docs(release): v0.21.0's notes describe what v0.21.0 actually shipped

The file had accumulated the gamescope-HDR work as it landed — one "New: HDR on
the gamescope path" section plus nine Under-the-hood bullets (the gamescope
patches, Vulkan 10-bit, the per-codec probe, the NVENC HDR leg, the EFC BT.2020
model, the managed-spawn flag check, the CI wiring). All of it landed AFTER
d0889338 was tagged: `PUNKTFUNK_GAMESCOPE_HDR` does not exist at v0.21.0 and
neither does packaging/gamescope/, verified by content rather than by SHA.

The live release body never carried any of it — it was PATCHed from the trimmed
file at c4e80fd4 and not re-synced since — so this only restores the file to the
7030 bytes users actually see on the release page. That matters because
announce.yml's apply_release_notes re-asserts this file over the live body: a
re-announce or a tag re-point would have rewritten a shipped release to
advertise features it does not contain.

The content itself is not lost; it is where it belongs, in v0.22.0.md.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

docs/releases/v0.21.0.md

@@ -12,23 +12,6 @@ A machine you pin from the host's own configuration shows the choice in the cons
 
 For headless and unattended setups, two new commands help you find and verify the right monitor without a screen attached: `punktfunk-host list-monitors` lists what the machine has (names, geometry, which one is pinned), and `punktfunk-host mirror-test` confirms frames are actually flowing from the one you picked.
 
-## New: HDR on the gamescope path (opt-in)
-
-Streaming Steam Gaming Mode from a Linux box has always been SDR — not because the encoder couldn't
-do better, but because gamescope hands its picture to Punktfunk already tone-mapped down to 8-bit.
-That gap is now closed, with a small companion package: install **`punktfunk-gamescope`** (gamescope
-plus a patch that adds the 10-bit BT.2020 PQ formats to its capture output — offered upstream) and
-set `PUNKTFUNK_GAMESCOPE_HDR=1`, and games render in real HDR while the stream carries HDR10 to an
-HDR-capable client.
-
-It sits *beside* your system's gamescope rather than replacing it — your own Gaming Mode is
-untouched — and Punktfunk only uses it for the sessions it starts itself. On Bazzite it rides in
-the Punktfunk sysext; there's an Arch package, a NixOS option, and a build script for everything
-else. `punktfunk-host hdr-probe` tells you exactly which pieces are in place.
-
-Opt-in for this release while it soaks: without the knob, or without the extra package, the
-gamescope path streams SDR exactly as before.
-
 ## Fixed: newer KDE Plasma silently fell back to a slower, less reliable way to arrange displays
 
