (the immutable
Fedora-Atomic gaming distro): the two install paths (rpm-ostree layering vs the
bootc image), udev + the `input` group, host.env knobs (gamescope-default), the
systemd --user service, firewall ports, verification, and troubleshooting — all
grounded in the packaging/ files. Flags the operator-run COPR, the loopback-only
mgmt port, and that the bundled unit runs the GameStream `serve` host (not m3-host).
Linked from packaging/README.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
PUNKTFUNK_GAMEPAD=dualsense now routes a session's gamepad through a real virtual
DualSense (UHID + hid-playstation) end to end:
- host: a `PadBackend` enum (m3.rs) selects `GamepadManager` (uinput xpad, default)
or the new `DualSenseManager` (dualsense.rs) per session. The manager keeps each
pad's full DsState so touchpad + motion (rich-input plane) persist across
button/stick frames, and services the !Send /dev/uhid fd only on the input thread
(which cycles <=4ms, so the GET_REPORT init handshake completes).
- feedback: `service()` now returns `DsFeedback { hidout, rumble }`. Motor rumble
stays on the universal 0xCA plane (so non-DualSense clients still feel it; manager
dedups change); lightbar / player LEDs / adaptive-trigger effects ride the new
0xCD HID-output plane (host->client) as `HidOutput`.
- rich input: touchpad contacts + motion ride the 0xCC plane (client->host) as
`RichInput`, applied via `DualSenseManager::apply_rich` (merged with button state;
touch normalized 0..65535 -> the touchpad resolution).
- connector + C ABI: `NativeClient::next_hidout` / `send_rich_input`, exported as
`punktfunk_connection_next_hidout` (-> PunktfunkHidOutput) and
`punktfunk_connection_send_rich_input` (<- PunktfunkRichInput); header regenerated.
- reference client: `--rich-input-test` drives the DualSense touchpad + motion and
logs the 0xCD feedback that comes back.
Validated live on-box: a synthetic-source m3-host + client-rs created the real
kernel DualSense, drove 0xCC, and decoded 12 live 0xCD events (the kernel's actual
lightbar/trigger init reports) with the data plane unaffected (600/600 frames).
Adversarial review fixes folded in: the input loop no longer skips the rich drain +
feedback pump on a dropped gamepad event, and the touch contact id is clamped to its
slot. Remaining: the Apple client renders triggers/rumble on a real DualSense.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
4-agent feasibility read converged on three independent walls, any one fatal:
- host capture needs a kernel rebuild (CONFIG_USB_DUMMY_HCD off → no UDC for an
f_uac2 composite gadget; everything else for the gadget IS present);
- near-zero Linux supply (only ~5-10 Proton titles via custom Wine patches emit
it; hid-playstation/Steam-Input/RPCS3 don't);
- Apple client can't faithfully replay PCM haptics (CoreHaptics is discrete
pattern-based; no public CoreAudio channel-3/4 routing).
Advanced haptics ride the DualSense USB *audio* interface, not HID, so the UHID
backend structurally can't carry them. Defer; the reachable 80% ("real DualSense
feel") is adaptive triggers over the HID 0x02 path we already parse + two-motor
rumble. New docs/dualsense-haptics.md records the walls + conditions for a future
go; roadmap §5 updated (HID DualSense backend built & live-validated).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Both directions of the audio plane, on CoreAudio's built-in Opus codec
(kAudioFormatOpus — no bundled libopus; OpusCodec.swift, round trip unit-tested):
- Playback: a drain thread pulls nextAudio() packets, decodes, and writes a priming
jitter ring feeding an AVAudioSourceNode (~20 ms prefill, adaptive to the device's
render quantum so large-buffer devices don't oscillate prime/dropout; a high-water
clamp sheds stall backlog so one network hiccup can't permanently lag audio behind
video; underrun re-primes — one dip, not sustained crackle).
- Mic: a second engine taps the input device, resamples to 48 kHz stereo, Opus-encodes
20 ms chunks and sendMic()s them into the host's virtual PipeWire source. Permission
via AVCaptureDevice (NSMicrophoneUsageDescription added to the Xcode target).
- Settings: Speaker + Microphone pickers (CoreAudio HAL enumeration, persisted by
device UID — "System default" leaves the engine unpinned so it follows macOS device
changes) and a "Send microphone" toggle (default on). Applies from the next session.
- Audio starts with streaming, never during the trust prompt (no host sound — and no
mic uplink — before the user trusted the host); teardown stops audio before close().
Adversarial-review fixes baked in: stop() and the dangling mic-permission callback
share one lock+flag protocol (no hot mic with no owner), the connect-success handler
bails when the attempt was abandoned mid-handshake (no session/mic for a dead window),
SessionAudio gets a deinit backstop (a dropped instance can't pin the connection via
its drain thread), and the render scratch buffer is block-owned (was leaked per
session).
Verified live against the box: remote test decodes 100 host Opus packets to PCM and
the host opens its virtual mic on the first uplinked frame ("punktfunk/1 virtual mic
ready"); on-glass session runs with both engines up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Roadmap #5 (rich DualSense). A UHID device presents a real Sony DualSense to the kernel's
hid-playstation driver (matched by VID 054C/PID 0CE6), which exposes the full controller —
gamepad, motion sensors, touchpad, lightbar/player LEDs, adaptive triggers — unlike the
uinput X-Box-360 pad.
- inject/dualsense.rs: hand-rolled /dev/uhid codec (no bindgen) mirroring the uinput style;
the canonical inputtino 232-byte USB HID report descriptor + the feature-report replies
(calibration 0x05 / pairing 0x09 / firmware 0x20) — answering hid-playstation's GET_REPORTs
during init is REQUIRED or it creates no input devices. DsState::from_gamepad maps a
GameStream/XInput frame → the DualSense input report (buttons/sticks/triggers/dpad, +
touchpad/motion fields); service() answers GET_REPORTs and parses HID OUTPUT (rumble /
lightbar RGB / player LEDs / adaptive triggers) into quic::HidOutput.
- scripts/60-punktfunk.rules: grant /dev/uhid to the 'input' group (like /dev/uinput).
- `punktfunk-host dualsense-test`: standalone validation (no streaming session).
Validated live: `dualsense-test` → hid-playstation binds + loads ff_memless + led_class_
multicolor; the kernel creates "Punktfunk DualSense 0" (event/js gamepad + Motion Sensors +
Touchpad + Headset Jack) at VID 054c/PID 0ce6, plus the lightbar at /sys/class/leds/
input*:rgb:indicator; js shows the Cross button firing + the left-stick sweep. Clippy/fmt
clean, workspace tests green. Wiring into the session (pad-type select, touchpad/motion
routing, HID-output back-channel) is the next commit.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adopts the new ABI surface (still v2, additive):
- PunktfunkConnection.sendMic(_:seq:ptsNs:) — Opus mic frames (48 kHz) to the host's
virtual PipeWire source; enqueue-only, empty data = DTX silence. Wiring the actual
Mac microphone (AVAudioEngine input → Opus) into the app is the follow-up, alongside
audio playback (README note 5).
- PunktfunkInputEvent.touchDown/touchMove/touchUp — absolute pixels + surface size in
flags, host injects via libei ei_touchscreen. Built for the iOS variant; nothing on
macOS emits them yet.
- Loopback round trip now also sends touch events and mic frames (incl. a DTX frame)
through the wrapper.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The new features were Linux-built only and broke the documented macOS gate
(cargo build/test/clippy --workspace) four ways, all fixed following the existing
platform-gating conventions:
- m3.rs: mic_service_thread split into the Linux worker and a non-Linux stub that
drains and drops (sessions still count the datagrams) — opus/PipeWire are
Linux-gated deps, same pattern as audio_thread.
- punktfunk-client-rs: the new `opus` dependency moved into the Linux target table and
--mic-test gated with a warn-and-skip stub (only the synthetic-tone test rig needs
the encoder; the mic uplink itself is portable).
- gamestream/audio.rs: SAMPLE_RATE import gated to any(linux, test) (the frame_sizing
test uses it everywhere, the data plane only on Linux).
- tests/c_abi.rs: the harness's macOS link flags gained Security + CoreFoundation —
the quic feature now pulls rustls's platform verifier into the staticlib.
Also: two clippy match-ref-pats lints in the new rich-input/HID-output decoders
(clippy -D warnings is the repo gate), the regenerated punktfunk_core.h committed (the
checked-in copy predated the rich-input/HID-output constants — CI fails on drift), and
web's inlang cache dir gitignored.
cargo build/test/clippy/fmt --workspace: green on macOS, 122 tests passing.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Foundation for rich DualSense support (roadmap #5). The fixed 18-byte InputEvent (0xC8) can't
hold the DualSense touchpad/motion or HID feedback, so two new variable-length, kind-tagged
datagram families join the side-plane (mouse/keyboard/gamepad/touch keep the fixed InputEvent):
- RICH_INPUT_MAGIC 0xCC, client→host: `[0xCC][kind][fields]`
Touchpad{pad,finger,active,x,y} (x/y normalized 0..65535; host scales to the pad)
Motion{pad, gyro[3], accel[3]} (raw i16, straight into the DualSense report)
- HIDOUT_MAGIC 0xCD, host→client: `[0xCD][kind][pad][fields]` — the rich analog of the 0xCA
rumble datagram (rumble stays on 0xCA):
Led{rgb} PlayerLeds{bits} Trigger{which, effect} (adaptive-trigger params to replay)
`RichInput`/`HidOutput` enums with encode/decode; unknown kinds + truncation decode to None
(forward-compatible). +2 round-trip/disjointness tests; quic suite green, clippy/fmt clean.
Wiring (host UHID device, capture, C ABI, client) lands in following commits.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Roadmap #5 (touch, ahead of the XL UHID DualSense work). Touch fits the existing 18-byte
InputEvent: code = touch id, x/y = client pixels, flags = (w<<16)|h — the same absolute
mapping as MouseMoveAbs.
- core: InputKind::{TouchDown=9, TouchMove=10, TouchUp=11} + from_u8 + roundtrip test.
- host inject/libei.rs: request the RemoteDesktop Touchscreen device type, bind the Touch
capability, and inject ei_touchscreen down/motion/up (one event = one frame, per the
protocol rule), mapping coordinates into the device region like the abs pointer. wlroots
has no virtual-touch protocol wired — no-ops there.
- client-rs --touch-test: drags a synthetic finger (touch id 0) in a circle.
Validated live on headless KWin: the portal GRANTS the Touchscreen device type
(Keyboard|Pointer|Touchscreen), proving the request path — but KWin's EIS server creates no
touchscreen *device*, so touch currently no-ops on this KWin (now logged once, not silent).
The injection code is correct and will land on a backend that exposes ei_touchscreen
(gamescope / a newer compositor / the real touch-client path). Workspace green, clippy/fmt
clean, +1 unit test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A 4-agent read of the host crate: a Windows host is an "add a backend" job, not a parallel
port — ~95% reuse (core/protocol/FEC/crypto/C-ABI, QUIC, GameStream, mgmt, m3/pipeline are all
platform-agnostic and already cfg-isolated). New cfg(windows) backends behind the existing
traits: DXGI Desktop Duplication (capture), Media Foundation / NVENC-SDK (encode), SendInput +
ViGEm (input), WASAPI loopback + virtual mic (audio). The blocker is the virtual-display
feature — no user-mode Windows API; it needs a signed kernel-mode IDD driver (XL).
docs/windows-host.md records the per-subsystem effort + a phased plan (Phase 0 = a "basic
Windows host" capturing an existing monitor, smallest surface). Deferred: large and unbuildable
on the Linux dev box, per the request to only take it on if manageable. roadmap.md marks
#1/#2/#4 done, #3 packaged, and adds #7 Windows.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Roadmap #3 (install on other devices). Bazzite already ships gamescope + PipeWire + the
NVIDIA stack, so the host slots in with minimal new deps (ffmpeg-libs from RPM Fusion + opus
+ libei).
