Skip to content

The truly free, open-source, local-first, cross-platform CapCut replacement. Scriptable through a JSON-RPC API & MCP.

chore: drop the audits and ARCHITECTURE.md

Jareer4a859fe8e961
diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md
deleted file mode 100644
index dd06ba2a..00000000
--- a/ARCHITECTURE.md
+++ /dev/null
@@ -1,680 +0,0 @@
-# How Concat works, and where it stands
-
-A map of the engine and the window for someone who wants to change one
-part without reading the rest, written against the tree at `7223d64`
-(11 October 2026, 71 commits past the 0.2.6 tag). Every section names the
-crate and the file to open when the picture is not enough; the last
-section says what is known to be unfinished or wrong today, with the
-audit that says so in detail. Diagrams are Mermaid, which GitHub renders.
-
-Concat is a video editor in Rust. The engine is a set of crates with no
-window in them; the window is a Slint application over that engine; a
-command-line tool and a socket server drive the same engine through the
-same API. One document format, one command set, one renderer, three ways
-in. Nothing is spawned at run time: FFmpeg is linked, whisper.cpp and
-sherpa-onnx are compiled in, ONNX Runtime is loaded for the models.
-
-## 0. The tree in numbers
-
-| | |
-|---|---:|
-| Crates in the workspace (`src/crates/`) | 16 |
-| Rust | 91 697 lines |
-| Slint (the window's `ui/`) | 29 790 lines in 65 files |
-| WGSL (effect packages) | 1 764 lines |
-| Java (the Android bridge) | 291 lines |
-| Effect packages, built in | 123 (98 on format 2 with a shader; 25 sound packages on FFmpeg chains) |
-| Locales | 14, 1 016 keys each, 0 missing |
-| Downloadable models | 23, every one with a SHA-256, mirrored on the `models-v1` release |
-| Tests (`#[test]`) | 719 in 93 files |
-| `unsafe` sites | 68, all in FFmpeg, JNI, Wayland, mmap and platform calls |
-| Largest file | `concat/src/studio.rs`, 10 689 lines |
-| Toolchain | Rust 1.93, pinned in `src/rust-toolchain.toml`; Slint 1.18.1; wgpu 30; FFmpeg 7+ through ffmpeg-the-third 6 |
-
-## 1. The crates and who depends on whom
-
-Every arrow points one way. Nothing below the window knows a window
-exists; nothing but `concat-media` knows FFmpeg exists; `concat-core` has
-no dependencies at all. The edges are the `[dependencies]` tables of the
-manifests, not a wish.
-
-```mermaid
-graph BT
- core["concat-core · 2 897<br/>Rational time, arenas, Frame, Timeline,<br/>animation, speed curves, ShaderPass<br/>std only"]
- project["concat-project · 10 285<br/>the document: model, commands,<br/>placement rule, undo, concat.json"]
- media["concat-media · 9 897<br/>FFmpeg: probe, decode, encode, audio mix,<br/>reader pool, prefetch, hardware decode"]
- render["concat-render · 7 450<br/>FramePlan; the wgpu compositor;<br/>SSIM; the CPU oracle (tests only)"]
- effects["concat-effects · 7 321<br/>packages: manifest + WGSL or chain,<br/>catalogue, looks from .cube"]
- vision["concat-vision · 2 760<br/>cutout masks, brushes, enhance<br/>(ONNX Runtime behind `infer`)"]
- text["concat-text · 2 607<br/>titles as pixels; no engine deps"]
- export["concat-export · 3 871<br/>document → clip list → Timeline;<br/>the frame loop; the monitor's frame"]
- host["concat-host · 12 589<br/>Session, projects, playback, monitor,<br/>proxies, scheduler, jobs, titles, cards,<br/>templates, record, updates, dirs"]
- speech["concat-speech · 3 109<br/>whisper, Kokoro / Pocket / Chatterbox"]
- api["concat-api · 2 352<br/>one dispatcher: every verb as JSON"]
- server["concat-server · 1 411<br/>JSON-RPC lines, gRPC, tokens, Hub"]
- cli["concat-cli · 514"]
- window["concat · 23 080 Rust + 29 790 Slint<br/>the editor window"]
- android["concat-android · 462 + Java<br/>the activity"]
-
- project --> core
- media --> core
- render --> core
- effects --> core
- effects --> project
- vision --> core
- vision --> project
- export --> core
- export --> media
- export --> render
- export --> effects
- export --> project
- export --> vision
- host --> export
- host --> text
- host --> media
- host --> render
- host --> vision
- speech --> host
- speech --> media
- api --> host
- api --> export
- api --> effects
- api --> project
- server --> api
- cli --> server
- cli --> api
- window --> server
- window --> speech
- window --> api
- android --> window
-```
-
-Three things the graph says that matter when adding code:
-
-- **`concat-render` does not know what an effect package is.** It takes a
- `concat_core::ShaderPass` (compiled source, laid-out uniform bytes) and
- draws it. The catalogue in `concat-effects` makes those; `concat-export`
- is where the two meet (`shader_passes_at`).
-- **`concat-text` is an island**: fonts, shaping and painting with no
- engine type in its signature. `concat-host/src/titles.rs` is the seam
- that turns a title into an image clip for the rest of the pipeline.
-- **The window depends on `concat-api`**, not to drive itself through it
- (it has its own `Session`), but for the Remote page, which embeds a
- server, and for the register of open project folders the two share.
-
-## 2. From the document to a pixel
-
-A clip exists in three shapes on its way to the screen, and each crossing
-is one function in one file.
-
-```mermaid
-flowchart LR
- doc["Project<br/>concat-project::model<br/>f64 seconds, string ids,<br/>keys as fractions of the clip"]
- flat["Vec&lt;ExportClip&gt;<br/>concat-export::flatten<br/>one list, tracks by index,<br/>titles rejoined as stills"]
- tl["core::Timeline<br/>concat-export::resolve<br/>Rational time, arena ids,<br/>quantised to the frame grid,<br/>sizes, chains, passes, masks decided"]
- plan["FramePlan<br/>concat-render::plan_frame(t)<br/>what is on screen at t: media,<br/>source time, placement, opacity,<br/>blend, per layer, bottom first"]
- filled["FramePlan, filled<br/>concat-export<br/>+ decoded picture, effects resolved<br/>for t, mask, transitions, treatments"]
- gpu["WgpuCompositor<br/>concat-render::gpu<br/>the GPU, or WARP / lavapipe"]
- enc["Encoder<br/>concat-media::encode"]
- tex["a texture on the<br/>window's device"]
-
- doc -- "flatten_timeline" --> flat
- flat -- "build_timeline" --> tl
- tl -- "plan_frame(t)" --> plan
- plan -- "decode + passes_at" --> filled
- filled --> gpu
- gpu -- "composite: read back" --> enc
- gpu -- "composite_texture: stays on the GPU" --> tex
-```
-
-- **The document** (`concat-project/src/model.rs`) is what a person edits
- and what is saved. Times are `f64` seconds, ids are strings, every
- keyframe is a fraction of its clip's length, and serde names are
- camelCase because that is the file's spelling. Every change is a
- `Command` (`commands/`), and `Clip::tidy` is the one place clamps live:
- every time field is held to `MAX_TIME` (a hundred hours), every size to
- `concat_core::frame::MAX_SIDE` (8 192) through `VideoSettings::is_sane`.
-- **Placement is a model rule** (`placement.rs`): no clip ever covers
- another on its lane. Every command that places, moves or lengthens a
- clip asks it, and so does the window's drag, so the echo under the
- pointer shows where the drop will land.