 On current KDE Plasma (6.7 and newer), Punktfunk's direct way of talking to the desktop about display layout stopped seeing any displays at all, with no visible error — every session quietly fell back to an external helper tool that's known to hang under load, exactly the thing the previous release's Plasma fix was meant to stop relying on. Fixed: Punktfunk now recognizes both the way older and newer Plasma releases announce their displays.
@@ -42,16 +25,6 @@ On current KDE Plasma (6.7 and newer), Punktfunk's direct way of talking to the
 - **libei absolute-coordinate resolution changed from matching by mode size to `mapping_id` → origin → size → first**, since two outputs can share a size but never a top-left, and a mirrored head's region is not the client's stream size at all. A named anchor that matches nothing falls back down the ladder rather than stranding input, and both a miss and a match now log once per distinct answer. `punktfunk-host anchor-test` exercises the ladder against a live compositor with two same-sized outputs — the one case a unit test can only simulate.
 - **`SWAYSOCK` is now derived rather than required to be inherited** — by the compositor's known PID, then the newest socket owned by the host's user — closing the last session variable a `systemd --user` host lacked for sway. A new drop-in (`PartOf`/`WantedBy` on `graphical-session.target`, opt-in, shipped under `/usr/share`) restarts a desktop-login host with its desktop, so a Wayland socket and portal connection that die with a Plasma/GNOME restart don't leave the daemon silently unable to recover.
 - **KWin's in-process output-management path now also binds `kde_output_device_registry_v2`**, not just the per-output `kde_output_device_v2` globals it originally shipped with — KWin 6.7 stopped advertising the latter, so the module written specifically to avoid shelling out to `kscreen-doctor` (0.19.x) saw zero devices and silently degraded to it on every session. Both models are supported now; registry-sourced devices arrive one Wayland round-trip later, so the handshake gains one conditional extra barrier.
-- **gamescope HDR is decided statically, before the display exists.** The punktfunk/1 Welcome fixes a session's bit depth up front and cannot take it back (PQ frames on an 8-bit encoder are a deliberate hard error), so the capability answer is the identity of the gamescope binary the host will spawn — a `+pfhdr` marker in its `--version` banner, probed once per boot — never an optimistic negotiation. The capture-side gate became source-aware (`capturer_supports_hdr_for(compositor)`), the HDR negotiation-failure latch became per-source (a wedged monitor mirror no longer disables a gamescope session's HDR, or vice versa), and the keep-alive reuse key gained `hdr` so a display brought up SDR can never be handed to an HDR session. The GameStream plane's live BT.2100 monitor probe is now scoped to the portal source — a headless gamescope box has no monitor to be in HDR mode.
-- **The gamescope patch mirrors code already in gamescope's tree.** Its PipeWire node additionally offers `xRGB_210LE`/`xBGR_210LE` with MANDATORY SMPTE ST.2084 + BT.2020 properties, mapped to `DRM_FORMAT_XRGB2101010`/`XBGR2101010` — the same 10-bit capture texture the HDR AVIF screenshot path allocates — and `paint_pipewire()` composites into them with the HDR screenshot LUT set and `EOTF_PQ`, which is exactly the `bHDRScreenshot` branch. The new formats are listed last, so every existing consumer keeps negotiating the 8-bit stream bit-for-bit, and the fixed PQ container is deliberately *not* conditional on the focused app being HDR (a stream's colourimetry must not follow what the game happens to render).
-- **Vulkan Video learned 10-bit**, which is what keeps AMD/Intel HDR on the good path: an HDR session opens a 10-bit video profile (HEVC Main10 / AV1 Main at 10 bits) with a `G10X6…3PACK16` picture and DPB, headers carrying the depth and the BT.2020/PQ CICP triplet — an SPS `bit_depth_*_minus8 = 2` for HEVC, `high_bitdepth` plus the matching sequence-header OBU bits for AV1 — and a new `rgb2yuv10.comp`: the 8-bit BT.709 shader's twin, doing a pure BT.2020 NCL matrix on the already-PQ-encoded samples (there is no transfer function to apply, and applying one would be wrong) and writing 10-bit values into the high bits of `R16`/`RG16` scratch planes that are size-compatible with the picture's. That keeps the two things HDR would otherwise cost on this vendor: real RFI loss recovery, and the compute CSC's cursor blend — the only way a gamescope pointer reaches the stream at all.
-- **GameStream advertises its 10-bit codec bits per codec.** `ServerCodecModeSupport` layered `SCM_HEVC_MAIN10` and never `SCM_AV1_MAIN10` — a blanket omission on the theory that the GameStream AV1 path was unconfirmed, even though the SDR baseline has always offered AV1 Main8, so the depth was never the uncertain part. Each 10-bit bit is now gated on that codec's own `can_encode_10bit` probe AND the baseline already advertising it, and the RTSP honor degrades a session whose NEGOTIATED codec can't carry 10 bits rather than labelling an 8-bit stream PQ. `host_hdr_capable` became codec-agnostic to match (any 10-bit-capable codec makes the host HDR-capable; which one a session gets is the session's question).