- packaging/rpm/punktfunk.spec — builds punktfunk-host from source (cargo), installs the
binary + udev rule + systemd user unit + headless helpers; Requires/Recommends mapped from
the Ubuntu bootstrap deps to Fedora.
- packaging/bootc/Containerfile — layer punktfunk into a bazzite-nvidia bootc image for
atomic, image-based installs.
- packaging/bazzite/host.env — gamescope-default appliance config (spawned per session).
- packaging/copr/ + packaging/README.md — COPR build-from-SCM settings + install docs
(rpm-ostree and bootc paths), and why not Flatpak.
- LICENSE-MIT + LICENSE-APACHE — materialize the declared `MIT OR Apache-2.0` (was unfiled);
the RPM ships them.
Not buildable on the Ubuntu dev box (no rpm tooling) — the COPR/Fedora build is operator-run;
all spec-referenced files verified present and the cargo build is green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The inverse of the host→client audio path: the client's mic, Opus-encoded, rides a
new 0xCB QUIC datagram to the host, which decodes it into a virtual PipeWire
Audio/Source its apps can record from (voice chat, etc.).
Protocol (punktfunk-core):
- MIC_MAGIC 0xCB + encode/decode_mic_datagram (mirror of the 0xC9 audio datagram).
- NativeClient::send_mic(seq, pts_ns, opus) over a new outbound channel + worker task
(mirror of send_input); C ABI punktfunk_connection_send_mic for native clients.
Host:
- audio::VirtualMic + PwMicSource: a PipeWire output stream tagged media.class=
Audio/Source (Direction::Output) — a recordable microphone node, fed decoded PCM.
- MicService: host-lifetime owner of the source + Opus decoder (mirror of
InjectorService / the audio capturer slot); lazily opened, persists across sessions,
self-heals. The per-session datagram reader now demuxes 0xCB→mic / 0xC8→input over a
single read_datagram loop (two loops would race).
- Adaptive jitter buffer in the producer: primes to ~3 consumer quanta before emitting,
so the 5 ms push / N ms pull clock skew never underruns — without it ~58% of output
was silence; with it, glitch-free across consumer quanta.
Client: punktfunk-client-rs --mic-test streams a synthetic 440 Hz Opus tone as the mic
uplink (opus dep added) for end-to-end validation without a real microphone.
Validated live on headless KWin: client tone → host source → pw-record shows the
punktfunk-mic Audio/Source node, 440 Hz dominant (Goertzel power 20.7 vs <0.001
elsewhere), RMS 0.179 ≈ the ideal 0.177, 0.3–0.4% silence at both 256 ms and 10 ms
consumer quanta. Tests +1 (mic datagram roundtrip); workspace green, clippy/fmt clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A full Plasma login starts several pieces our bare headless session was missing, which
surfaced as three separate failures while streaming the KDE desktop:
- Steam (and other X11 apps) failed "can't open display": Xwayland runs, but KWin only
sets DISPLAY for its own children — apps launched via the plasma menu / D-Bus activation
never saw it. Detect the Xwayland display after KWin is ready and export it into the
systemd/D-Bus activation environment.
- Discover / PackageKit couldn't install apps: polkitd (the policy engine) was running but
no authentication *agent* (the prompt) was — so privileged installs got no authorization.
Start polkit-kde-authentication-agent-1 (forcing the Qt Wayland platform, or it exits).
- The streamed desktop showed app windows but no wallpaper/panels: plasmashell had crashed
and the old unsupervised `plasmashell &` never brought it back. Supervise it — restart for
as long as KWin lives, so the desktop shell self-heals.
Validated live on this box: DISPLAY=:0 now in the --user environment (xdpyinfo on :0 works),
the polkit agent registers ("Listener online"), and plasmashell stays up under the supervisor.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Under rapid client reconnects, KWin's libei/EIS input setup intermittently wedged
with "EIS setup timed out", causing total input loss for affected sessions. Root
cause: each punktfunk/1 session opened (and tore down) its own RemoteDesktop-portal
CreateSession for pointer/keyboard injection, and back-to-back reconnects raced a
prior session's portal teardown before it settled.
LibeiInjector is only a Send channel handle to a worker thread that owns the portal
session, so the injector can live for the whole host run instead of per session.
Adds InjectorService: one host-lifetime thread owns the (!Send) injector, opened
ONCE (lazily, on the first event) and reused across every session — the portal grant
is established a single time and held. Sessions forward pointer/keyboard events to it
over a clonable Send channel; gamepads stay per-session (uinput, no portal). The
service self-heals — reopen after a 2s backoff if open fails or the backend worker
dies (covers a gamescope EIS socket that respawns with its nested session).
Mirrors the existing host-lifetime audio-capturer slot; the audio capturer is Send
(a slot works), the injector is !Send (needs the owning thread + channel).
Validated live on headless KWin: 8 rapid back-to-back input sessions →
"input injector ready (host-lifetime)" exactly once, ZERO "EIS setup timed out",
8/8 sessions injected input. Tests green, clippy/fmt clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Scroll was wired to GCMouse's scroll dpad, which only fires for plain HID wheel
deltas — trackpad and Magic Mouse scrolling are gesture events that never reach
GameController, so scrolling was dead on the default Mac setups. The stream view now
overrides scrollWheel (while captured the cursor is parked mid-view, so it receives
every scroll event) and feeds InputCapture.sendScroll: precise gesture deltas are
pixels (~0.1 notch/px, SDL's factor → ×12 for WHEEL_DELTA(120)), classic wheels are
lines (×120), fractional remainders accumulate, and the GC scroll handler is gone so
wheel mice can't double-deliver. Signs pass through as-is, preserving the local
(natural-)scrolling preference.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Adopts punktfunk_connect_ex from the compositor-selection batch: a Compositor enum on
PunktfunkConnection (auto/kwin/wlroots/mutter/gamescope, with the host's name aliases
for env parsing), a "Host compositor" picker in Settings (default Automatic — a
concrete choice is honored only if that backend is available host-side), and
PUNKTFUNK_COMPOSITOR / PUNKTFUNK_REMOTE_COMPOSITOR pass-throughs for the autoconnect
dev hook and the remote first-light test. The wire change is backward-compatible
(optional trailing byte), so no behavior changes at the default.
Validated live against the box: host with no compositor env (auto-detect = KWin)
logged "honoring client compositor request compositor=gamescope" and streamed 60/60
decoded frames from the spawned gamescope.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A client can now request which compositor backend the host drives its virtual
output on (gamescope/KWin/Mutter/wlroots). The host honors the request if that
backend is available, else falls back to auto-detect and reports the resolved
choice back — wire-compatible both directions (no ABI bump).
Protocol (punktfunk-core):
- New CompositorPref (config.rs): Auto|Kwin|Wlroots|Mutter|Gamescope with
u8/name mappings. Appended as one optional byte to Hello (client preference)
and Welcome (host's resolved choice). Both decoders already tolerate trailing
bytes, so old↔new interop is preserved — ABI_VERSION stays 2. Round-trip +
back-compat (truncated-message) tests.
- C ABI: punktfunk_connect_ex(compositor) + PUNKTFUNK_COMPOSITOR_* constants;
punktfunk_connect delegates with AUTO, so the existing symbol is unchanged.
NativeClient::connect / worker_main thread the preference through.
Host:
- vdisplay::available() enumerates usable backends via cheap, side-effect-free
probes (KWin zkde global, gamescope binary+version, GNOME/Sway env), plus
Compositor id/label/as_pref/from_pref/all helpers.
- m3 handshake resolves the preference to a concrete backend during the
handshake (pick_compositor pure + resolved logging), reports it in Welcome,
and threads it into virtual_stream (replacing the unconditional detect()).
- mgmt GET /v1/compositors lists every backend with availability + the
auto-detected default (OpenAPI regenerated).
Client:
- punktfunk-client-rs --compositor NAME; logs the host's resolved choice from
the Welcome ("session offer … compositor=…").
Web console:
- Host page gains a Compositors card (availability + default badges) via the
codegen'd useListCompositors hook; en/de strings added.
Also fixes a pre-existing, env-dependent test-isolation bug:
mgmt::tests::paired_clients_list_and_unpair seeded the real
~/.config/punktfunk/paired.json (AppState::new loads it), so a real
GameStream-paired client leaked into body[0] on a dev box — now cleared first.
Live-validated against headless KWin: --compositor kwin honored, --compositor
mutter falls back to kwin (available=[kwin, gamescope]), resolved choice
round-trips to the client. Tests: +6 (wire/back-compat, resolution precedence,
endpoint); workspace green, clippy/fmt clean, C ABI harness PASS at abi_version=2,
web typecheck + build clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Hardens the virtual-display → capture → encode bring-up against the transient
failures that surfaced as black screens / wrong refresh on cold KDE sessions.
- m3: build_pipeline_with_retry wraps the initial vd.create() + first-frame with
bounded exponential backoff (4 attempts, 500ms→2s). is_permanent_build_error
classifies config/version/missing-tool failures so they fail fast instead of
burning the retry budget. Encoder + frame clock now pace to the *achieved*
refresh reported in VirtualOutput::preferred_mode, not the requested rate.
- capture/linux: PortalCapturer::Drop sends a pipewire channel quit and joins the
thread, so a dropped/failed/retried capturer releases its PipeWire thread + EGL/
CUDA context promptly instead of leaking it to process exit. First-frame timeout
now reports the node id and distinguishes "format never negotiated" from
"negotiated but no buffers arrived" via a negotiated flag set in param_changed.
- vdisplay/kwin: set_custom_refresh reads back the active mode from kscreen-doctor
and returns the refresh KWin actually gave us (a rejected custom mode silently
leaves the output at 60Hz); create() carries it into preferred_mode.
- vdisplay/gamescope: find_gamescope_node requires the Video/Source object (the
node.name=gamescope tag is on two objects; the other wedges the link); a version
check warns on <3.16.22 (the PipeWire-1.6 capture-deadlock signature).
Live-validated against headless KWin: 720p120 build with requested=120 achieved=120,
zero-copy CUDA frames, and no per-session thread accumulation across back-to-back
sessions. Tests: +3 unit (retry classifier, gamescope version parse); 49 host tests
green, clippy/fmt clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Capture used to engage whenever the app became active, so the click that activates the
window — on the title bar (a drag) or a resize edge — got the cursor warped away
mid-gesture, and raw deltas kept streaming to the host while the user fought the window.
Reworked Moonlight-style, with capture as a deliberate, reversible state owned by
StreamLayerView:
- Engage: automatically once when the stream starts / trust is confirmed (one-shot, can
never fire surprisingly later), or by clicking into the video (that click's
press/release are suppressed toward the host; acceptsFirstMouse makes it one click
from another app). NEVER on app re-activation.
- Release: ⌘⎋ (toggles, key-window-scoped), focus loss — now including same-app window
switches (⌘, / ⌘N / ⌘M resign key without resigning the app; previously the new
window inherited a hidden frozen cursor and its typing was double-delivered to the
host) — and disconnect.
- While released: nothing is forwarded (InputCapture.forwarding gates the GC handlers;
held keys/buttons are flushed host-side so nothing sticks), the cursor is free, and
the HUD (now showing the capture state) is clickable.
- The no-beep behavior moved from the NSEvent monitor to first-responder key
consumption — swallowing at the monitor risked starving GC's own delivery (the
"input broken altogether" report). The monitor now only intercepts ⌘⎋.
- Adversarial-review fixes: a second session preempts the previous one cleanly instead
of leaving it captured with dead GC handlers (onPreempted); the engage click's
suppression latch can't outlive the click (mouseUp backstop); ⌘⎋'s physical Esc can't
type into the host in either toggle direction (suppressedVK latch + Esc-while-⌘
guard); capture callbacks defer out of the SwiftUI update pass.