-- **Flatten and resolve** (`concat-export/src/flatten.rs`, `resolve.rs`)
- turn the document into the engine's `Timeline`: rational time quantised
- to the frame grid so equality is exact, plus the per-clip facts the model
- has no field for (decode sizes, filter chains, shader passes, cutout
- masks, layer treatments). `render` and the preview read what `resolve`
- built and never look at an `ExportClip` again.
-- **The plan** (`concat-render/src/plan.rs`) is pure: no files, no pixels.
- Geometry (crop, flips, fit, centre, scale, turn) and weighing (fades
- folded into a scale and offset, wipes into edges, mask, opacity) are
- computed once here, so no backend can drift on them. The export fills
- the plan with the decoded picture, the effects resolved for that instant
- and the treatments live over the stack, and hands it to a compositor,
- which takes a `FramePlan` and nothing else.
-- **One compositor.** `WgpuCompositor` draws every frame, the monitor's
- and the export's; a machine without a GPU runs it on the platform's
- software adapter (WARP on Windows, lavapipe on Linux), and a machine
- with neither is told so. The CPU compositor that was the reference is
- kept in the tests alone (`concat-render/src/reference.rs`) as the oracle
- the parity suite (`gpu/tests.rs`) holds the GPU's geometry, masks,
- transitions and blending to, at a structural similarity above 0.99
- (`metrics.rs`). Every device, the compositor's own and the window's,
- carries an uncaptured-error handler that logs, so a driver error no scope
- caught is a wrong frame and a line in the log, never a dead thread.
-
-### 2.1 Colour
-
-```mermaid
-flowchart LR
- sdr["SDR source<br/>8-bit RGBA, Rec. 709, gamma"]
- hdr["HDR source<br/>HLG or PQ, Rec. 2020<br/>decoded deep: 16-bit, its own signal"]
- work["Working space<br/>linear light, extended Rec. 709,<br/>half floats, 1.0 = SDR white (203 nit)<br/>(scRGB / extended linear sRGB)"]
- sdrout["SDR frame<br/>gamma, 8-bit"]
- hdrout["HDR file<br/>HLG or PQ on Rec. 2020,<br/>16-bit integers → HEVC / AV1 10-bit,<br/>MaxCLL / MaxFALL measured as written"]
- screen["an SDR screen<br/>BT.2390 roll-off"]
-
- sdr -- "COPY_SHADER on upload" --> work
- hdr -- "DEEP_SHADER on upload:<br/>transfer undone, primaries to 709,<br/>conformed on an SDR timeline" --> work
- work -- "resolve" --> sdrout
- work -- "deliver_hdr" --> hdrout
- work -- "resolve, rolled off" --> screen
-```
-
-Blending, opacity and fades happen in light, as in Resolve and Final Cut.
-A Rec. 2020 source is decoded deep (`DecodeOptions::deep`: sixteen bits a
-channel, no CPU tone map) wherever its frame goes straight to the GPU. A
-timeline has a colour it is output in (`VideoSettings::color_space`: SDR,
-HLG or PQ); `Editor::follow_first_hdr` turns an SDR timeline HLG when a
-command puts its first HDR clip on it, in that command's undo step. A clip
-with a cutout keeps the eight-bit tone map in the decoder. (`gpu.rs`,
-`COPY_SHADER`, `DEEP_SHADER`, `FramePlan::output`, `Compositor::deliver_hdr`,
-`Encoder::create_hdr`.)
-
-## 3. Effects
-
-An effect is a folder under `concat-effects/packages/`: `effect.toml`
-(`format = 2`) naming its knobs, and `effect.wgsl` declaring a `Params`
-struct and `fn effect(uv)`, drawn on the GPU in light. A sound package is
-an FFmpeg chain template instead. Nothing in Rust names an individual
-effect: the window shows the knobs, the engine runs the backend, and a
-document stores `{ id, params, enabled }` and nothing more.
-
-```mermaid
-flowchart TD
- pkg["package folder<br/>effect.toml + effect.wgsl (+ lut.cube)<br/>123 built in, user folders at run time"]
- cat["Catalogue<br/>concat-effects::catalogue<br/>parsed, validated, kept;<br/>categories and shelves from the manifests"]
- stitch["shader.rs<br/>host prelude stitched round the body,<br/>parsed by naga, Params layout read,<br/>bindings and loops checked"]
- trial["install trial (user packages)<br/>drawn once on a sibling compositor<br/>at Default, Min and Max, 3 s timeout"]
- pass["ShaderPass (concat-core)<br/>compiled source, uniform bytes,<br/>values by key, intensity, LUT, stages"]
- gpu["GPU: pipeline cached by key,<br/>params as a uniform buffer,<br/>stages into pooled pictures"]
-
- pkg --> cat --> stitch --> trial
- cat -- "shader_passes_at(effects, t)" --> pass --> gpu
-```
-
-- A shader that binds anything the host did not declare, loops without a
- break, or asks for a look-up table over 65 a side is refused at load.
- A `[[wgsl.pass]]` draws a picture earlier passes cannot read and later
- ones sample through `<target>_at(uv)`; `shrink` lets a blur run at a
- fraction of the layer's pixels. Only the last pass mixes by intensity.
-- A shader works in light (`space = "linear"`), in the display encoding
- (`"display"`, colour looks), or in log (`"log"`, ACEScct). A `.cube`
- imported as a look (`looks.rs`) is a format 2 filter reading its table
- through `look()`.
-- `[[replaces]]` lets a package stand in for a retired one when a project
- opens, and `Catalogue::add` refuses a user package that claims a
- built-in's id, a bare id, or one already stood in for, with a reason.
-- Knobs are numbers; a `wheel` is three (`.x`, `.y`, `.m`), a `curve` up to
- eight points, a `color` is RGBA packed into one. `concat/src/grading.rs`
- and `ui/inspector/grading.slint` draw them.
-- Every package draws its own **card** (`concat-host/src/cards.rs`): one
- reference still through the package at its defaults, drawn by the same
- shader and compositor a clip's frames use, kept as a JPEG named by a
- fingerprint of everything it was drawn from.
-
-## 4. Decoding, caching and scheduling
-
-Export decodes every frame once, in order, with one paced decoder per
-clip, and does not touch the pool, with one exception: a clip that runs
-backwards or on a speed curve is sought frame by frame through
-`ReaderPool::frame_at`, which bakes the chain in the same way. Everything
-interactive goes through the pool and the one scheduler.
-
-```mermaid
-flowchart LR
- transport["transport<br/>(playback, scrub)"]
- cursor["Cursor {time, direction, rate}"]
- sched["Prefetcher<br/>concat-media::prefetch<br/>2-4 threads, one queue, one kept<br/>clear of background work<br/>Playback > Filmstrip > Artwork > Proxy"]
- pool["ReaderPool<br/>concat-media::pool<br/>source cache: (file, level, frame),<br/>a share of RAM (host::memory)<br/>treated cache: + crop, fit, chain<br/>one warm reader per (file, level)"]
- dec["Decoder<br/>concat-media::decode<br/>seek, turn, scale, libavfilter chain,<br/>pace, deep for HDR"]
- hw["hardware.rs<br/>VideoToolbox · D3D11VA · MediaCodec · VAAPI<br/>only where it wins; falls back at open,<br/>copy-back and send"]
- proxy["proxy<br/>concat-host::proxy<br/>a smaller H.264 of anything over HD<br/>in cache/proxy, written on the Proxy lane"]
- gpupool["GPU texture pool<br/>gpu.rs: POOL_TEXTURES 1024,<br/>also the source-texture cache by frame id"]
-
- transport --> cursor --> sched
- sched -- "frames ahead, pinned until passed" --> pool
- pool --> dec --> hw
- proxy -- "adopted as the file's stand-in<br/>for playback and filmstrips" --> pool
- sched -- "written once" --> proxy
- pool -- "upload" --> gpupool
-```
-
-- **The source cache** is keyed by the file, the level it was decoded at
- (the file's own size halved while it still covers what was asked) and
- the frame index, and nothing else. The crop, the fit and the effect
- chain are applied on the way out and kept in a second, smaller cache,
- so turning a knob costs a filter per frame and a scrub back over
- covered ground costs a lookup.
