diff --git a/quest/m1/README.md b/quest/m1/README.md index 5c1d42933a3..e53e7bc7670 100644 --- a/quest/m1/README.md +++ b/quest/m1/README.md @@ -49,7 +49,6 @@ moves what remains of them once their claims merge. - [Socket endpoints split into servers and clients](/quest/m1/socket-roles.md) - servers announce `.server/` and clients `.client/`, so plain grants keep a client from posing as a server or reading another client - [CARLA's driver ids stay private](/quest/m1/carla-driver-ids.md) - 128-bit driver ids, and `status` shows a handle that isn't the endpoint id - [TS fidelity and recovery](/quest/m1/ts-reliability/README.md) - both modes have explicit profiles, automated validation, loss tests, and a dedicated guide -- [WebGPU on Safari](/quest/m1/webgpu-safari.md) - the WebGPU renderer is verified on Safari 26 for macOS and iOS - [Session policy lives in moq-net](/quest/m1/session-policy/README.md) - accept and connect deadlines, admission, window widening and reconnect live once in moq-net; moq-tokio and moq-relay's io_uring path call them - [Vendor and trim the WebRTC stack](/quest/m1/webrtc-vendor.md) - str0m, dimpl, `is`, and sctp-proto live in moq-webrtc with unused code removed, so moq-rtc can take rind - [MoQ on rind](/quest/m1/rind/README.md) - every MoQ crate takes its crypto from rind, a pure-Rust graviola fork, so no build compiles C crypto diff --git a/quest/m1/webgpu-safari.md b/quest/m1/webgpu-safari.md deleted file mode 100644 index 81f421b2812..00000000000 --- a/quest/m1/webgpu-safari.md +++ /dev/null @@ -1,90 +0,0 @@ -# [S] The WebGPU renderer is verified on Safari 26 - -## Goal - -`@moq/video`'s WebGPU path is verified on Safari 26, on macOS and iOS: H.264 -and biplanar frames import through `importExternalTexture`, orientation is -right, and a frame that fails to import is handled as decided below, with -evidence. - -## Plan - -Planned 2026-10-10 as a follow-up of -[moq-dev/moq#5138](https://github.com/moq-dev/moq/pull/5138), which shipped the -renderer without Safari verification. The original concerns were H.264 -frames, biplanar formats, and iOS orientation with `importExternalTexture`. - -- Run `` and the `` preview with `backend` set to - `auto` and `webgpu` on real Apple hardware or a hosted Safari session, over - H.264 and the other catalog codecs, in each orientation. -- [moq-dev/moq#5138](https://github.com/moq-dev/moq/pull/5138) decided an - unimportable frame stays loud (it throws and that frame is - skipped). Revisit with what Safari actually does: keep it, or report it - through `out.error` under its own error name. No valid macOS frame failed to - import in the run below, so this decision remains unchanged. -- Fix what breaks at its source. If the probe passes but frames fail, consider - extending the probe rather than a runtime fallback, which that PR rejected. - -Verification on 2026-11-03 used physical Apple M4 hardware (Mac16,13), macOS -26.6.2 build 25G83, and Safari 26.6.2 through its existing WebDriver setup. -The source was `976fece17349666bb2ab6dad93109408ac272e8d`. The adapter reported -`vendor: apple` and `isFallbackAdapter: false`; neither was overridden. -Pixel readback checked four colored quadrants, with no renderer changes. - -- Direct WebGPU rendering passed 80/80 cases: RGBA, NV12, I420, H.264 High, - Main and Baseline, HEVC, VP8, VP9 and AV1, each rotated 0/90/180/270 degrees - and mirrored or unmirrored. Decoded frames were NV12. VP9 and AV1 used - independent 640x360 FFmpeg fixtures; Safari's own encoders failed, and its - decoders also refused the earlier 64x32 fixtures. -- `` passed 112/112 cases: those seven encoded codec/profile - variants, both `auto` and `webgpu`, all four catalog rotations and both - mirror settings. Broadcasts used a page-local origin, so this is rendering - evidence, not transport interoperability evidence. -- `` passed 10/12 landscape image-source preview cases, covering - `source` and `encoded` with both backends for H.264 High, HEVC and VP8. Both - HEVC `encoded` cases failed before rendering with - `preview: decode error EncodingError: Decoder failure`: Safari accepts - `hevc: { format: "annexb" }` but returns length-prefixed chunks (prefix - `00 00 00 3b 4e 01`) with a 109-byte hvcC description. - -✅ Safari HEVC: probe and refuse (maintainer, 2026-11-03, on -[moq-dev/moq#7765](https://github.com/moq-dev/moq/pull/7765)). The encoder -setup probe in `@moq/publish` now checks that H.264 and HEVC output opens with -an Annex B start code and otherwise skips that codec, so publishing and the -encoded preview pick the next one. No runtime fallback. On the same Safari -26.6.2 Mac, the probe refused `hev1.1.6.L93.B0` and resolved `avc1.640028`, -whose output was Annex B. - -Remaining evidence includes real iOS capture and physical orientation, -publisher preview orientation, and a valid frame that actually fails import. -A deliberately closed RGBA frame did not induce an import exception or -validation error in the macOS session; it does not verify failure handling. -Keep the renderer's existing error behavior pending such evidence. - -Public API: possibly a new `out.error` name. Wire: none. - -## Waiting - -- Maintainer: run Safari 26 on a physical iPhone or iPad (a simulator does not - count), with Web Inspector attached from a Mac. Serve the `main` demo - (`just demo`) to the device over HTTPS, since camera capture needs a secure - context. Then: - 1. Open `publish.html`, publish the rear camera, then the front camera. In - the console, run `document.querySelector("moq-publish").backend = "webgpu"`, - then `"auto"`. For each, rotate the device through portrait, landscape - left, landscape right, and upside down. Check that the preview stays - upright and matches the scene. - 2. Open `watch.html` on the same broadcast and set - `document.querySelector("moq-watch").backend` to `"webgpu"`, then `"auto"`. - Repeat the rotations and check that the picture stays upright. - 3. In the console, run `(await navigator.gpu.requestAdapter()).info`. Record - its result, the iOS and Safari versions, any `moq-publish: skipping` - warning, and any `importExternalTexture` or WebGPU validation error in - this quest. - - Done when those results are recorded here, so the remaining work can be - fixed at its source or the quest finished. - -## Related - -- [WebGPU HDR](/quest/m5/webgpu-hdr.md) - builds on the same renderer