Validated live against the box: 16185 input datagrams injected during a captured
session (gamescope EIS), title-bar drag/resize free while released, and visible
cursor + typing on a streamed KWin desktop, all user-confirmed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
GCKeyboard reads the HID state directly, so the key NSEvents kept traveling the
responder chain unhandled — and an unhandled keyDown makes NSWindow play the
"invalid input" sound on every keystroke. InputCapture now installs a local event
monitor for its lifetime that swallows key events, except ⌘-combos, which still
reach the local app (the HUD's ⌘D disconnect, ⌘Q) in addition to the host.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The pairing/renegotiation batch bumped the punktfunk/1 ABI to v2 and the host now
hard-rejects v1 Hellos (m3.rs), so streaming from the Mac was dead until the bundled
PunktfunkCore.xcframework is rebuilt — it is gitignored, so that is a per-checkout step:
bash scripts/build-xcframework.sh. The Swift wrapper itself was already adapted upstream;
this lands the app on top of it.
- ClientIdentityStore: persistent client identity in the login Keychain, presented on
every connect so paired hosts recognize this Mac. Keychain access failure throws
instead of regenerating (a fresh identity would silently un-pair this Mac from every
--require-pairing host); a lost first-run race resolves toward the stored identity;
pairing uses the strict loadForPairing() so a memory-only identity can't strand a
ceremony.
- PairSheet: the SPAKE2 PIN ceremony, reachable from a host card's context menu and from
the trust prompt's "Pair with PIN instead…" (which drops the live session first — the
host's accept loop is sequential). Success pins the verified fingerprint and connects;
an in-flight ceremony self-discards when the sheet is dismissed, so a late success
can't pin + auto-connect behind the user's back. Wrong PIN and Keychain failures get
distinct, actionable error text.
- Tests: identity unit tests; the full pairing ceremony + --require-pairing gate on
loopback (test-loopback.sh arms a second host, parses its PIN from the log, and gives
both hosts throwaway config homes — no more writes to the real ~/.config/punktfunk);
remote pairing + pinned stream over the LAN (PUNKTFUNK_REMOTE_PIN, _PORT).
Validated live against the box: SPAKE2 ceremony with the host's arming PIN → verified
fingerprint → pinned + identified 720p60 session (host persisted the client identity);
first light 60/60 AUs decoded to pixels; vkcube on glass through the app.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A synchronous systemctl try-restart of the portal chain (xdg-desktop-portal is Type=dbus,
waits for its bus name) blocked the script ~30-40s before plasmashell started. --no-block
queues the restart and returns immediately — the portal only needs to be ready before the
first client streams (seconds later), not before plasmashell. Validated: plasmashell up in
1s (was ~30s); a virtual capture session against the fresh session streamed 720/720 frames
@720p120, zero-copy CUDA, no black screen.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Headless KDE startup was a chain of timing-sensitive handoffs gated by a blind `sleep 2`,
the dominant source of black screens. Phase-1 fixes:
- New `punktfunk-host probe-compositor` subcommand: exits 0 iff the detected compositor is
up AND ready to create a virtual output now. KWin gets a real check (connect + registry
roundtrip + the privileged zkde_screencast global must be advertised — what the backend
needs); gamescope/Mutter/wlroots create on demand so the probe just confirms Linux.
(vdisplay::probe dispatcher + kwin::probe; reuses kwin.rs's existing roundtrip path.)
- run-headless-kde.sh: replace `sleep 2` with an active readiness wait (poll probe-compositor
until ready, 30s deadline, and bail with kwin's log if kwin_wayland exits during init).
Move the portal restart to AFTER readiness, and precede it with `systemctl --user
import-environment` + `dbus-update-activation-environment` (the missing env import — the
Sway script does this; without it a restarted portal inherits a stale/empty WAYLAND_DISPLAY,
which is the "streams but eats no input/audio" failure). kwin's stderr → a log file.
Validated: probe-compositor exits 0 "Kwin ready" against the live session, exit 1 with a
clear diagnostic when the compositor is absent. 114 tests green, clippy/fmt clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Research-grounded sequence + per-goal approach/effort. Decisions: start with KDE startup
reliability; Bazzite via COPR RPM then bootc image; commit to full UHID DualSense.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Multi-agent security review of 9856c04 (4 dimensions, 2-skeptic verification):
- CRITICAL functional+security: the session cookie inherited h3's Secure=true default;
browsers DROP Secure cookies over plain http://, so login silently failed on a LAN HTTP
client (worked only on localhost, a secure context — which is why the live test passed).
Now set the cookie attributes explicitly: HttpOnly + SameSite=Lax + Path=/, and Secure
only when PUNKTFUNK_UI_SECURE=1 (behind TLS). Verified: Set-Cookie no longer has Secure.
- Gate bypass: isPublicPath allowlisted any path ending in .json/.css/.png/etc., so
/api/v1/openapi.json (served unauthenticated on the mgmt side too) leaked the whole API
schema through the token-injecting proxy. Now /api is ALWAYS gated and the generic
extension allowlist is gone (client assets are all under /assets/, still allowlisted).
Verified: /api/v1/openapi.json and /api/v1/status.json → 401.
- Session lifetime: added maxAge (7d) — bounds a stolen cookie (cookie Max-Age + iron seal
TTL); previously never expired.
- Open redirect: the post-login `next` accepted protocol-relative `//evil.com`. Hardened
client + added safeNextPath() (same-origin path only).
Re-validated end to end: login assets public (200), /api/openapi.json gated (401), authed
/api/v1/status (200), unauth /→302. tsc + build green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Single-user, LAN-reachable-but-gated. The web server is a backend-for-frontend:
- Login: POST /_auth/login {password} checks PUNKTFUNK_UI_PASSWORD (constant-time) and
sets a SEALED session cookie (h3 useSession / AES-GCM). server/middleware/auth.ts gates
every request — pages 302 → /login, /api → 401 — and FAILS CLOSED (503) when
PUNKTFUNK_UI_PASSWORD is unset, so a misconfigured LAN-exposed server admits no one.
- The management API stays loopback-only + token (never LAN-exposed). The proxy
(server/routes/api/[...].ts) injects PUNKTFUNK_MGMT_TOKEN server-side and drops the
browser's cookie before forwarding — the token never reaches the browser, which only
holds the session cookie.
Nitro doesn't auto-scan a server/ dir, so the Nitro plugin gets an explicit scanDirs to
pick up middleware + routes. Client: removed the localStorage token (server injects it);
the fetcher bounces to /login on 401; new /login page (bare, no shell); Settings drops the
token field and gains a Sign-out button; en/de strings.
Validated live end to end: unauth /→302, /api→401; wrong pw→401; right pw→200+cookie;
authed /api/v1/status→200 (proxied, mgmt token injected — the host required it); logout→
session cleared→401. tsc + build green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The earlier "render the shell with a custom script" was a hack. The real issues were a
version matrix and a missing server target:
- TanStack Start's start-plugin-core peer-requires Vite >= 7; on Vite 6 the build's
prerender/post-build buildApp plugin hook silently doesn't run (Vite 6 lets a
config-level builder.buildApp suppress plugin buildApp hooks; Vite 7 runs both). Pinned
Vite ^7 + @vitejs/plugin-react ^5 (v5 ↔ Vite 7; v6 needs Vite 8 / vite/internal).
- Added @tanstack/nitro-v2-vite-plugin with the `bun` preset — the server/deploy target.
`bun run build` → .output/ (bun-runnable server + .output/public). `bun run start` =
`bun run .output/server/index.mjs`.
- Full SSR instead of SPA mode: SPA-shell prerender points its preview server at the old
dist/server/server.js path that Nitro relocates, breaking the build. The Nitro server
renders the shell per request; React Query fetches client-side after hydration.
- Nitro routeRules proxy /api/** → PUNKTFUNK_MGMT_URL (default 127.0.0.1:47990), so the
browser stays same-origin (bearer token rides along, no CORS).
Toolchain is now Bun (package manager + runtime): bun.lock replaces pnpm-lock.yaml;
scripts/prepare/start use bun. Validated live: bun build → .output, bun server SSR-renders
the console on :3000 and proxies the API (health/host return through it). tsc clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
TanStack Start's dev server requires a React Refresh plugin; without it `/@react-refresh`
404s, the client entry 500s, and nothing hydrates (blank screen — the production build was
unaffected since rollup handles JSX there). Pinned to the v4 line: plugin-react 6 imports
`vite/internal` (Vite 7 only) and we're on Vite 6. Must sit after tanstackStart() in the
plugin list.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Browser UI for the host's management REST API (mgmt.rs / docs/api/openapi.json).
Stack, exactly as specified:
- TanStack Start (Vite, SPA mode) — file-based routes, SSR shell + client hydration.
- React Query via orval codegen from the checked-in OpenAPI spec: a custom fetch mutator
(src/api/fetcher.ts) centralizes the base URL, the bearer token (Settings → localStorage),
JSON, and a throwing ApiError; the query client skips retries on 4xx. orval returns the
response body directly (includeHttpResponseReturnType:false) so a query's `.data` is the
typed payload; GET→useQuery, POST/DELETE→useMutation by method.
- shadcn/ui on Tailwind v4 (CSS-first tokens, dark-first) — button/card/badge/input/label/
table/skeleton primitives hand-authored from the canonical source.
- Paraglide i18n (en + de) with a reactive useLocale() hook and a language switcher.
Pages: dashboard (live status — video/audio/session/stream, stop-session + request-IDR,
2s polling), host (identity/codecs/ports), clients (paired list + unpair), pairing (PIN
submit, polls pin_pending), settings (API token + language).
Dev server proxies /api → 127.0.0.1:47990 (same-origin, no CORS; PUNKTFUNK_MGMT_URL to
override). Generated code (orval client, paraglide runtime, routeTree) is gitignored and
reproduced by `pnpm codegen` (prepare/pre* scripts). Validated live against `serve`: API
shapes match, dev proxy works, SSR shell renders the localized nav, build + tsc green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Triaged the multi-agent review of the renegotiation + pairing + Sway + AV1/surround batch
(1 critical, 11 major/minor confirmed). Fixes:
CRITICAL — PIN pairing was offline-brute-forceable. The HMAC-of-PIN proof let an active
MITM who terminates the TOFU ceremony recover the 4-digit PIN by offline dictionary search
(all other inputs observable) and forge a correctly-bound proof. Replaced with **SPAKE2**
(balanced PAKE, `spake2` crate) + key-confirmation MACs, binding both cert fingerprints as
the SPAKE2 identities: an attacker gets exactly ONE online guess, no offline search, and
mismatched cert views (a real MITM) never reach a shared key. Also reworked the UX to an
"arming PIN" — one PIN per arming window shown at host startup (the SPAKE2 client needs the
PIN to build its first message, so it can't be minted per-connection). Validated live:
wrong PIN rejected in 0.1s, right PIN pairs + persists + the paired identity streams.
Pairing hardening: `--allow-pairing`/`--require-pairing` must arm pairing (default rejects
unsolicited ceremonies); per-host cooldown bounds online guessing; the client flushes its
CONNECTION_CLOSE so a refused ceremony can't wedge the sequential host for the full timeout;
atomic (temp+rename) paired-store writes.
Protocol: control/pairing messages use a distinct CTL_MAGIC (PKFc) — fully disjoint from
the positional Hello namespace (a future abi_version can't be misparsed as a control
message); all typed decodes are length-exact. ABI_VERSION → 2 (punktfunk_connect signature
gained the identity params; header regenerated).
Renegotiation: drain the reconfig channel to the NEWEST mode (one rebuild, not one per
stale step); validate refresh_hz; build the new pipeline BEFORE dropping the old so a
rebuild failure keeps the session on its current mode instead of killing it.
GameStream: packetDuration snaps to {5,10} (an in-between value isn't a legal Opus frame
size and would kill audio). Sway: chooser file moved to $XDG_RUNTIME_DIR (was a fixed
world-writable /tmp path — DoS / capture-misdirection by another local user).