-- **The scheduler** is one per process (`concat_host::scheduler()`), and
- owns the pool the monitors read. The frames ahead of the playhead come
- first, then filmstrips, the bin's artwork, then proxies; an import of
- twenty files is twenty jobs on the same few threads, never twenty
- decoders.
-- **Hardware decode** is a process-wide preference the window turns on
- at start, asked for stream by stream only where the hardware is faster
- (not for 8-bit H.264). Any failure falls back to software from the last
- delivered frame.
-- **The monitor** (`concat-host/src/preview.rs`, `concat/src/panes/
- monitor.rs`) asks for a frame in two calls: `frame_sources` on a worker,
- `texture_of` on the thread that owns the device, so decoded pictures go
- up once and no pixel comes back down. Full quality plays the files
- themselves; Half and Quarter play the proxies.
-
-## 5. Sound
-
-Sound is FFmpeg's in both places, and has exactly one definition of what
-speed, fades and gain mean: `concat_media::audio::mix_graph`, a pure
-function from audible clips to one filtergraph.
-
-```mermaid
-flowchart LR
- subgraph playback["Playback (concat-host::playback)"]
- span["each audible clip's span<br/>decoded once to PCM, chain and<br/>speed baked in, in cache/audio,<br/>memory-mapped, byte-budgeted"]
- mix["one cpal stream mixes the maps<br/>sample by sample; gain and fades at mix<br/>time; the device's sample counter<br/>is the clock; supervised across<br/>device changes"]
- span --> mix
- end
- subgraph export["Export (concat-export, concat-media::audio)"]
- graph_["mix_graph: one filtergraph<br/>for every audible clip"]
- file["mix_to_file: one pass"]
- mux["mux: joins sound and picture<br/>into a scratch sibling; the output<br/>path is touched once, by a rename"]
- graph_ --> file --> mux
- end
- samples["samples.rs: one audio decoder for<br/>peaks, the playback cache, the transcriber"]
- peaks["peaks.rs: min/max buckets<br/>at a fixed rate, streamed"]
- rec["record.rs: the microphone<br/>to a WAV in the project's audio folder"]
- meters["concat/src/meters.rs: playback's<br/>peaks onto the Meters global, 30 Hz"]
- samples --> span
- samples --> peaks
- mix --> meters
-```
-
-## 6. The window
-
-The window is one Slint tree published from Rust. The controller is
-`Studio` (`concat/src/studio.rs`); each pane owns its state and is changed
-only by its own messages. Two kinds of state live in `Studio` and the line
-between them is the whole design: **the edit** is the engine's, read back
-from the open `Session` and written only as `Command`s; **the view**
-(selection, playhead, zoom, tool, the dock, locks, what the dialogs show)
-is the window's and never reaches the document.
-
-```mermaid
-sequenceDiagram
- participant Slint
- participant lib as lib.rs (callbacks)
- participant Studio as Studio::handle
- participant Pane as pane.update(msg, &mut Studio)
- participant Host as host.rs
- participant Worker
- Slint->>lib: callback (a click, an edit)
- lib->>Studio: handle(Msg::Pane(msg)) under one RefCell borrow
- Studio->>Pane: take the pane out, update
- Pane->>Host: spawn_in_project(work, then)
- Host->>Worker: a thread, stamped with the project epoch
- Worker-->>Host: result
- Host-->>Studio: invoke_from_event_loop → deliver: dropped if the epoch moved
- Studio->>Slint: publish(): every pane's data(), rows synced by diff
-```
-
-- **Threads.** Slint's models and properties are touched on the
- event-loop thread only, so the state lives in a plain `RefCell` with no
- lock. Everything slow runs through `host::spawn` or
- `host::spawn_in_project` and comes back through
- `slint::invoke_from_event_loop`; a result for a project that has since
- closed is dropped in one place, by epoch. `host::after_dialog` opens a
- native file dialog on its own turn of the loop with nothing borrowed,
- because on macOS a dialog pumps Slint's timers and the transport and
- meter timers borrow the studio too.
-- **Panes** (`concat/src/panes/`): captions, export, media bin, monitor,
- project sheet, relink, settings, speech, start, timeline view,
- voiceover. Each is `state + Msg + update + data`.
-- **What stays on the controller:** the gestures. A clip dragged,
- trimmed or razored, a picture moved on the stage, a brush stroke: one
- gesture spans the lanes and the stage over an *echo* of the document, a
- clone the pointer mutates and commits as one command on release, so undo
- undoes the drag and not a pixel of it. The clipboard holds a selection
- with each clip's lane row, and a paste lays the group down with its shape
- kept, in this or another timeline, as one undo step (`paste_plan`).
-- **Publishing** rebuilds each pane's Slint data on every event; row
- models go through `sync`, which diffs against the last published rows.
- The lanes report their width and the controller publishes only the clips
- that intersect the visible window plus one screen either side. During
- playback `spawn_frame` publishes only the picture.
-- **The monitor** draws on the window's own wgpu device
- (`concat/src/gpu.rs`), created before the backend is selected so Slint
- and the compositor share it; a frame is a texture Slint samples with no
- readback. On Android the backend owns its device and none is shared.
-- **The dock** (`dock.rs`, `ui/workspace/`) is a tree of splits and
- views walked flat for Slint; the workspace's panes are the monitor, the
- timeline, the library (media, effects, text, stickers), the inspector,
- keyframes, meters and config.
-- **Words.** Every string a person reads passes through `i18n::t` or
- `I18n.t` by a dotted key; `en.json` sits under every language; a user's
- `locales/<code>.json` in the config folder lays over a shipped one.
- `fonts.rs` lends the base fonts to Slint and a Noto Sans CJK subset as
- the fallback for Chinese, Japanese and Korean. `presets.rs` makes text
- presets data: TOML folders, built in or the user's own.
-- **The launch screen** (`ui/start.slint`) is a rail of verbs beside the
- projects this machine has opened; the new-project form is a sheet at the
- window root. **Preferences** (`prefs.rs`) are one small JSON file: the
- theme, chosen models, languages; nothing about a project.
-- **Platform** (`platform.rs`) keeps the differences in one file with the
- reason at each branch: the backend, pickers, the title strip, drops from
- outside (`wayland_drop.rs` listens for the Wayland drop itself), and the
- phone's publish hook.
-
-### 6.1 The phone shell
-
-`ui/phone/` is what the editor is on Android and iOS (`platform::phone`,
-or `CONCAT_PHONE=1` on a desk) in place of the title strip and the dock:
-a top bar, the monitor, a transport, the lanes with the playhead held at
-their middle, and a bar of tools that is the library's pages with nothing
-selected and the clip's verbs with one selected. A tool with more to say
-opens a sheet over the lanes holding the same inspector page or shelf the
-desk shows. `concat-android` is the activity: it points the host's XDG
-bases at the app's own folders, forwards the log to logcat, hosts a
-headless fragment for the document picker, and publishes a finished
-export into the phone's Movies through the media store
-(`ConcatFiles.java`).
-
-## 7. The document, undo and the file
-
-```mermaid
-flowchart LR
- cmd["Command<br/>commands/{clips,properties,audio,<br/>tracks,timelines,media}.rs"] --> run["run(&mut Project)<br/>placement asked, Clip::tidy clamps"]
- run --> snap["snapshot: Arc-shared,<br/>only the touched clip copied"]
- snap --> undo["undo stack, depth 200;<br/>a named gesture folds into one step;<br/>view state never enters"]
- file["concat.json<br/>version 1 + document"] -- "1. migrate, oldest shape first;<br/>a newer document is refused" --> parse["2. parse: serde derives,<br/>defaults for the missing,<br/>unknown fields kept in `extra`"] --> settle["3. settle: orphans dropped,<br/>tidy pass, active timeline exists"] --> proj["Project"]
- proj -- "derive Serialize; temp file, sync_all, rename" --> file
-```
-
-A hand-edited or older file degrades to something openable; nothing short
-of a document that is not an object, or has no timeline, fails the load.