-- **The Vulkan encode backend is now capability-probed per codec AND depth**, mirroring what the direct-SDK NVENC path already does with its GUID probe. `vkGetPhysicalDeviceVideoCapabilitiesKHR` is asked against the very profile the session open will build, so the dispatcher's prediction cannot disagree with reality: a device that can encode 10-bit keeps the Vulkan path, one that can't routes to libav VAAPI *before* burning a failed session open, and `can_encode_10bit` reports the union of what VAAPI and Vulkan Video can do rather than VAAPI's answer alone (which was under-reporting 10-bit on hardware that could do it).
-- **A gamescope session can now be zero-copy end to end.** gamescope keeps the pointer out of its PipeWire node (it lives on a hardware plane for scanout), so the host always reconstructed it from XFixes and blended it into every frame — and *that blend* is what forced the encode path onto its compute colour-conversion arm, since the zero-copy RGB-direct source hands the captured buffer to a fixed-function front end with no blend stage. A second carried gamescope patch adds `--pipewire-composite-cursor`, which paints it in using the same `MouseCursor::paint` call the scanout composite uses; the repaint test grows the cursor's state alongside the commit ids, so a pointer-only move still produces a frame and a hidden cursor is erased. The host reads a monotonic `+pfhdr<N>` patch level from the `--version` banner — one probe now answering both "can it do HDR" and "does it paint the cursor" — and stops attaching the XFixes reader and blending when the answer is yes.
-- **The two indirect spawn modes now verify that their flags arrived.** A host-managed `gamescope-session-plus` receives the HDR and cursor flags through `GAMESCOPE_BIN` + `PF_HDR_ARGS`, and SteamOS through a PATH shim — conventions, not guarantees. A session that ignored either would exec the distro's gamescope with none of them, and while the HDR half fails loudly (capture negotiation times out against the bit depth the Welcome already fixed), a lost `--pipewire-composite-cursor` was **silent**: the host had been told the compositor would paint the pointer, so it painted none, and the stream simply had no cursor. Both managed paths now read the running compositor's `/proc/<pid>/cmdline` once its node appears and refuse the session on a missing flag; since the plan is fixed by then (`cursor_blend` feeds the encoder open, which precedes the display), the capability is latched off for the process and the retry resolves a correct SDR host-composited session — one rejected attempt per boot, then it converges. The check fails **open** at every ambiguity: no flags expected, or no readable gamescope in `/proc`, says nothing.
-- **`punktfunk-gamescope` is now built by CI on the channels that ship it.** `rpm.yml` builds it in the matching Fedora container and hands it to `build-sysext.sh --gamescope`; `arch.yml` builds `packaging/gamescope/PKGBUILD` into the same pacman repo. Both are cached on `packaging/gamescope/**` (that tree depends on nothing else in the repo, so a normal push restores a binary rather than spending ten minutes on someone else's C++) and both are best-effort, because the packages those workflows exist to publish must not hinge on a gamescope build. The Arch PKGBUILD and the nix derivation had each drifted to a stale patch list — the PKGBUILD naming two patches when there are three, the nix override naming the level-1 banner patch after level 2 landed — so both now read the patch *directory*, and the PKGBUILD delegates the whole build to the shared script (asserting its own pinned rev matches) instead of re-deriving the meson invocation, which had already lost the `force_fallback_for=wlroots` that keeps the binary startable off the build host. The nix override's `gamescope.unwrapped` requirement was wrong on current nixpkgs, where `gamescope` *is* the buildable derivation; it now prefers `.unwrapped` where it exists and checks the result is something `overrideAttrs` can actually patch.
-- **The zero-CSC RGB-direct (EFC) source works in HDR too.** The `VK_VALVE_video_encode_rgb_conversion` probe now asks for the BT.2020 model and the captured 10-bit packed-RGB format instead of assuming BT.709/BGRA, and the session selects the matching model — so an HDR session with no pointer to composite (the GameStream desktop mirror) hands the captured buffer straight to the encoder's fixed-function front end and runs no host CSC at all. Sessions that DO composite a pointer keep the compute CSC, as before: the EFC cannot blend.
-- **NVIDIA gained a zero-copy HDR leg**, and the VAAPI fallback needed no new encoder code. The VAAPI path already ingested XR30 dmabufs into `format=p010:out_color_matrix=bt2020`. On NVIDIA the packed 10-bit frame now travels LINEAR dmabuf → Vulkan bridge → CUDA → NVENC `ARGB10`/`ABGR10`, letting NVENC do the BT.2020 CSC itself: no host CSC pass, no depth loss, and the cursor-blend compute shader gained two 10-bit modes so the pointer survives. The tiled EGL de-tile blit is still 8-bit and HDR never routes through it. A host without the direct-SDK NVENC backend keeps HDR on the CPU path, since libav's HDR route swscales into a P010 hardware frame that a packed-10-bit CUDA buffer cannot fill.
 - **CI-only fix:** the winget release-verification step's `envs:` allow-list was a step-level sibling of `with:`/`env:` instead of nested inside `with:`, so `appleboy/ssh-action` never actually received it as an input and the step kept failing on every tag. Moved to match how `REGISTRY_TOKEN` is already forwarded elsewhere.
 
 See [v0.20.1](v0.20.1.md) for the fixes carried by that release (Windows install/winget, GameStream opt-in default, the gamescope Game Mode takeover hardening, a laptop-panel stall fix, and the PyroWave latency-creep bundle) — all included here too, since this release supersedes it.