Swift: fixed two compile breakers in the new pairing/identity APIs (Int32 status .rawValue,
UInt cap cast). New SPAKE2 + namespace-disjointness + pairing-roundtrip unit tests; the
in-process pairing test now also exercises the arming PIN + cooldown. 114 tests green,
clippy -D warnings clean (both feature sets), fmt, C-ABI harness.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Renegotiation (no reconnect on resize): the handshake bi-stream stays open; the client
sends Reconfigure{mode} (typed post-handshake message), the host validates + acks
Reconfigured and rebuilds capture/encoder/virtual output at the new mode while the data
plane (keys, ports, FEC) runs untouched — the first new-mode AU is an IDR with in-band
parameter sets. NativeClient::request_mode / punktfunk_connection_request_mode; mode()
reflects the active mode. Validated live on KWin: one continuous stream, 225 frames
@1280x720 then 395 @1920x1080, ~90 ms pipeline rebuild (ffprobe shows both resolutions).
PIN pairing (mutual trust, kills TOFU MITM): clients get persistent self-signed
identities presented via QUIC client auth (generate_identity / client auth offered but
optional server-side — legacy clients still connect). Ceremony on the control stream:
PairRequest{name} → host shows a 4-digit PIN (log) + PairChallenge{salt} → client proves
with HMAC-SHA256(PIN‖salt, client_fp‖host_fp) — binding both certs means a MITM can't
forward a proof, single attempt per PIN, constant-time compare → PairResult; host
persists the fingerprint (~/.config/punktfunk/punktfunk1-paired.json), client pins the
host's. m3-host --require-pairing gates sessions on the paired set.
NativeClient::pair + punktfunk_pair/punktfunk_generate_identity in the ABI; reference
client: --pair PIN --name LABEL + auto-generated persistent identity, --remode for live
renegotiation testing. Swift wrapper: ClientIdentity/generateIdentity()/pair(),
requestMode()/currentMode(); README handoff updated.
Tested: reconfigure/pairing wire roundtrips, C-ABI mode switch ack, full in-process
ceremony (wrong PIN → Crypto, anonymous-vs-gate rejection, success → pinned session);
live wrong-PIN ceremony against the serving host (PIN logged, proof rejected).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Codec negotiation (M2 polish):
- ServerCodecModeSupport now advertises what we encode: H264|HEVC|AV1_MAIN8
= 65793 (flags verified against moonlight-common-c Limelight.h). The old
placeholder 3843 wrongly claimed HEVC Main10 + 4:4:4 and no AV1. Main10
bits stay off on purpose: Moonlight ties 10-bit to HDR, and capture is
8-bit SDR BGRx with no HDR metadata path (av1_nvenc -highbitdepth was
validated working for later).
- RTSP ANNOUNCE: bitStreamFormat 0/1/2 -> H264/HEVC/AV1 (already plumbed to
av1_nvenc; validated e2e via `m0 --codec av1` + ffprobe av01), and a
dynamicRangeMode!=0 request now logs + falls back to 8-bit SDR.
Surround audio (M2 polish):
- ANNOUNCE x-nv-audio.surround.{numChannels,AudioQuality} +
x-nv-aqos.packetDuration -> per-session AudioParams; DESCRIBE advertises
all six Opus configs (normal before HQ per channel count). Normal-quality
mappings are pre-rotated for the client's GFE-order LFE swap
(RtspConnection.c, verified verbatim) so its derived decoder mapping
equals our encoder mapping — including 7.1, where Sunshine's rotate only
covers [3,6) and scrambles LFE/SL/SR.
- 5.1/7.1 encode via libopus multistream (audiopus_sys, the sys layer the
opus crate already links) with Sunshine's layouts/bitrates, RAII wrapper;
the live-validated stereo wire is byte-identical (plain Opus, no FEC).
- Surround sessions add Sunshine-style RS(4,2) audio FEC (packetType 127 +
AUDIO_FEC_HEADER, the OpenFEC parity matrix both ends hardcode, nanors
gemm semantics verified from nanors/rs.c).
- PipeWire capture generalized to the negotiated channel count with explicit
FL FR FC LFE RL RR [SL SR] positions; missing sink channels are zero-
filled by the channel-mixer. PwAudioCapturer now tears down cleanly on
Drop (pipewire channel -> loop quit), so a channel-count change can
reopen without leaking a capture stream.
Tests: serverinfo mask, RTSP codec/audio param parsing, DESCRIBE contents,
surround-params strings + client-swap round trip, FEC parity self-recovery
and packet layout, real-codec 5.1 channel-identity round trip, and an
ignored live test (ran green against a 6ch null sink monitor).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The fourth VirtualDisplay backend: `swaymsg create_output` adds a HEADLESS-N
output (name found by diffing get_outputs), `output <NAME> mode --custom
WxH@HzHz` sets the client's exact mode (and the refresh clock a fresh headless
output needs to produce frames at all), and the PipeWire node comes from the
ScreenCast portal. Headless output selection is non-interactive via
xdg-desktop-portal-wlr's chooser hook: a managed config (chooser_type=simple,
chooser_cmd cats /tmp/punktfunk-xdpw-output; portal try-restarted when the
config changes) plus a per-session `Monitor: <NAME>` written to that file.
Teardown is RAII: drop ends the portal thread (zbus connection drop ends the
cast) then `swaymsg output <NAME> unplug`. swaymsg commands go after `--` so
tokens like `--custom` reach sway instead of swaymsg's getopt.
Validated live on headless sway 1.11 (gles2-on-NVIDIA, xdpw 0.8.1), zero-copy
dmabuf→CUDA on both runs: 720p60 257 frames p50 0.77 ms, 1080p60 480/480
frames p50 1.18 ms, output unplugged with the session both times. The
checked-in xdpw.config sample now matches the managed config (the old
chooser_type=none/HEADLESS-1 form would pin capture to the wrong output).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The awaiting-trust and streaming phases rendered StreamView in different switch
branches, so confirming trust dismantled and recreated the NSView — the fresh pump had
already missed the opening IDR (infinite GOP: no other keyframe ever comes) and decoded
nothing. One session branch now hosts a single StreamView; the trust card is an overlay
on the blurred stream and only the capturesCursor flag flips on confirmation.
Verified live against the box (gamescope+vkcube at 720p60, 11.7 Mb/s on glass). Note for
host runs: without PUNKTFUNK_COMPOSITOR=gamescope + PUNKTFUNK_GAMESCOPE_APP, m3-host
auto-picks KWin and streams its (black, empty) session — looks identical to a client
bug but isn't one.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The portal processes bind to the compositor that existed when they started; after a kwin
restart the stale instances point at a dead socket and RemoteDesktop/EIS input injection
times out ("EIS setup timed out"). Hit live: a fresh session streamed fine but ate no
mouse/keyboard until the portals were restarted.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The app grows from a dev connect form into a real client shell:
- Home is a grid of saved hosts (UserDefaults-persisted; context menu: Remove / Forget
Identity), "+" in the toolbar opens the add-host sheet, the stream mode moved into
Settings (⌘, / gear) — native resolution stays the only mode, no scaling.
- Trust is now explicit: the protocol always supported certificate pinning, but the app
passed no pin and discarded the observed fingerprint — silently trusting any host.
First connect now shows the host's SHA-256 fingerprint (compare with the "clients pin
this fingerprint" line in the host log) over the live-but-blurred stream; the stream
must pump immediately (the opening IDR is the only guaranteed one), so StreamView gains
a capturesCursor switch to keep the cursor free while the prompt needs clicking, and
input capture starts only after confirmation. Trusting pins the fingerprint per host;
a changed host identity then refuses to connect.
- PUNKTFUNK_AUTOCONNECT keeps working (auto-trusts, doesn't touch the saved hosts).
Host→client authorization (pairing PIN) remains a punktfunk-core roadmap item — the host
still accepts any client that can reach its port.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
punktfunk_Logo.icon (Icon Composer 2.0) in App/, ASSETCATALOG_COMPILER_APPICON_NAME set.
Compiles with Xcode 27 beta's actool; Xcode 26.5's actool crashes on EVERY .icon file
(known regression, Apple FB20183399, expo/expo#46121) — build with the beta (or 26.4.1)
until a 26.x fix lands. The icon itself is fine.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
clients/apple was a bare Swift package — fine for swift run/test, but app icons, a real
bundle (Info.plist, signing identity, TCC), and the normal Xcode build/run flow need an
app target. Punktfunk.xcodeproj (synchronized-folder format) wraps the SAME sources as
the CLI dev shell (Sources/PunktfunkClient) plus App/Assets.xcassets, and links
PunktfunkKit from the local package — no source duplication, both flows stay green:
swift build / swift test / swift run PunktfunkClient, and xcodebuild -scheme Punktfunk.
The asset catalog ships an empty AppIcon slot ready for the Icon Composer .icon
(drag in + set as App Icon + drop the placeholder; see README — including the actool
crash observed with the current icon bundle). Package tests on ⌘U need one GUI step
(Edit Scheme → Test → +); a hand-written package-test scheme reference doesn't resolve
headlessly.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The host renders its own cursor from our raw deltas, so the local macOS cursor both
stays visible and drifts away from the remote one — and it can wander out of the window,
where a click focuses another app. While the stream has focus, do what Moonlight does:
warp the cursor mid-view, disconnect it from mouse movement
(CGAssociateMouseAndMouseCursorPosition(false) — GCMouse still delivers raw HID deltas),
and hide it. Released on app deactivation (Cmd+Tab is the escape hatch), view teardown,
and disconnect; re-captured when the stream regains focus. The HUD's Disconnect gains ⌘D
since a hidden, frozen cursor can't click it.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Full project rename, decided 2026-06-10:
- Crates/binaries: punktfunk-core / punktfunk-host / punktfunk-client-rs.
- C ABI: punktfunk_* symbols, Punktfunk* types, include/punktfunk_core.h,
PUNKTFUNK_FEATURE_QUIC guard (header regenerated; cbindgen renames updated, incl.
PUNKTFUNK_BTN_*/PUNKTFUNK_AXIS_* wire constants).
- Protocol: punktfunk/1 — control-plane magic LMN1 → PKF1, nonce salt lmn1 → pkf1.
WIRE BREAK: clients must be rebuilt from this revision.
- Env knobs: PUNKTFUNK_VIDEO_SOURCE / PUNKTFUNK_COMPOSITOR / PUNKTFUNK_ZEROCOPY / ….
- Host config dir: ~/.config/punktfunk (the box's dir was migrated in place — the
persistent identity is unchanged, pinned fingerprints stay valid).
- Swift package: PunktfunkKit + PunktfunkCore.xcframework + PunktfunkConnection
(Sources/PunktfunkClient app + tests renamed with it); build-xcframework.sh updated.
- scripts/: 60-punktfunk.rules, punktfunk-host.service; OpenAPI doc regenerated.
Also: scripts/headless/run-headless-kde.sh — full headless Plasma bringup. Root cause of
"desktop but no apps/settings" over the stream: plasmashell launched without
XDG_MENU_PREFIX=plasma-, so the launcher resolved a nonexistent applications.menu and
rendered an empty menu. The script sets the complete KDE session env (menu prefix,
KDE_FULL_SESSION, session version) and rebuilds ksycoca before starting plasmashell.
Gate: 97/97 tests, clippy -D warnings (both feature sets), fmt, C-ABI harness PASS,
zero lumen references left outside .git.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Mutter backend validated on headless GNOME Shell 50.1: RecordVirtual virtual monitor at
the client's mode, 360/360 frames @1080p60, p50 1.28 ms, tiled-dmabuf zero-copy. All
three VirtualDisplay backends (KWin/gamescope/Mutter) are now proven live.
- gamescope 3.16.22 system install validated (720/720 @720p120 via /usr/local/bin, the
PATH override is retired from the docs).