-`DOCUMENT_VERSION` is 1 and the format is frozen by the documents that
-exist: `wire::clip_kind` falls back to `Video` for a kind it does not know,
-which is why shapes saved by a newer build vanish in 0.2.5 (section 12).
-
-## 8. The API and the three doors
-
-```mermaid
-flowchart LR
- window["the window<br/>its own Session, in-process"] --> host["Session, jobs, export<br/>concat-host"]
- remote["Remote page: embeds a Server<br/>with sessions of its own"] --> hub
- cli["concat-cli api<br/>JSON-RPC lines on stdin/stdout"] --> api
- json["JSON-RPC lines<br/>TCP or a Unix socket<br/>MAX_LINE 4 MB · 64 seats · 10 s to auth"] --> hub
- grpc["gRPC (`grpc` feature)<br/>same methods, JSON in a protobuf envelope"] --> hub
- hub["Hub: one thread owns the one Api;<br/>every caller in turn; catch_unwind;<br/>events fanned out to subscribers"] --> api["Api (concat-api)<br/>dispatch(Request) → Response + Events"]
- api --> host
-```
-
-- **The verbs** (`concat-api/src/message.rs`): `project.{create, open,
- close, list, get, save, document}`, `edit.{apply, undo, redo}` carrying
- `concat_project` commands as they are, `media.{probe, import}`,
- `catalogue.list`, `template.{list, save, instantiate}`, `export.run` and
- `export.cancel` as a job with `export.{progress, done, failed}` events,
- `preview.frame`, `cutout.progress`. `version` reports `capabilities`.
-- **The door is narrow.** `Server::start` mints a 128-bit token when none
- is configured, loopback included; every connection presents it first and
- the comparison is constant-time (`token.rs`). The API writes only under
- its roots (`Config::roots`, the home folder by default) and bounds a
- frame or an export by `MAX_SIDE`. There is no encryption: a bind off
- loopback belongs behind something that provides it.
-- The window is not a client of its own API; what the embedded server and
- the window share is the export slot (one export at a time across them)
- and `concat_api::OpenProjects`, so neither opens a folder the other is
- editing.
-- The README names an MCP transport. There is none in the tree; the
- socket transports are JSON-RPC and gRPC, and an MCP server would be a
- third transport over the same `Api`.
-
-## 9. Speech, vision and text
-
-```mermaid
-flowchart LR
- manifest["models/manifest.toml<br/>23 models, each with bytes + SHA-256,<br/>upstreams pinned to commits,<br/>mirrored on release `models-v1`"]
- dl["download: mirror first, upstream second;<br/>.part then rename; refused without a digest<br/>(concat-host::models and each crate's table)"]
- whisper["concat-speech::transcribe<br/>whisper.cpp in-process (Metal on a Mac)<br/>6 model sizes → timed caption segments"]
- tts["concat-speech::tts<br/>Kokoro (built-in speakers) and Pocket TTS<br/>(a voice from a recording) through sherpa-onnx;<br/>Chatterbox Turbo through ONNX Runtime,<br/>desktop only, behind `chatterbox`"]
- cutout["concat-vision::segment<br/>RVM person · ISNet object · SlimSAM brush,<br/>ONNX Runtime: CoreML / DirectML / NNAPI;<br/>a small person model compiled in"]
- enhance["concat-vision::enhance<br/>Real-ESRGAN compact, tiled,<br/>×4 averaged down to ×2 or ×1"]
- store["masks: 8-bit PNGs in the project<br/>folder, keyed by media and instant,<br/>MASK_RATE 10 a second; strokes in<br/>source fractions beside them"]
- text["concat-text<br/>face found, lines shaped (rustybuzz),<br/>outlines painted: plate, shadow, outline,<br/>fill, colour emoji probed not trusted;<br/>frame-sized canvas, sizes as fractions<br/>of the frame height"]
- titles["concat-host::titles<br/>a title or shape becomes an image clip,<br/>keyed by everything that changes its pixels"]
-
- manifest --> dl
- dl --> whisper
- dl --> tts
- dl --> cutout
- dl --> enhance
- cutout --> store
- text --> titles
-```
-
-Every long job is one at a time through a `SingleFlight`
-(`concat-host/src/jobs.rs`): export, transcription, synthesis, cutout,
-brush, enhance, reverse, a model download. `begin` refuses a second run
-and every run gets its own cancel flag, so a cancel can only stop the job
-that is running.
-
-## 10. Threads and where work runs
-
-| Thread or pool | Owner | What runs there |
-|---|---|---|
-| The event loop | Slint | every callback, `Studio::handle`, `publish`, drawing the monitor's texture, the transport and meter timers |
-| `host::spawn` threads | `concat/src/host.rs` | probes, saves, imports, caption and speech runs, anything a pane starts; results return by epoch |
-| The scheduler, 2-4 threads | `concat_host::scheduler()` → `Prefetcher` | frames ahead of the playhead, filmstrips, artwork, proxies; one thread kept clear of the low lanes |
-| Playback decode workers (2) and the cpal callback | `concat-host/src/playback.rs` | spans to PCM; the mix |
-| The export thread | `concat-host/src/export.rs` → `concat-export::render_on` | the paced decoders, the compositor on its own device, the encoder, the mix, the mux |
-| Job threads | `SingleFlight` owners | cutout, brush, enhance, reverse, downloads |
-| The cards painter | `concat-host/src/cards.rs` | draws missing effect cards on a thread of its own |
-| The Hub thread | `concat-server/src/hub.rs` | the one `Api`; transports are threads that wait on it |
-| The recorder thread | `concat-host/src/record.rs` | owns the cpal input stream from open to close |
-
-## 11. On disk
-
-```mermaid
-flowchart TB
- subgraph app["App folders (concat-host::dirs)"]
- cfg["config: settings.json, recents, prefs,<br/>locales/, text-presets/, templates/, packages/<br/>(%APPDATA% · ~/.config · Application Support)"]
- data["data: whisper-models/, tts-models/, cutout models,<br/>titles/, cards/, logs/<br/>(%LOCALAPPDATA% · ~/.local/share · Application Support)"]
- portable["or one `portable/` folder beside the exe:<br/>nothing written anywhere else"]
- end
- subgraph project["A project folder"]
- json["concat.json (written at creation; atomic saves)"]
- audio["audio/ (voiceover takes, narration)"]
- cache["cache/ audio/ (PCM maps, budgeted) · proxy/ (named by path, size, mtime) ·<br/>masks and strokes · waveforms · filmstrips"]
- end
-```
-
-The recents list is the machine's, not the project's, so a folder copied
-elsewhere carries no one's history with it. A project is a real thing from
-the moment it is created, and `projects::save` is the one code path that
-writes `concat.json` (temp file, `sync_all`, rename).