- Gamepads validated live over lumen/1: client datagrams -> uinput "Microsoft X-Box 360
pad" created (event9/js1), destroyed with the session (udev rule + input group active).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The clients/apple scaffold is now a working macOS client, validated live against this
repo's host across the LAN: gamescope virtual output → NVENC HEVC → lumen/1 (GF(2¹⁶) FEC +
AES-GCM over UDP, QUIC control) → VideoToolbox → AVSampleBufferDisplayLayer at 720p60,
mouse/keyboard flowing back as QUIC datagrams into the host's gamescope EIS injector
(~3.7k events injected in one session).
LumenKit:
- LumenConnection: the predicted cbindgen compile fixes (C17 header spells the typedefs as
integers while the enum constants import as a distinct Swift type — bridge by rawValue);
close() is now safe from any thread (a close flag + pumpLock held across the blocking
poll enforce the C contract "never close with a next_au in flight"; flag prevents
lock-starvation by back-to-back polls).
- StreamView: per-pump cancellation token (reconnects can't double-pump), flush + re-gate
on the next in-band parameter sets when the layer fails, no stale enqueue after restart.
- InputCapture: fractional-delta accumulation (sub-pixel motion isn't truncated away),
pressed-state tracking with release-all on focus loss and stop() (nothing sticks down
host-side), global-singleton ownership guard (GC has one handler slot per process),
X1/X2 buttons, horizontal scroll, full keypad/CapsLock/ISO-102nd/PrintScreen/Menu VKs.
- LumenClient app shell (swift run LumenClient): connect form, fps/Mb-s HUD,
LUMEN_AUTOCONNECT/LUMEN_MODE for scripted first-light runs.
- Tests: Annex-B byte-level units; real-codec round trip (VTCompressionSession-encoded
HEVC rebuilt as the host's wire shape → AnnexB → VTDecompressionSession → pixels);
test-loopback.sh (Swift client vs a real local m3-host over loopback — the Swift twin of
c_abi_connection_roundtrip); RemoteFirstLightTests (full pipeline over the LAN).
Host/build fixes that fell out:
- The workspace builds on non-Linux again: gamestream audio (opus) and sendmmsg batching
are now platform-gated with stubs/fallback, per the crate's "compiles everywhere" rule.
- Horizontal scroll was inverted end-to-end: the injectors negated BOTH axes onto the
ei/wl axes, but GameStream's horizontal convention is positive = right
(moonlight-qt/Sunshine pass it through unnegated) — only vertical flips now. This also
un-inverts real Moonlight clients.
- AnnexB drops all zeros preceding a start code (trailing_zero_8bits padding), ffmpeg's
policy, instead of leaking them into the preceding NAL.
- build-xcframework.sh: deployment targets pinned to the package floor + an otool guard —
cargo does not fingerprint MACOSX_DEPLOYMENT_TARGET, so warm caches can silently ship
too-new minos objects.
Adversarially reviewed (5-dimension multi-agent pass, every finding refutation-verified):
14 confirmed findings, all fixed above; the send-while-polling core-contract gap flagged
here is closed by the lumen/1 session-planes work (&self pulls + per-plane borrow slots).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
m3-host is now a real host, not a one-shot demo. Everything validated live on this box
(two back-to-back sessions, pinned + TOFU, ~200 audio pkts/s, p50 0.84 ms at 720p60).
lumen-core:
- quic.rs: QUIC-datagram side planes demuxed by first byte — Opus audio 0xC9
([magic][u32 seq][u64 pts_ns][opus], host→client) and rumble 0xCA ([magic][pad][low][high]).
- Trust: endpoint::server_with_identity (persistent PEM identity) and
endpoint::client_pinned — SHA-256 cert-fingerprint pinning with TOFU (observed
fingerprint reported back for persisting). The verifier checks the TLS 1.3
CertificateVerify signature for real (an MITM replaying the host's public cert without
its key is rejected; cert pinning alone would not prove key possession).
- client.rs: NativeClient gains pin + host_fingerprint, audio/rumble receivers
(next_audio / next_rumble); pull methods take &self so the C ABI's per-plane threads
never alias a &mut (per-plane mutexed borrow slots in abi.rs).
- abi.rs: lumen_connect(pin_sha256, observed_sha256_out) + lumen_connection_next_audio /
next_rumble. input.rs: documented gamepad wire contract (GameStream buttonFlags bits,
XInput axis conventions, +y = up) — exported as LUMEN_BTN_*/LUMEN_AXIS_* (bare BTN_*
collides with <linux/input-event-codes.h> at different values).
lumen-host (m3):
- Persistent accept loop: sessions back to back on one endpoint (--max-sessions, 0 =
forever); per-session failures log and the loop keeps serving; 10 s handshake deadline
so a silent client can't wedge the sequential accept queue; teardown on every exit path
(stop flag → conn.close → join audio+input threads).
- Audio plane: desktop PipeWire capture → Opus 48 kHz stereo 5 ms CBR → datagrams; ONE
capturer reused across sessions via an AudioCapSlot (PipeWire streams have no cheap
teardown — per-session opens would leak a thread + core connection + live node each).
- Gamepad routing: incremental GamepadButton/GamepadAxis datagrams accumulate into
per-pad state feeding the uinput xpad manager; force feedback returns as rumble
datagrams, with current state re-sent every 500 ms (idempotent-state healing for the
lossy channel). QUIC endpoint serves the persistent ~/.config/lumen identity and logs
the pinnable fingerprint.
lumen-client-rs: --pin (malformed values abort — never silently downgrade to TOFU),
TOFU fingerprint logging, audio/rumble datagram counters, gamepad events in --input-test.
clients/apple: scaffold synced — pinSHA256/hostFingerprint (wrong-size pin throws,
fail-closed), nextAudio/nextRumble, gamepad event constructors; README handoff updated
(persistent listener, audio decode notes, trust UX).
Adversarially reviewed (5-dimension multi-agent pass over the diff, 2-skeptic
verification): fixed the MITM signature-check gap, a Y-axis contract inversion, header
macro collisions, ABI aliasing UB, the PipeWire per-session leak, the missing handshake
deadline, fail-open pin parsing, and teardown-on-error paths.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The shared-core architecture pays off: platform clients now link ONE Rust library that
does the entire lumen/1 protocol, and only add decode/present/input on top.
lumen-core:
- client.rs (quic feature): NativeClient — QUIC handshake + UDP data plane + input
datagrams on internal threads; embedder surface = connect / next_frame / send_input.
- abi.rs: lumen_connect / lumen_connection_next_au (borrow-until-next-call, matching
lumen_client_poll_frame semantics) / lumen_connection_send_input / lumen_connection_mode /
lumen_connection_close. Guarded in the generated header by LUMEN_FEATURE_QUIC (cbindgen
[defines] mapping), so the checked-in header is stable across feature sets.
- error.rs: append-only LumenStatus additions Timeout (-9) and Closed (-10).
- TESTED end-to-end through the C ABI: in-process lumen/1 host, lumen_connect pulls 25
byte-verified frames, sends input, closes (m3.rs::c_abi_connection_roundtrip).
Apple client (clients/apple — SCAFFOLD, written on Linux, first Xcode build pending):
- scripts/build-xcframework.sh: cargo per Apple target → universal staticlib + header
(LUMEN_FEATURE_QUIC pre-defined) + modulemap → LumenCore.xcframework.
- Package.swift (LumenKit) + Swift sources: LumenConnection (ABI wrapper), AnnexB
(in-band VPS/SPS/PPS → CMVideoFormatDescription, Annex-B → AVCC CMSampleBuffers with
DisplayImmediately), StreamView (SwiftUI over AVSampleBufferDisplayLayer — stage-1
presenter that hardware-decodes compressed HEVC itself), InputCapture (GCMouse raw
deltas + GCKeyboard HID→VK).
- README.md is the full handoff for the next (Mac-side) agent: build steps, ABI contract,
first-light test recipe against the Linux host, stage-2 (VT+Metal pacing) plan, and the
known host-side gaps (single-session m3-host, no lumen/1 audio yet, gamepad kinds not
yet routed in m3's injector, seed-stage trust).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Resolutions: serve() keeps main's AppState::new() with our persisted-pairing load folded
into it; main.rs keeps both the m3 and mgmt modules; mgmt's test LaunchSessions gain the
new appid field; Cargo.lock re-resolved. Full gate green (92 tests, clippy, fmt).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The native protocol now does the real thing, end to end:
- Hello carries the client's requested mode; the host creates a NATIVE virtual output at
exactly that size/refresh (same vdisplay backends as the GameStream path) and streams
NVENC HEVC through the M1 Session (GF(2^16) Leopard FEC + AES-GCM, QUIC-negotiated).
- Input rides QUIC DATAGRAMS — encrypted, congestion-managed, no ENet retransmission
spikes — decoded into lumen_core InputEvents and fed to the session's input injector.
- Frames are stamped with the capture wall clock; the reference client computes per-frame
capture→reassembled latency percentiles and writes a playable .h265.
- m3-host gains --source synthetic|virtual + --seconds; the client gains --mode WxHxFPS,
--out, --input-test (scripted mouse/keyboard datagrams).
VALIDATED live (gamescope session, xev nested): client requested 1280x720@120 → host
created gamescope at that mode → 1680/1680 frames over 14s, zero loss, valid HEVC;
pipeline latency p50 0.83ms / p95 1.2ms / p99 1.3ms (capture→encode→FEC→crypto→UDP→
reassembled, same-host clock); 176 input datagrams sent → injector (GamescopeEi) → 164
X events observed inside the nested session.
Known follow-on: slice-level sub-frame pipelining needs the NVENC SDK directly (libavcodec
emits whole AUs only) — the next big latency lever.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The first end-to-end run of lumen's own protocol, past the GameStream compatibility layer.
- lumen-core/src/quic.rs (behind the `quic` feature): the lumen/1 handshake — Hello/Welcome/
Start as length-prefixed LE binary on one QUIC bi-stream. Welcome carries the COMPLETE
data-plane Config: mode, FEC scheme incl. GF(2^16) Leopard (inexpressible in GameStream),
shard sizing, AES-GCM key + per-direction salt, data UDP port. Plus quinn endpoint helpers
(self-signed server; accepts-any client — pinning lands with the trust model) and framed
async IO. Round-trip unit-tested.
- lumen-host m3-host: serves one lumen/1 session — QUIC handshake, then a NATIVE thread
(no async on the frame path — design invariant) streams deterministic 64KB test frames
through the hardened M1 Session over UdpTransport.
- lumen-client-rs: from scaffold to working reference client — connects, negotiates, brings
up the client Session over UDP, reassembles + FEC-recovers + byte-verifies every frame.
VALIDATED END-TO-END on localhost: 300/300 frames verified, 0 mismatches, through
QUIC-negotiated GF(2^16) FEC + AES-GCM over real UDP sockets. M4 (decode+present) builds on
this exact client skeleton.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- gamestream/apps.rs: an app catalog (loaded from ~/.config/lumen/apps.json, with defaults:
Desktop + gamescope entries when gamescope/steam/vkcube are installed). /applist renders
it; /launch?appid=N selects the entry; RTSP PLAY resolves it and the stream honors the
app's compositor + nested command — so a Moonlight client picks "Steam" and gets a
gamescope session at its native resolution, or "Desktop" for the KWin/GNOME desktop.
- Persistent pairing: the paired-client cert allow-list now survives restarts
(~/.config/lumen/paired.json), saved on each successful pairing, loaded at boot.
- Quit semantics: /cancel now actually stops the media threads (streaming/audio flags),
tearing down the per-session virtual output / gamescope process via the capturer's RAII.
- scripts/lumen-host.service (systemd user unit) + scripts/host.env.example (config file
consumed by it) — the host runs as a managed service instead of an SSH shell.
Smoke-tested: serve boots, /applist serves the catalog (Desktop + vkcube gamescope entry
auto-detected on this box). GNOME backend validation still pending gnome-shell install;
wlroots vdisplay backend deliberately deferred (not in the priority compositor trio).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The missing zero-copy path is closed. NVIDIA's EGL won't sample LINEAR and the CUDA driver
rejects raw dmabuf fds — but Vulkan imports dmabufs (VK_EXT_external_memory_dma_buf) and
exports OPAQUE_FD memory that CUDA officially imports. zerocopy/vulkan.rs (ash):
dmabuf fd → VkBuffer (import cached per fd) → vkCmdCopyBuffer (GPU) →
exportable VkBuffer → vkGetMemoryFdKHR(OPAQUE_FD) → cuImportExternalMemory → CUdeviceptr
The exportable buffer + CUDA mapping are per-resolution; per frame it's one GPU buffer copy
(fence-waited) + one pitched CUDA copy into the encoder's pool. No CPU touches pixels.
EglImporter::import_linear now routes through the bridge (lazy init; any failure still falls
back to the CPU mmap path). cuda::ExternalDmabuf gained import_owned_fd for the
Vulkan-exported fd.
Validated live: gamescope 720p120 → "Vulkan→CUDA exportable staging buffer ready
size=3686400" (exactly 1280*720*4), full-rate 122.7 fps, decoded frame pixel-correct
(vkcube). KWin's tiled EGL path regression-tested intact. NV12 negotiation dropped — moot
now that BGRx is fully zero-copy.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
stream_body no longer sends: each frame's packet batch goes over a depth-2 bounded queue to a
dedicated send thread, so a send spike can never stall capture/encode (a full queue drops the
NEWEST batch — FEC/RFI covers the client — rather than ever blocking). The sender ships packets
with sendmmsg (≤64/syscall: ~375 syscalls/s instead of ~24k at 5K@240) in 16-packet chunks
paced across ~3/4 of the frame interval — microburst shaping for real links without per-packet
sleep jitter. Client-gone detection moved to the sender (clears `running`); the LUMEN_VIDEO_DROP
FEC test knob moved with the send path. Loopback-tested: batches arrive complete and
byte-identical through the paced sendmmsg path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Full controller path for the SteamOS-like session, mirroring Sunshine byte-for-byte
(wire formats verified against moonlight-common-c + Sunshine source; ioctl numbers and
struct layouts verified by compiling against this box's <linux/uinput.h>, locked in with
const asserts):
- gamestream/gamepad.rs: decode MULTI_CONTROLLER (magic 0x0C, mixed BE-size/LE-body) incl.
the Sunshine buttonFlags2 extension (paddles/touchpad/Misc — our appversion already
advertises Sunshine, so clients send it) and CONTROLLER_ARRIVAL (0x55000004); build the
0x010B rumble plaintext (with the mandatory 4-byte filler). Unit-tested.
- inject/gamepad.rs: VirtualPad clones the kernel xpad identity ("Microsoft X-Box 360 pad",
045e:028e, exact button/axis codes + absinfo) so SDL/Steam/Proton match their built-in
mapping with zero config. GamepadManager creates/destroys pads from activeGamepadMask
(hotplug), emits button transitions + axes (+Y-up → evdev +Y-down negation, D-pad as
HAT0X/Y) per frame. Rumble: non-blocking FF pump answers UI_BEGIN/END_FF_UPLOAD/ERASE
(games block in EVIOCSFF until answered), tracks effects with replay expiry + FF_GAIN,
mixes to (low, high) motor levels, dedups.
- control.rs: channel_limit 8 → 0x30 — Moonlight sends gamepad input on ENet channel
0x10+n, so the old limit silently discarded ALL controller input. Gamepad events route to
the manager; rumble is sealed with the client's detected GCM scheme direction-flipped
(V2 marker 'H?', own seq counter) and sent on the control peer every service tick.
- scripts/60-lumen.rules: udev rule (Sunshine-style) granting the input group /dev/uinput.
Live validation needs the udev rule installed (root-only /dev/uinput on this box) + a
Moonlight client with a controller; everything else is gated and unit/static-checked.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Third compositor on the VirtualDisplay seam, via Mutter's direct D-Bus APIs (the
gnome-remote-desktop headless model, no portal grant): RemoteDesktop.CreateSession →
ScreenCast.CreateSession({remote-desktop-session-id}) → Session.RecordVirtual (creates a
virtual monitor) → Start → PipeWireStreamAdded(node_id). A keepalive thread owns the zbus
connection (sessions die with it — RAII teardown); select with LUMEN_COMPOSITOR=mutter or
XDG_CURRENT_DESKTOP=GNOME.
Mutter sizes its virtual monitor FROM the PipeWire format negotiation, so VirtualOutput
gains preferred_mode (w, h, refresh_hz), threaded into the consumer's format pods as the
default size/framerate. KWin/gamescope set it too (their outputs are already exact-size;
the preference just confirms it) — both regression-tested intact.
Compile/clippy/test clean; live validation needs gnome-shell installed (then
`gnome-shell --headless` + m0 --source kwin-virtual with LUMEN_COMPOSITOR=mutter).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
gamescope only offers LINEAR dmabufs, which the EGL/GL interop path can't handle (NVIDIA's
EGL lists no LINEAR modifier for sampling). Attempt a direct CUDA external-memory import
(cuImportExternalMemory OPAQUE_FD, cached per buffer fd, one DtoD copy per frame into the
pooled buffer): the FFI + plumbing are in place, and LINEAR(0) is now advertised alongside
the tiled EGL modifiers (tiled first, so KWin still prefers it — regression-tested).
Empirically the 595 desktop driver rejects raw dmabuf fds as OPAQUE_FD (CUDA_ERROR_UNKNOWN),
matching the documented limitation — true LINEAR GPU import needs a Vulkan interop bridge
(import dmabuf via VK_EXT_external_memory_dma_buf, GPU-copy into an exportable allocation,
hand that to CUDA), noted as future work. So the importer now degrades instead of dying:
on GPU-import failure it logs once, disables itself, and falls through to the CPU mmap path.
Validated: gamescope + LUMEN_ZEROCOPY=1 runs full-rate (122.9 fps @720p120, valid HEVC) via
the fallback; KWin keeps real zero-copy.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
gamescope runs its own EIS server and exports the socket to its children as LIBEI_SOCKET —
no portal involved. The gamescope backend now launches the nested app through a tiny shell
wrapper that relays that value to /tmp/lumen-gamescope-ei; the libei injector gains an
EiSource enum (Portal | SocketPathFile) and connects a UnixStream directly to gamescope's
socket (polling until the app has started), then runs the identical reis sender flow.
Backend::GamescopeEi is auto-selected when LUMEN_COMPOSITOR=gamescope
(LUMEN_INPUT_BACKEND=gamescope overrides).
Validated end-to-end: input-test against a headless gamescope running xev — 129
MotionNotify/KeyPress/ButtonPress events delivered into the nested X app ("Gamescope
Virtual Input" device bound, sender handshake + emulation working).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Deep investigation (gdb + daemon traces) proved the gamescope capture stall is a gamescope
3.16.20 bug, not ours: it calls pw_loop_iterate() without pw_loop_enter()/leave(), and under
PipeWire 1.6's loop locking its main thread permanently holds the loop mutex — the pw thread
deadlocks, gamescope never acks the daemon's port_set_param(Format), and the link parks in
"negotiating" silently. Stock gst pipewiresrc fails identically. Fixed upstream by gamescope
commit e3ed1ea7 ("pipewire: Fix pipewire loop locking", pipewire#5148); first release 3.16.22.
Ubuntu 26.04 ships 3.16.20 (built ten days before the fix) — patch/upgrade required.
Consumer-side improvements from the investigation (all verified correct vs gamescope's pods,
and needed once the producer is fixed):
- discover the node from gamescope's own "stream available on node ID: N" log line (its
node.name appears on two objects; the advertised id is authoritative); pw-dump fallback
- CPU path accepts mappable dmabufs: Buffers param now offers MemPtr|MemFd|DmaBuf (gamescope
counter-offers exactly DmaBuf when its modifier pod wins, never MemPtr), mmap the fd
ourselves when MAP_BUFFERS didn't (Vulkan-exported dmabufs aren't flagged mappable), and
treat chunk.size==0 as the computed span
- warn_once on every silent frame-drop path in the process callback
- node.dont-reconnect on our capture streams: an orphaned stream re-targeted by wireplumber
onto a fresh node wedges it — and a stuck link head-blocks the daemon's shared work queue,
stalling ALL new link negotiation system-wide (this poisoned whole test sessions)
- LUMEN_GAMESCOPE_NODE (attach to an existing gamescope) + LUMEN_PW_FIXED_POD (negotiation
bisection) debug knobs
KWin path regression-tested (zero-copy intact). gamescope end-to-end validation pending the
patched gamescope build.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Five confirmed findings from a 46-agent review panel:
- Empty --mgmt-token no longer satisfies the non-loopback token gate
(critical: 'Bearer ' with an empty token authenticated; parse_serve now
bails on blank tokens and mgmt::run treats blank as none)
- axum's built-in body rejections (400/415/422) now wear the documented
ApiError envelope via an ApiJson extractor, and the spec documents them
- GET /health carries security([{}]) in the spec, matching the server's
auth exemption
- unpairClient's description no longer claims revocation the TLS layer
doesn't enforce yet (gamestream/tls.rs accepts any cert — known gap)
- CLAUDE.md/README.md no longer reference the deleted web.rs
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A versioned control-plane REST API (/api/v1) on its own port (default
127.0.0.1:47990) serving host info, runtime status, paired-client
management, the pairing PIN flow, and session control (stop / force-IDR).
The OpenAPI 3.1 document is generated from the handlers by utoipa, served
live at /api/v1/openapi.json (+ Scalar docs at /api/docs), printable via
`lumen-host openapi`, and checked in at docs/api/openapi.json for client
codegen — a test fails if it drifts, mirroring the cbindgen header rule.
Auth: optional bearer token (--mgmt-token / LUMEN_MGMT_TOKEN), enforced on
everything but /health, and mandatory for non-loopback binds. PinGate
gains a waiter count so the API can report pin_pending; logs moved to
stderr so stdout stays machine-readable. Supersedes the web.rs stub.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Third compositor on the VirtualDisplay seam. gamescope's model differs from KWin/Mutter: it's
not a runtime protocol but a micro-compositor we spawn — `gamescope --backend headless -W -H -r
-- <app>` — which composites at the client's size AND refresh natively (so no separate
refresh step), runs the app nested, and exports a built-in PipeWire node named "gamescope".
The backend spawns it, discovers that node via pw-dump, and returns a VirtualOutput whose
keepalive owns the process (drop = kill = teardown). App via LUMEN_GAMESCOPE_APP. Select with
LUMEN_COMPOSITOR=gamescope; m0's virtual source now honors LUMEN_COMPOSITOR so any backend is
testable without a client. Input (gamescope's libei/EIS socket) is a follow-up.
Builds/clippy/fmt clean. Needs gamescope installed to validate; headless capture on the
proprietary NVIDIA driver is plausible-by-architecture but unproven — validate empirically.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
KWin creates virtual outputs at a hardcoded 60 Hz and zkde stream_virtual_output has no
refresh argument, so the *source* composited at 60 Hz even when the client asked for 120/240
(confirmed live: stream paced a stable 240 fps but only ~60 unique frames/s). KWin 6.6+ allows
custom modes on virtual outputs, so after creating the output we install + select a mode at the
client's refresh, before capture connects PipeWire. First cut shells out to kscreen-doctor
(output is "Virtual-<name>"); the in-process kde_output_management_v2 client is a follow-up.