-
-## 12. Build, CI and release
-
-```sh
-cd src
-cargo dev # the window, --profile quick: optimised engine, unoptimised window, incremental
-cargo build --profile app -p concat # the shipping binary: fat LTO, panic=abort, stripped
-cargo test --workspace # 719 tests; needs FFmpeg's DLLs on PATH on Windows
-cargo run -p concat-cli -- serve # the API on 127.0.0.1:7420
-python scripts/locales.py --check # every locale against the inventory
-python scripts/models.py --check # models/manifest.toml against the engine's tables
-```
-
-A build needs FFmpeg 7+ development libraries (`FFMPEG_DIR` on Windows),
-cmake and a C++ toolchain for whisper.cpp, and on Linux the fontconfig and
-freetype headers for Skia. The window draws with Slint's Skia renderer
-from prebuilt binaries; `src/patches/i-slint-renderer-skia` is a vendored
-copy of that renderer with one function changed so glyphs sit on whole
-pixels at 1x on Windows (#278), and has to be carried across every Slint
-upgrade. `flake.nix` builds the window with every native dependency
-pinned on Linux.
-
-```mermaid
-flowchart LR
- push["push / PR"] --> ci
- subgraph ci["ci.yml"]
- engine["engine (ubuntu): fmt, locales, models,<br/>clippy -D warnings, test -j1 with lavapipe<br/>(CONCAT_REQUIRE_GPU),<br/>grpc clippy + tests"]
- desktop["desktop: macos-15 cargo check;<br/>windows-2022 full tests, 120 min job timeout"]
- end
- tag["tag v*"] --> rel
- subgraph rel["release.yml"]
- check["check: tag = workspace version"]
- gate["ci.yml on the tag"]
- build["build-app.yml ref: tag<br/>mac arm + x64 (dmg, notarised), linux x64 + arm<br/>(deb, rpm, AppImage, Arch), win x64 + arm (setup exe, msi)"]
- phones["mobile.yml ref: tag, sign: true<br/>APK signed in its own step; iOS ipa ad-hoc"]
- publish["GitHub release: assets, checksums,<br/>manifest.json for self-update,<br/>notes from the commits"]
- check --> gate --> build --> publish
- gate --> phones --> publish
- end
- nix["nix.yml: nix build + ldd check"]
- models["models.yml (manual): fill the models-v1 mirror"]
-```
-
-Self-update (`concat-host/src/updates.rs`) reads the release list from
-GitHub, the package for this machine from the chosen release's
-`manifest.json`, checks the bytes against the manifest's digest, and
-hands the package to the platform: the installer on Windows, a copy over
-the bundle on macOS, a write over the AppImage, the desktop's installer
-for a .deb, .rpm or pacman package. Only releases from 0.2.5 on are
-offered. The chain trusts the manifest beside the package and nothing
-signs it; see section 13.
-
-## 13. Testing and measuring
-
-| Suite | Where | What it holds |
-|---|---|---|
-| Unit tests, 719 | every crate but concat-android | the arithmetic, the commands, the placement rule, the reader, the plan, the tolerant document reader |
-| Export end to end | `concat-host/tests/export.rs` | every edit a person can make exports through real `Session` commands over synthetic media and is read back, pixels pinned; the rename path with sound |
-| Parity | `concat-render/src/gpu/tests.rs` | the GPU against the CPU oracle by SSIM, one plan per feature; runs on lavapipe in CI, skips on WARP unless `CONCAT_REQUIRE_GPU` |
-| Hostile packages | `concat-effects/src/shader.rs` tests | the unbounded loop, the extra binding, the oversized table are refused |
-| Package fixtures | `fixtures.toml` per package | colours in and out through the shader (`[[probe]]`); chain strings at default, min and max |
-| Locales | `concat/src/i18n.rs` tests, `scripts/locales.py` | every shipped locale covers the inventory; the CJK face covers every character the locale files use |
-
-What is not measured: the Slint repaint itself (`SLINT_DEBUG_PERFORMANCE`
-needs the window), and the window's 23 000 lines of Rust have 73 tests,
-the Slint tree none.
-
-## 14. Where to look
-
-| To change | Open |
-|---|---|
-| what a clip can be | `concat-project/src/model.rs` |
-| what an edit does | `concat-project/src/commands/` |
-| where a clip may land | `concat-project/src/placement.rs` |
-| the file format | `concat-project/src/doc.rs` |
-| how a frame is planned | `concat-render/src/plan.rs` |
-| how it is drawn | `concat-render/src/gpu.rs`, `compositor.rs` |
-| an effect | `concat-effects/packages/<id>/`; the contract in `src/shader.rs` |
-| decoding, the cache | `concat-media/src/decode.rs`, `pool.rs`, `prefetch.rs`, `hardware.rs` |
-| the sound | `concat-media/src/audio.rs` (export), `concat-host/src/playback.rs` (live) |
-| the export loop | `concat-export/src/lib.rs` (`render_on`), `resolve.rs` |
-| the monitor | `concat-host/src/preview.rs`, `concat/src/panes/monitor.rs` |
-| a sheet or a pane | `concat/src/panes/<name>.rs` and `concat/ui/` |
-| the gestures, the clipboard | `concat/src/studio.rs` |
-| threads and dialogs in the window | `concat/src/host.rs` |
-| a platform difference | `concat/src/platform.rs` |
-| the API's verbs | `concat-api/src/message.rs`, `lib.rs` |
-| the server | `concat-server/src/lib.rs`, `json.rs`, `grpc.rs`, `token.rs`, `hub.rs` |
-| a model's source and digest | `models/manifest.toml`, `concat-host/src/models.rs` |
-| the release | `.github/workflows/release.yml`, `build-app.yml`, `mobile.yml` |
-
-## 15. Where we stand today
-
-The audits in `audit/` (23 and 28 September, 4, 9 and 11 October) score
-every crate and feature and name each finding with a file and a line;
-`audit/AUDIT-2026-10-11.md` is the current one and this section is its
-one-page shape. Whole-app scores on 11 October: architecture 8, user
-likeability 7, code quality 6, design philosophy 6, performance 6, testing
-6, documentation 6, maintainability 5, scalability 5, security 4, CI and
-release 4.
-
-**What is solid and should stay that way.** The crate boundaries hold:
-one compositor, one plan, one placement rule, one definition of sound,
-one code path that writes the document, one scheduler. Rational time with
-checked rates. Hardware decode that falls back at every stage. Model and
-update downloads verified by digest. Atomic saves and, since 9 October,
-atomic exports with sound. A document that cannot panic the GPU
-(`MAX_SIDE`, `MAX_TIME`, uncaptured-error handlers). Constant-time tokens
-and bounded reads on the socket. Every fix of the last fortnight landed
-with a test that fails without it.
-
-**What is unfinished or wrong, by layer.**
-
-- *The window.* `studio.rs` is 10 689 lines and one `impl` on a struct of
- about a hundred fields; the gestures, the clipboard and the stage still
- live on the controller, and the panes have moved out one at a time
- (eleven so far). 73 tests cover 23 000 lines of window Rust. Package
- trials, "Relink all", the cache clear and the proxy sweep still run on
- the event-loop thread. About fifteen user-visible strings bypass i18n.
- Two `accessible-*` properties in 29 790 lines of Slint; no keyboard
- reach into dialogs beyond Escape on the confirm sheet.
-- *The document.* `DOCUMENT_VERSION` is still 1, so a shape clip written
- by 0.2.6 opens as a video clip in 0.2.5 and the kind is lost on save.
- `SetClipSpeedCurve` can lengthen a clip over its neighbour, the one
- command that bypasses the placement rule. Every undo snapshot clones the
- media bin.
-- *The engine.* The export is serial with a readback per frame; the
- mixer scans every clip per sample; pinned prefetch frames sit outside
- the pool's byte budget. The dissolve's hold plays keyed effects early.
- The reader pool caches by `DefaultHasher` names for proxies. HDR side
- data is sized by a Rust mirror of structs FFmpeg says are not ABI.
-- *The API and server.* Nine or more window operations have no verb
- (voiceover, meters, the font picker, presets, click-to-preview, copy and
- paste, the overwrite confirmation). Every JSON connection leaks a file
- descriptor; seats are taken before auth; a caller that stops reading
- pins its seat; `Server::stop` blocks the window until a running API
- export ends; API reads are unconfined and caller strings reach FFmpeg's
- opener with no protocol whitelist; gRPC has none of the JSON caps; the
- embedded API opens a second GPU device.
-- *Trust.* Self-update trusts the digest beside the package and strips
- quarantine; nothing is signed. FFmpeg, sherpa-onnx and ONNX Runtime are
- fetched in CI from moving releases with no digest; actions are pinned by
- tag; the Windows installers are unsigned and the Mac builds ad-hoc
- signed. Clear cache deletes reversed and enhanced media.
-- *CI.* The Windows test job has not passed since 4 October: the hanging
- cards test now skips on WARP, and the job still fails inside `cargo
- test` for a reason the public logs do not show. `scripts/locales.py`
- crashes on a Windows console (cp1252) unless Python runs with `-X utf8`.
-- *Phones.* iOS cannot create a project (the default location is a
- Desktop folder the sandbox refuses, #299) and the Choose button is gone;
- Android's export encoders and media-store publishing have not been seen
- on real hardware; the Android bridge has three ways to lose an import.
-- *Users are asking for* (open issues, 89): exports that survive an 8 GB
- GPU running out of memory (#297, #223), frame-step buttons and pinch
- zoom on phones, shape masks back (#239), SRT import and export, audio
- ducking, beat detection, a crop tool with free aspect, a command
- palette, a Shortcuts tab (PR #300), an activity panel for long jobs,
- Flatpak, and the Chatterbox DirectML failure (#291).
-- *Docs.* `src/README.md`'s crate table omits concat-text and concat-vision. A few comments still describe the CPU compositor fallback
- that no longer exists (`concat-export/src/lib.rs`, `concat-host/
- Cargo.toml`, two sites in `gpu.rs`), and `src/Cargo.toml` says wgpu is
- pinned to Slint 1.17's.
-
-**The order to take it in** is the audit's section 8: the Windows CI job
-first, so that a tag can pass its gate; then the server's seats, fds and
-`stop`; then the document version for shapes; then the window's thread
-hygiene and i18n; then signing. After those, phase 4 of the HDR plan, a
-true HDR preview.
diff --git a/audit/AUDIT-2026-09-23-QUICK.md b/audit/AUDIT-2026-09-23-QUICK.md
deleted file mode 100644
index c5bb97b5..00000000
--- a/audit/AUDIT-2026-09-23-QUICK.md
+++ /dev/null
@@ -1,129 +0,0 @@
-# Concat audit, 23 Sep 2026 — the quick version
-
-Same audit as `AUDIT-2026-09-23.md`, one screen. Each bar has 10 slots:
-🟢 filled = good, ⚪ empty = missing. Severity: 🔴 critical · 🟠 high · 🟡 medium · ⚪ low.
-
-## 🎯 Verdict
-
-**Structure: done. Guarantees: not yet.** The 16 Sep plan landed in full. What is left is rules the docs promise and the code does not hold yet: bounds, checks, one-place enforcement. Cheaper to fix than what came before.
-
-## 📊 The whole app
-
-| Aspect | Bar | Score |
-|---|---|:-:|
-| 🧹 Code quality | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 |
-| 🏛️ Architecture | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 |
-| 🧭 Design philosophy | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 |
-| 🔧 Maintainability | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 |
-| 😊 User likeability | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 |
-| ⚡ Performance | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 |
-| 📈 Scalability | 🟢🟢🟢🟢⚪⚪⚪⚪⚪⚪ | 4 |
-| 🔐 Security | 🟢🟢🟢🟢⚪⚪⚪⚪⚪⚪ | 4 |
-| 🧪 Testing | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 |
-| 📚 Docs | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 |
-| 🤖 CI and release | 🟢🟢🟢🟢⚪⚪⚪⚪⚪⚪ | 4 |
-
-## 🧱 Per crate (average of its six scales)
-
-| Crate | Bar | Avg | One thing |
-|---|---|:-:|---|
-| concat-core | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7.3 | 🟢 no deps, rational time · ⚪ linear clip lookup |
-| concat-vision | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.2 | 🟢 ORT fully wrapped · ⚪ serial batch-1 |
-| concat-effects | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.2 | 🟢 strict manifests · ⚪ no host-version or licence field |
-| concat-text | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.0 | 🟢 zero engine deps · ⚪ copies the font per render |
-| concat-api | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.6 | 🟢 every verb JSON, docs match · 🔴 any path, any size |
-| concat-server | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.2 | 🟢 constant-time token · 🔴 no limits at all |
-| concat-media | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.7 | 🟢 unsafe is sound · 🔴 a panicking job wedges the scheduler |
-| concat-project | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 | 🟢 commands only · 🟠 ripple copies every clip |
-| concat-speech | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.2 | 🟢 whisper on Metal · 🔴 downloads unverified |
-| concat-export | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.8 | 🟢 one composite path · ⚪ 2 403-line lib.rs |
-| concat-host | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.7 | 🟢 small modules · 🔴 corrupt doc opens empty |
-| concat (Rust) | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.7 | 🟢 panes + echo-and-commit · 🔴 results land in the wrong project |
-| concat (Slint) | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.8 | 🟢 one Modal pattern · ⚪ accessibility 2/10 |
-| concat-render | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5.3 | 🟢 FramePlan · 🟠 7 CPU kernels for 147 shaders |
-
-## 🎛️ Per feature
-
-| Feature | Bar | Score | State |
-|---|---|:-:|---|
-| Launcher | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | ✅ |
-| Import and bin | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | ✅ |
-| Timeline editing | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | ✅ |
-| Undo / redo | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | ✅ sharing lost on ripple |
-| Relink | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | ✅ |
-| Themes | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | ✅ |
-| Effects and filters | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ⚠️ CPU path never pixel-compared |
-| Titles | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ CPU-heavy |
-| Keyframes | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ |
-| Cutout | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ |
-| Captions | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ |
-| Text-to-speech | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ |
-| Templates | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ |
-| LUTs | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ |
-| Export | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ serial loop |
-| Hardware decode | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | ✅ not zero-copy |
-| Transitions | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | ⚠️ fades and wipes missing from preview |
-| Crop, blend, masks | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | ⚠️ Darken/Lighten wrong at partial opacity |
-| Enhance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | ⚠️ one model, no native path |
-| Playback and audio | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | ⚠️ 400 decodes per speed-curved clip |
-| Remote API | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | ⚠️ unsafe defaults, window not a client |
-| Languages | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | ⚠️ 13 of 14 at ~71 %, no RTL |
-| Custom packages | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | ⚠️ trial never runs, multi-pass unwired |
-| Proxies | 🟢🟢🟢🟢⚪⚪⚪⚪⚪⚪ | 4 | ❌ no UI, never swept |
-| Accessibility | 🟢🟢⚪⚪⚪⚪⚪⚪⚪⚪ | 2 | ❌ two accessible-* properties |