Best-effort — failure leaves the source at 60 Hz (stream still works). Verified the mode is
applied (Virtual-lumen -> 1280x720@120). Empirically de-risked that this headless QEMU VM's
software vsync accepts >60 Hz.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GB203 has two NVENC engines. A single HEVC p1 session tops out ~1 Gpix/s, so 5120x1440@240
(1.77 Gpix/s) is encoder-bound on one engine; split-frame encode runs it across both (~1.8x,
latency-neutral, output is standard HEVC the client decodes normally). NVENC's AUTO split
won't engage below ~2112px height, so force split_encode_mode=2 when the pixel rate exceeds
~1 Gpix/s (HEVC/AV1 only — not H.264). Below that (e.g. 5K@120) stay single-engine to avoid
the ~2% BD-rate cost. Override with LUMEN_SPLIT_ENCODE. Verified: engages at 240, not at 120.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The zero-copy import did real per-frame GPU churn that capped high-fps throughput: a fresh
~29MB cuMemAllocPitch + cuMemFree, a cuGraphicsGLRegisterImage/unregister, and a map of the
*same* persistent blit texture — every frame. Two fixes:
- BufferPool: a recycled free-list of pitched device buffers per resolution. DeviceBuffer
returns its allocation to the pool on drop (after the encoder synchronized) instead of
freeing — kills the per-frame 29MB alloc/free that took the device allocator lock and
serialized against the GPU.
- RegisteredTexture: register the (reused) GL_RGBA8 blit destination with CUDA ONCE when the
GlBlit is built; each frame only maps → copies the array → unmaps, instead of
registering/unregistering every frame.
This is the "zero-copy should be overhead-free" cleanup. Verified the import still produces
correct frames; the remaining per-frame cuCtxSynchronize pair (shared-context coupling) is
the next step (CUDA stream + events). lumen-host builds, clippy/fmt/tests clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
At 5120x1440 the stream froze on a ~2s cadence. Two compounding causes (confirmed by a
profiling pass + adversarial review):
1. Periodic IDR every 2s (set_gop(fps*2)). A keyframe at 5K is ~20-40x a P-frame — a
recurring multi-millisecond encode+packetize+send spike. Fix: infinite GOP (gop_size=-1),
one IDR at stream start, P-frames only; forced-idr makes a client recovery request (RFI via
request_keyframe) emit an IDR on demand — the Moonlight/Sunshine low-latency model.
2. Two pacing timers summing on the capture/encode thread: a per-packet thread::sleep pacer
(spread a frame's packets across a whole frame interval) PLUS a backstop sleep on top, so
every frame cost 1-2x the interval and the big IDR blew through it (the 2->120 oscillation).
Fix: delete both; send at line rate and drive cadence from a single absolute deadline.
(Proper microburst pacing belongs on a dedicated send thread — a follow-up.)
Also: honor the client's fps (pacing clamp 60->240) and add an env-gated (LUMEN_PERF)
per-stage timing log (enc/pkt/send µs + unique-vs-reencoded frames + max packet burst) for
diagnosing the remaining throughput ceiling. Verified live: freeze gone at 5120x1440.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ubuntu 26.04 ships FFmpeg 8.0 (libavcodec 62); bump ffmpeg-next 7.1 -> 8.1 to bind it
as the intended pairing. No source changes needed — the encode API surface we use
(avcodec_send_frame, hwframe contexts, AV_PIX_FMT_CUDA, av_log) is stable across 7->8.
Workspace builds + all tests green; clippy/fmt clean. Refresh the 7.x doc references.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Honor the client's requested resolution by rendering a compositor virtual output at
exactly that size — native, headless, no scaling. There is no cross-compositor Wayland
protocol for this, so it's a per-compositor backend behind the (previously stubbed)
VirtualDisplay trait.
- vdisplay.rs: VirtualDisplay::create(mode) now returns a live VirtualOutput
{ node_id, remote_fd: Option<OwnedFd>, keepalive } with RAII teardown (drop releases
the output) instead of an inert OutputHandle + explicit destroy. Add compositor
detect() (LUMEN_COMPOSITOR / XDG_CURRENT_DESKTOP).
- vdisplay/kwin.rs: the KWin backend — the zkde_screencast_unstable_v1 stream_virtual_output
client (vendored protocol XML + wayland-scanner codegen). Creates a WxH output, returns
its PipeWire node (default daemon, remote_fd=None); a keepalive thread holds the Wayland
connection until dropped. (Moved here from capture/kwin.rs — it's a vdisplay backend, not
capture.)
- capture: generalize the PipeWire consumer to Option<OwnedFd> (portal remote vs. default
daemon) and add capture_virtual_output(vout), compositor-agnostic, owning the keepalive.
- gamestream/stream.rs: LUMEN_VIDEO_SOURCE=virtual creates a virtual display sized to the
client's cfg and captures it (self-contained, not pooled — a reconnect at a new
resolution gets a fresh output).
- m0: --source kwin-virtual goes through the trait.
Verified end-to-end against the running headless KWin: the request reaches the compositor
and is handled cleanly. Native creation needs a backend implementing createVirtualOutput —
the DRM backend, or the VirtualBackend since KWin 6.5.6; on this box's --virtual 6.4.5 it
returns "Could not find output" (expected; validates after the KWin upgrade). wlroots/Mutter
backends are the next ones to land on the same seam.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Clients pick the resolution via mode=WxHxFPS / RTSP clientViewportWd-Ht, so the
host must bound attacker/typo-controlled dimensions before allocating buffers or
opening NVENC. Add encode::validate_dimensions: reject zero, odd, and over-limit
modes (H.264 ≤ 4096px/side; HEVC/AV1 ≤ 8192) with a clear message instead of a
buffer-math overflow or an opaque NVENC open failure. Gate both the stream path
(before any allocation) and open_video (also covers m0). Unit-tested.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The PipeWire dmabuf now reaches NVENC with no CPU touch. Verified live against
headless KWin: a tiled BGRx dmabuf is imported and encoded to a pixel-correct
H.265 stream (decoded frame matches the captured desktop — no tiling artifacts,
no colour swap). The CPU-copy path stays the default and the runtime fallback.
Capture side (zerocopy::egl): desktop NVIDIA can't register a dmabuf EGLImage
with CUDA directly (cuGraphicsEGLRegisterImage is Tegra-only; cuGraphicsGLRegisterImage
rejects EGLImage-backed textures), so we follow OBS/Sunshine — bind the EGLImage
to a GL texture, render it through a fullscreen-triangle shader into an immutable
GL_RGBA8 texture (de-tiling + .bgra swizzle to the BGRx the encoder wants), then
register that texture with CUDA and copy it device-to-device into an owned buffer
so the dmabuf returns to the compositor immediately.
Encode side (encode/linux::submit_cuda): take a *pooled* CUDA surface via
av_hwframe_get_buffer and device→device-copy our imported buffer into it, instead
of wrapping our own pointer in a bare AVFrame. A bare frame is rejected with
EINVAL (NVENC ignores frames with null buf[0]; the encode path's av_frame_ref
needs a refcounted buffer), and a fresh device pointer every frame would thrash
NVENC's bounded resource-registration cache — the pool recycles a small set.
Also: gate FFmpeg AV_LOG_DEBUG behind LUMEN_FFMPEG_DEBUG for diagnosing
hw-frame rejects, and refresh the now-accurate module docs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wire the capture side of zero-copy (LUMEN_ZEROCOPY=1):
- EGL importer now opens the headless EGLDisplay on the NVIDIA EGL device
(EGL_PLATFORM_DEVICE_EXT) and queries its importable DRM modifiers
(eglQueryDmaBufModifiersEXT).
- The PipeWire stream advertises a BGRx dmabuf format with those modifiers as a
mandatory enum Choice + a dmabuf-only Buffers param; the compositor fixates an
importable tiled modifier. param_changed reads the negotiated modifier; the
process callback imports the dmabuf (eglCreateImage with explicit LO/HI
modifier) and would copy it into a CUDA buffer for the encoder.
Validated against headless KWin (Plasma 6.4): negotiation succeeds (13 NVIDIA
modifiers advertised, KWin fixates one, stream reaches Streaming with a real
tiled dmabuf) and `eglCreateImage` succeeds. The remaining blocker is
`cuGraphicsEGLRegisterImage` returning CUDA_ERROR_INVALID_VALUE on the
dmabuf-imported EGLImage — the likely fix is to bind the EGLImage to a GL
texture (glEGLImageTargetTexture2DOES) and register that via
cuGraphicsGLRegisterImage (OBS/Sunshine's path), which needs a GL context.
The CPU-copy path stays the default and is unaffected (regression-checked: real
KWin capture → HEVC). LUMEN_ZEROCOPY is opt-in/experimental until the CUDA
registration lands.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Scaffolding for dmabuf zero-copy (plan §9), opt-in via LUMEN_ZEROCOPY:
- src/zerocopy/{cuda,egl}.rs: hand-rolled CUDA Driver-API FFI (no Rust crate
exposes the EGL-interop calls / CUeglFrame) with a shared process-wide
CUcontext + pitched device buffers; an EGL importer (GBM platform on the
NVIDIA render node) that turns a dmabuf into an EGLImage, registers it with
CUDA, and copies it device-to-device into an owned buffer. `zerocopy-probe`
subcommand validates the FFI/linking/GPU access — confirmed on the box
(driver 595, EGL_EXT_image_dma_buf_import + modifiers).
- CapturedFrame gains a FramePayload enum (Cpu(Vec<u8>) | Cuda(DeviceBuffer));
the encoder branches: CPU keeps the expand+upload path, CUDA wraps the device
buffer in an AV_PIX_FMT_CUDA frame fed straight to hevc_nvenc (sharing our
CUcontext via a hand-declared AVCUDADeviceContext, since ffmpeg-sys doesn't
bind hwcontext_cuda.h). open_video/the encoder take a `cuda` flag derived from
the first frame's payload.
The capture-side dmabuf negotiation (which produces the Cuda frames) is the
next step; the CPU path is unchanged and remains the default + fallback. Builds
clean, clippy clean, tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The M2 teardown work added an `active` gate to the PipeWire capture callback
(idle by default so reconnects stay cheap, with the stream path calling
set_active(true) on PLAY). The `m0` subcommand was never updated, so its portal
capturer stayed inactive and the callback dropped every frame — `m0 --source
portal` failed with "no PipeWire frame within 10s" on every compositor. Call
set_active(true) before the capture loop.
Validated on headless KWin (Plasma 6.4) via the RemoteDesktop-anchored ScreenCast
session: real desktop frames flow (shm BGRx 1920x1080) and encode to valid H.265.
(Also folds in a rustfmt reflow of the input-test log line.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On RemoteDesktop-capable desktops (KWin, GNOME), select the ScreenCast source on
a session created via the RemoteDesktop portal and start it through RemoteDesktop,
so a single grant — pre-authorized headlessly via the `kde-authorized` permission,
exactly like the libei input path — also covers screen capture. Standalone
ScreenCast has no such bypass and would raise an un-clickable dialog on a headless
box. wlroots/Sway has no RemoteDesktop portal, so it keeps the plain ScreenCast
session; the choice keys off inject::default_backend(). The PipeWire consumer is
unchanged — the anchored session yields the same fd + node id.
Validated on headless KWin (Plasma 6.4): the portal grants the session with no
dialog and PipeWire negotiates the format (1920x1080 BGRx, Streaming). Frame
delivery on KWin still pends dmabuf import — KWin hands GPU dmabuf buffers and the
M0 consumer is CPU-copy/shm only (plan §9, zero-copy) — so it's the next step;
the CPU-copy path remains correct on wlroots.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a second input-injection backend that works on compositors implementing
the org.freedesktop.portal.RemoteDesktop interface (KWin, GNOME/Mutter), where
the wlroots virtual-input protocols are absent. Uses ashpd 0.13 to open a
RemoteDesktop session + EIS fd and reis 0.6.1 to drive it as an EI sender:
bind pointer/keyboard/scroll/button capabilities and, per device,
start_emulating → emit → frame. Runs on a dedicated thread with its own tokio
runtime (the portal session + EIS connection must stay alive and the event
stream must be polled continuously); open() returns immediately so a slow or
denied portal can never freeze the ENet control thread, with events enqueued
over an unbounded channel until devices resume.
Backend now auto-selects per session (inject::default_backend): wlr on Sway,
libei on KDE/GNOME; LUMEN_INPUT_BACKEND overrides. Refactor inject.rs into the
inject/{wlr,libei}.rs layout matching the capture/encode convention. Keyboard
codes are evdev (the same space our VK→evdev table produces) and the compositor
supplies the keymap, so no keymap upload and no modifier serialization — pressing
the modifier keys Moonlight sends is enough.