-
-## 🚨 Top findings
-
-| # | Sev | What | Where |
-|:-:|:-:|---|---|
-| 0 | 🔴 | **`main` is red**: CI fails at fmt (8 files), 6 tests fail underneath | `ci.yml`, §1.2 of the full report |
-| 1 | 🔴 | Corrupt project opens empty, next save overwrites it | `concat-host/src/session.rs:95` |
-| 2 | 🔴 | Panicking scheduler job wedges the scheduler for good | `concat-media/src/prefetch.rs:355` |
-| 3 | 🔴 | Worker results land in whichever project is open | `concat/src/studio.rs:5376` |
-| 4 | 🔴 | API writes any path; input sizes and line lengths unbounded | `concat-api/src/lib.rs:499,707` · `concat-server/src/json.rs:117` |
-| 5 | 🔴 | Window and Remote page are two engines on one folder | `concat/src/panes/settings.rs:403` |
-| 6 | 🔴 | 20 of 23 model downloads have no digest; 18 URLs unpinned | `models/manifest.toml` |
-| 7 | 🟠 | Hostile-package trial never runs; loop check trivial; no error scope | `concat-render/src/gpu.rs:850,1091` |
-| 8 | 🟠 | Ripple, freeze, fill-slot deep-copy every clip into undo | `concat-project/src/model.rs:1966` |
-| 9 | 🟠 | Stream start time ignored (MTS files offset) | `concat-media/src/decode.rs:743` |
-| 10 | 🟠 | Darken / Lighten maths wrong at partial opacity | `concat-render/src/compositor.rs:258` |
-| 11 | 🟠 | Art scan on every pointer event; full publish 30×/s | `concat/src/lib.rs:373` |
-| 12 | 🟠 | CI is Linux-only until a tag; perf gate never runs; parity skips | `.github/workflows/ci.yml` |
-| 13 | 🟡 | Clamps in four places, they disagree | `concat-project/src/commands/properties.rs:28` |
-| 14 | 🟡 | E2E suite checks timing, not looks | `concat-host/tests/export.rs:915` |
-| 15 | 🟡 | Caches with no bound: proxy, artwork, pinned frames, GPU passes | `proxy.rs`, `prefetch.rs:243`, `gpu.rs:1453` |
-| 16 | 🟡 | 3–5 copies per frame; filter graph rebuilt per treated frame | `decode.rs:934`, `treat.rs:85` |
-| 17 | 🟡 | Hub thread does heavy work in line; gRPC fan-out under a mutex | `hub.rs:52,71` |
-| 18 | 🟡 | ONNX Runtime likely linked twice; pinned to an rc | `src/Cargo.toml:41` |
-| 19 | 🟡 | Speech only reachable from the window | `crates/concat/Cargo.toml` |
-| 20 | 🟡 | Titles copy the whole font per render | `concat-text/src/lib.rs:268` |
-| 21 | ⚪ | Wrong H shortcut, ⌘ on every OS, "twelve" languages, dark colours in light theme | `settings.slint:671`, `tray.slint:127` |
-| 22 | ⚪ | Dead `ui/demo/` still imported | `app.slint:21` |
-| 23 | ⚪ | Stale docs: 429 tests (544), crate map, §7 diagram, mobile.yml, CONTRIBUTING | `ARCHITECTURE.md:276` |
-| 24 | ⚪ | 8 files unformatted; dead `core::Project`; disk IO in a wasm crate | `timeline.rs:500`, `model.rs:1896` |
-
-## 🧾 Today's checks
-
-| Check | Result |
-|---|---|
-| fmt | ⚠️ 8 files would change |
-| clippy | ✅ 0 warnings |
-| tests | ❌ 537 passed, **6 failed**, 1 ignored (3 window, 2 render transitions, 1 Chatterbox) |
-| perf --check | ✅ 14 of 14 within budget (GPU composite 3.2 ms, export 115 fps, hardware decode 326 fps) |
-
-## 🗺️ Do next, in order
-
-0. 🔴 **Make `main` green** — `cargo fmt --all`, then fix the 6 failing tests (5 are today's commits, 1 is locale coverage).
-1. 🔴 **Correctness trio** — not-found vs parse error in `Session::open`; `catch_unwind` around scheduler jobs; a session epoch on worker results.
-2. 🔴 **Server limits + root allowlist** — line cap, pre-auth timeout, connection cap, bounded outbox and sizes; share the window's exporter and sessions.
-3. 🔴 **Verify every model** — fill the digests, pin to commits, reject empty digests in release, generate the tables from the manifest.
-4. 🟠 **Refuse hostile packages for real** — trial at install, error scopes, binding-type check, pipelines keyed by source hash.
-5. 🟠 **Document maths** — `clips_where` in place of `clips_mut`; one clamp pass; `start_time`; fix Darken and Lighten.
-6. 🟡 **Stop the per-event work** — art refresh on change only; targeted publish; cache treat graphs; `av_frame_ref`; sweep proxy and artwork.
-7. 🟡 **CI** — macOS and Windows check; lavapipe with parity required; `perf --check`; release depends on CI; fix the stale docs.
-
-Then the ARCHITECTURE §10 list in its order: plan filled by the export → source textures pooled → zero-copy → one texture per track → native Enhance → multi-pass packages.
-
-## ✅ Keep
-
-Layering · `FramePlan` · commands-only document with Arc undo · hardware decode fallback · package load hygiene · JSON API contract and its docs · pane pattern · model staging and licences · tests that check behaviour.
diff --git a/audit/AUDIT-2026-09-23.md b/audit/AUDIT-2026-09-23.md
deleted file mode 100644
index 5bf7b32c..00000000
--- a/audit/AUDIT-2026-09-23.md
+++ /dev/null
@@ -1,467 +0,0 @@
-# Concat audit, 23 September 2026
-
-Where the codebase stands one week after the 16 September proposal landed.
-Every crate, every feature and the engineering process were read and scored
-on the same seven scales, plus security, testing and documentation where they
-turned out to matter. Every finding names a file and a line. The sibling
-`AUDIT-2026-09-23-QUICK.md` is the same audit as a one-screen scorecard.
-
-**How it was done.** Nine reviewers each read one part of the tree in full
-(the crates, the window's Rust, the Slint UI, the API and transports, the
-model crates, and the tests, CI, build and docs), scored it and reported
-findings with locations. The headline findings were then re-verified by hand
-against the code before they went in here. Line counts, test counts and the
-unsafe, unwrap and panic tallies come from the tree; `cargo fmt --check`,
-`cargo clippy --workspace --all-targets`, `cargo test --workspace` and
-`cargo run -p concat-perf --release -- --check` were run on this machine
-(Apple Silicon, macOS 26). The app itself was not launched: "user
-likeability" is judged from what the UI code does, not from users.
-
-**Scale.** 0 to 10. 8 and up means "keep doing this"; 5 to 7 means "works,
-with known debt"; 4 and under means "will bite at the next order of
-magnitude, or already bites".
-
----
-
-## 1. The numbers
-
-### 1.1 Size, tests and hygiene per crate
-
-| Crate | Rust lines | Slint / WGSL | Tests (files) | `unsafe` | `unwrap`/`expect` | `panic!` family |
-|---|---:|---:|---:|---:|---:|---:|
-| concat (window) | 16 058 | 23 636 slint | 40 (11) | 1 | 15 | 1 |
-| concat-project | 9 911 | | 100 (3) | 0 | 312 | 12 |
-| concat-host | 8 727 | | 53 (14) | 1 | 152 | 12 |
-| concat-media | 8 038 | | 79 (11) | 30 | 148 | 2 |
-| concat-render | 4 873 | | 62 (6) | 1 | 39 | 3 |
-| concat-effects | 4 664 | 2 205 wgsl | 46 (7) | 0 | 76 | 2 |
-| concat-export | 3 286 | | 25 (3) | 0 | 17 | 0 |
-| concat-speech | 3 087 | | 19 (3) | 3 | 23 | 0 |
-| concat-vision | 2 760 | | 31 (8) | 0 | 22 | 0 |
-| concat-core | 2 691 | | 45 (7) | 0 | 22 | 2 |
-| concat-api | 1 993 | | 18 (2) | 0 | 19 | 8 |
-| concat-server | 1 183 | | 10 (4) | 0 | 27 | 2 |
-| concat-text | 1 101 | | 11 (1) | 0 | 28 | 1 |
-| concat-perf | 804 | | 2 (1) | 0 | 32 | 0 |
-| concat-cli | 449 | | 3 (1) | 0 | 2 | 0 |
-| concat-android | 335 | | 0 (0) | 10 | 9 | 2 |
-| **Total** | **~71 000** | **~25 800** | **544 (82)** | **46** | **~940** | **47** |
-
-The unwrap and expect counts include test code (concat-project's 312 are
-nearly all in its 3 692-line test module). Outside tests, every crate but the
-window and the host is close to panic-free, and every `unsafe` block in
-concat-media carries a SAFETY note and was found sound.