Add a `lumen-host input-test` subcommand that injects a scripted mouse+keyboard
pattern through the session backend, so input injection can be validated without
a Moonlight client.
Live-validated on headless KWin (Plasma 6.4): mouse motion, left click, and the
'A' key inject correctly and are delivered to the focused client.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Disconnect/reconnect now works reliably. Previously each stream spawned its own
portal+PipeWire (and PipeWire audio) capture threads and never stopped them, so a
reconnect opened a SECOND screencast session that conflicted with the leaked
first one ("no PipeWire frame within 10s" → black screen on reconnect).
- The screen capturer and audio capturer are now persistent, held in AppState and
reused across streams (created on the first stream). One screencast session for
the host's lifetime → no conflict, and instant reconnect (no re-handshake).
Verified live: 3 stream cycles, 1 create + 2 "reusing capturer", clean every time.
- Capturer::set_active gates the (5K, ~1.3 GB/s) de-pad copy to active streams, so
the persistent video capturer is nearly free while idle between streams.
- AudioCapturer::drain discards buffered chunks on reuse so the client never hears
stale audio captured while idle.
- stream.rs / gamestream/audio.rs split into a borrow-the-capturer wrapper + the
encode/send body, so the capturer is always returned to its slot on exit.
This holds whether the client reconnects via /resume (Moonlight's "running →
play/continue") or a fresh /launch — both re-run RTSP PLAY → a new stream cycle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Graceful FEC behavior on a lossy link: at a realistic 2% packet loss the stream
is now steady 0% (was spiking 40-60%). Verified live.
- IDR/RFI handling: the control thread recognizes the client's recovery requests
(0x0301 invalidate-reference-frames, 0x0302 request-IDR, 0x0305) and sets a
shared force_idr flag; the video thread forces an NVENC keyframe on the next
frame (Encoder::request_keyframe → input frame pict_type = I). Without this, a
frame that exceeds the FEC budget broke the reference chain until the next GOP
IDR (~2s), cascading to most of the stream being undecodable.
- Min-parity floor: honor the client's x-nv-vqos[0].fec.minRequiredFecPackets
(it asks for 2). Small P-frames previously got m=ceil(k*20/100)=1 parity — a
single loss broke them; flooring m>=2 (capped so k+m<=255, wire pct recomputed)
protects them. This is what turned the 2% spikes into steady 0%.
- Send pacing: spread each frame's packets evenly across the frame interval
instead of blasting them at line rate (a real link drops microbursts), matching
Sunshine's rate-controlled sends; sub-500us sleeps skipped (unreliable).
Note: sustained ~8% uniform loss still degrades — that exceeds 20% FEC for
reference-frame video and real Sunshine degrades there too; real networks are
<1% or bursty, which this now handles cleanly.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Moonlight now reconstructs lost video shards from our parity (verified live:
under induced packet loss the picture recovers cleanly instead of failing with
"network connection too bad"; 0% added loss in normal operation).
The decisive finding: Moonlight's nanors uses a CAUCHY generator matrix
(M[j][i] = inv[(m+i)^j], GF(2^8) poly 0x1d), while reed-solomon-erasure is
Vandermonde — so its parity was NOT Moonlight-decodable, despite the old
gf8.rs comment claiming equivalence.
lumen-core:
- Swap the GF(2^8) backend from reed-solomon-erasure to a vendored fec-rs
(vendor/fec-rs, BSD-2), which builds the byte-identical Cauchy matrix. Pure
Rust, no FFI — keeps the "one core" hot path. This makes both lumen's own
protocol and the GameStream parity nanors-compatible.
- Lock it with a regression test against real nanors vectors
(k=4,m=2 [10,20,30,40] -> parity [136,0]) + an independent matrix-derived
cross-check + an erase/recover round-trip. Existing FEC/loopback tests stay
green, so lumen's own protocol is unaffected.
lumen-host video.rs:
- Generate m = ceil(k*pct/100) parity shards per FEC block via Gf8Coder; stamp
fecInfo with the recomputed wire pct (100*m/k) so the client derives the same
count; cap per-block data to 255*100/(100+pct) so k+m <= 255.
- CRITICAL byte-exactness: RS runs over the whole `blocksize` shard (Moonlight
decodes packetSize+16 bytes from the datagram start and PACKET_RECOVERY_FAILUREs
on a bad reconstructed `flags` byte). So the NV header fields RS must reproduce
(streamPacketIndex/frameIndex/flags/multiFec*) are written into data shards
BEFORE encode, and only the transport fields (RTP header/seq/timestamp +
fecInfo) are stamped AFTER — leaving the flags byte RS-covered. Matches
Sunshine stream.cpp. Unit-tested incl. flags recovery.
- fec_percentage wired from stream.rs (Sunshine default 20, LUMEN_FEC_PCT
override; 0 = data-only). LUMEN_VIDEO_DROP injects loss to test recovery.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A stock Moonlight client now gets video + full input + AUDIO from the
from-scratch GameStream host (verified live end-to-end on a macOS client).
Audio (audio.rs, audio/linux.rs, gamestream/audio.rs):
- Capture the default PipeWire sink's monitor (system output) as interleaved
f32 stereo @ 48kHz via stream.capture.sink, on its own thread.
- Opus-encode 5ms/240-sample stereo frames (RESTRICTED_LOWDELAY, CBR) and send
as GameStream RTP audio: 12-byte BE RTP_PACKET (packetType 97, seq+1/pkt,
timestamp += packetDuration, ssrc 0) on UDP 48000, after learning the client
endpoint from its port-learning ping.
- Encrypt the Opus payload with AES-128-CBC (PKCS7), key = launch rikey, IV =
BE32(rikeyid + seq) in [0..4]. Like the control stream, modern Moonlight
always decrypts audio regardless of the negotiated flags — plaintext makes it
log "Failed to decrypt audio packet" and play silence (diagnosed from the
client log). RTP header stays in the clear. Scheme cross-checked against
Sunshine stream.cpp/crypto.cpp + moonlight AudioStream.c.
- Pace each frame to its 5ms slot (PipeWire delivers ~1024-frame buffers) to
avoid bursts the client's jitter buffer hears as glitches. LUMEN_AUDIO_GAIN
applies optional linear gain for quiet sources.
- DESCRIBE SDP advertises the stereo Opus config (a=fmtp:97 surround-params).
Video (stream.rs): pace at a steady ≤60fps, re-encoding the last captured frame
when the compositor produces none. wlroots only emits on damage, so a static or
slow-updating desktop previously starved the client into a "network too slow"
abort; an unchanged frame costs a near-empty P-frame. Adds a non-blocking
Capturer::try_latest (portal drains to the freshest queued frame).
Misc: serialize pipewire init across the video + audio capture threads
(pwinit.rs, std::sync::Once) to avoid a concurrent pw_init race. Deps: opus,
cbc; libopus-dev in bootstrap-ubuntu.sh.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A stock Moonlight client can now drive the headless Sway desktop: mouse
movement, buttons, scroll, and keyboard all inject through the streamed
session (verified live end-to-end — typing, clicking, window management).
Control stream (gamestream/control.rs):
- Moonlight encrypts the ENet control stream with AES-128-GCM even though we
negotiate no media encryption (it detects our Sunshine `state` and turns it
on). Decrypt per-packet under the /launch `rikey`.
- The exact GCM scheme is auto-detected on the first authenticating packet
(nonce construction × key byte-order × tag position × AAD) since GCM gives no
partial credit. Our client uses the legacy 16-byte nonce (`iv[0]=seq&0xff`)
because we advertise no encryption; the 12-byte SS_ENC_CONTROL_V2 nonce is
also supported. Key/IV/tag layout cross-checked against Sunshine stream.cpp +
crypto.cpp and moonlight-common-c ControlStream.c.
Input decode (gamestream/input.rs):
- Decrypted control messages (`[u16 type][u16 len][NV_INPUT packet]`, type
0x0206) decode into lumen_core::input::InputEvent: relative/abs mouse, buttons,
vert/horiz scroll, keyboard down/up. Struct layout from moonlight Input.h
(size BE, magic LE, body BE; keyCode LE masked to the low-byte VK), dispatch
per Sunshine input.cpp (Gen5+). Unit-tested against real captured bytes.
Injection (inject.rs):
- WlrootsInjector: connects to Sway as a Wayland client and injects via the
wlroots virtual-pointer + virtual-keyboard protocols (uinput is invisible to a
compositor running WLR_LIBINPUT_NO_DEVICES=1). Uploads an evdev/US xkb keymap,
tracks modifier state, and maps Windows VK → Linux evdev (full table).
Deps: aes-gcm, wayland-client, wayland-protocols-{wlr,misc}, xkbcommon (+
libxkbcommon-dev in bootstrap-ubuntu.sh).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wire M0's portal desktop capture into the GameStream video plane: with
LUMEN_VIDEO_SOURCE=portal the stream captures the headless wlroots desktop
(PipeWire RGB) instead of the synthetic pattern, opens NVENC from the first
captured frame's format/size, and streams it. Verified live: a stock Moonlight
client shows the real 5120×1440 desktop at ~42 fps (release build).
- capture.rs: FastSyntheticCapturer (cheap fill pattern, real-time at 5K) so both
sources share the Capturer trait
- stream.rs: source select (portal | synthetic), encoder opened from the first
frame, wall-clock 90 kHz RTP timestamps (correct under a variable capture rate)
Note: the CPU-copy RGB→rgb0 path caps ~42 fps at 5K (single-threaded); dmabuf
zero-copy is the deferred optimization (plan §9).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A stock Moonlight client now decodes H.265 from the lumen host end-to-end
(verified at 5120×1440@120 on RTX 5070 Ti):
- control.rs: ENet control host on UDP 47999 (rusty_enet). Moonlight starts the
control stream before video (STAGE_CONTROL_STREAM_START precedes _VIDEO_), so it
must be up first — this was the blocker behind the earlier "error 35".
- stream.rs: video data plane — on RTSP PLAY, learn the client endpoint from its
ping, NVENC-encode at the negotiated mode, packetize (GameStream RTP/NV/FEC),
send over UDP 47998; stops when the client disconnects.
- rtsp.rs: ANNOUNCE → StreamConfig (resolution/fps/packetSize/bitrate/codec), PLAY
starts the stream, TEARDOWN stops it; PairStatus=1 over the mutual-TLS port.
P1.3 uses a synthetic test pattern + data-shards-only FEC (clean-LAN). Next: real
portal desktop capture, input injection (decode control → uinput), nanors-exact FEC,
encryption, audio.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Prepares the move to the NVIDIA-GPU Ubuntu VM where M0/M2 run (macOS can't drive the
Wayland/GPU stack). The repo carries the context, since Claude Code sessions are
machine-local and don't transfer.
- CLAUDE.md: project state + design invariants + don't-regress security notes. Auto-loads
every session, so a fresh session on the VM continues from here.
- scripts/bootstrap-ubuntu.sh: verifies the (already-installed) NVIDIA/NVENC stack,
installs rustup + PipeWire/portal/wlroots/Sway + DRM/EGL/GBM/VA dev deps; GATES the
FFmpeg -dev headers so apt can't clobber a custom NVENC build; checks nvidia-drm.modeset.
- scripts/headless/: headless-Sway + xdg-desktop-portal-wlr config templates, the
NVIDIA-wlroots env workarounds, run-headless-sway.sh, and a wf-recorder->hevc_nvenc
capture smoke test (proves capture->NVENC with no Rust).
- docs/linux-setup.md: M0 walkthrough + verified gotchas (modeset, headless backend,
vGPU NVENC licensing, dmabuf->NVENC CPU-copy fallback, FFmpeg-dev gate, crate versions).
Ubuntu 24.04 package names/versions verified against the live archive; scripts pass
shellcheck and `bash -n`.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.