-
-Other facts: 173 effect packages (147 with WGSL, 20 transitions, 25 audio);
-14 locale files; 606 commits since 25 August, 451 of them this month; 26
-release tags, latest `v0.2.3`; largest files `concat/src/studio.rs` 7 484,
-`concat-export/src/lib.rs` 2 403, `concat-render/src/gpu.rs` 1 794,
-`concat-media/src/decode.rs` 1 603; largest Slint files
-`workspace/media-pane.slint` 2 303 and `inspector/clip-inspector.slint` 2 131.
-
-### 1.2 The checks, run today
-
-| Check | Result |
-|---|---|
-| `cargo fmt --all -- --check` | 8 files would change |
-| `cargo clippy --workspace --all-targets` | 0 warnings |
-| `cargo test --workspace --no-fail-fast` | 537 passed, 6 failed, 1 ignored |
-| `cargo run -p concat-perf --release -- --check` | 14 of 14 scenarios within budget |
-
-**`main` is red.** The last four CI runs on `main` failed at the
-`cargo fmt --check` step, so CI has not run the test suite for today's
-commits. Run locally, six tests fail, five of them in files changed today:
-
-| Test | Crate | What it says |
-|---|---|---|
-| `transitions::tests::a_slide_arrives_from_the_far_side` | concat-render | "the outgoing picture dropped away": red where blue was expected (`transitions.rs:546`) |
-| `transitions::tests::every_shape_runs_from_one_picture_to_the_other` | concat-render | `smoothleft` at progress 1 is a blend, not the incoming picture (`transitions.rs:436`) |
-| `chatterbox::tests::the_text_is_normalised_the_way_resemble_does_it` | concat-speech | a double space survives after a comma (`chatterbox.rs:1100`) |
-| `panes::settings::listen_tests::a_missing_or_bad_half_gets_the_default` | concat | `"192.168.1.9:"` keeps its trailing colon as the host (`settings.rs:753`) |
-| `format::tests::zooming_in_reveals_the_fine_buckets_and_never_loses_a_peak` | concat | every column carries the spike, not one (`format.rs:448`) |
-| `i18n::tests::every_shipped_locale_parses_names_itself_and_keys_off_the_inventory` | concat | 13 locales lack about 290 of the 1 015 keys (`i18n.rs:254`) |
-
-The first five look like today's work outrunning its tests; the last is a
-rule (every locale covers the inventory) the locale files have not met since
-the inventory grew. The end-to-end export suite, the parity fixtures and the
-migration tests all pass.
-
-**The perf table**, release build, every scenario inside its budget:
-
-| Scenario | Measured | Budget |
-|---|---:|---:|
-| plan one frame of 200 clips on 4 tracks | 0.53 µs | ≤ 200 µs |
-| 200 clip additions and 200 undos | 0.41 ms | ≤ 150 ms |
-| ripple delete of 200 clips out of 400 | 0.14 ms | ≤ 20 ms |
-| save and open a 200-clip document | 1.49 ms | ≤ 60 ms |
-| choose the visible clips of 200 for one publish | 0.20 µs | ≤ 50 µs |
-| read an hour's waveform across 2048 columns | 4.33 µs | ≤ 2000 µs |
-| decode 1080p h264 in software | 1 637 fps | ≥ 120 |
-| decode 1080p h264 on the hardware (VideoToolbox) | 326 fps | ≥ 120 |
-| scrub forward over 90 frames, cold | 0.81 ms/frame | ≤ 25 |
-| scrub back over covered ground | 0.24 µs/frame | ≤ 200 |
-| decodes when an effect knob is turned | 0 | ≤ 0 |
-| compose 1080p, 3 layers, sepia, CPU | 54.7 ms | ≤ 400 |
-| compose 1080p, 3 layers, sepia, GPU | 3.19 ms | ≤ 25 |
-| export a 4 s 720p cut of two overlapping clips | 115 fps | ≥ 30 |
-
-Two readings worth keeping in mind: hardware decode is five times slower
-than software here, because the frame is downloaded and converted on the CPU
-after the chip decodes it (finding 16 and ARCHITECTURE §10), and the CPU
-composite is seventeen times the GPU one, which is what a machine without a
-GPU, and any export whose device dies mid-run, will see.
-
----
-
-## 2. Scorecard for the whole app
-
-| Aspect | Score | One line |
-|---|:---:|---|
-| Code quality | 7 | Clippy-clean, typed errors, sound and documented `unsafe`; long functions and copied blocks pull it down. |
-| Design architecture | 7 | The crate layering holds and FFmpeg and ONNX Runtime are contained; the window bypasses `Session` and the API, and chains bypass `FramePlan`. |
-| Design philosophy | 6 | The ideas are clear and written down. Several are stated and not enforced: one clamp place, hostile packages refused, publish by diff, stale results dropped in one place. |
-| Maintainability | 6 | Excellent comments and one shape per pane; four files over 1 800 lines and model tables kept in five places. |
-| User likeability | 7 | A broad CapCut-class feature set with real polish; accessibility is near zero and a few UX facts are stale. |
-| Performance | 5 | The planner and undo are microseconds; the per-frame copies, the UI's per-event art scan and the serial export loop are not. |
-| Scalability | 4 | Linear clip lookups, undo that copies every clip on a ripple, caches with no bound, one server thread, unbounded socket input. |
-| Security | 4 | The token is done right. A token holder can write any path; input sizes are unbounded; 20 of 23 model downloads have no digest. |
-| Testing | 6 | 544 behaviour tests and a real end-to-end suite; it checks timing, not looks, and the GPU parity suite silently skips in CI. |
-| Documentation | 7 | Substantial and well written, with a handful of stale claims. |
-| CI and release | 4 | `main` is red at the fmt step; Linux only until a tag; the perf gate the docs describe does not run. |
-
----
-
-## 3. Scores per crate
-
-Columns: code quality, maintainability, scalability, performance,
-architecture, design philosophy. A dash means the scale does not apply.
-
-| Crate | CQ | Maint | Scale | Perf | Arch | Phil | Note |
-|---|:-:|:-:|:-:|:-:|:-:|:-:|---|
-| concat-core | 8 | 8 | 6 | 6 | 8 | 8 | No deps, rational everywhere; linear `clip_on_track_at`, gcd on every add. |
-| concat-project | 7 | 6 | 5 | 5 | 7 | 6 | Commands are the one way to write; clamps live in four places; ripple copies every clip. |
-| concat-media | 7 | 7 | 5 | 6 | 8 | 7 | Unsafe is sound; 3 to 5 copies per frame; scheduler cannot survive a panicking job. |
-| concat-render | 7 | 5 | 4 | 5 | 6 | 5 | `FramePlan` is excellent; 7 CPU kernels for 147 shaders; GPU memory unbounded at 8K. |
-| concat-effects | 8 | 8 | 6 | 7 | 7 | 7 | Strict manifests and a shared WGSL prelude; format lacks host version, capabilities, licence. |
-| concat-export | 6 | 4 | 6 | 5 | 7 | 7 | One composite path for export and preview; a 2 403-line lib.rs and 12 side tables per clip. |
-| concat-host | 6 | 6 | 5 | 6 | 5 | 6 | Small clear modules; `Session::open` swallows parse errors; proxy and artwork caches never swept. |
-| concat (Rust) | 7 | 5 | 5 | 5 | 6 | 6 | Panes and echo-and-commit are sound; 67 callbacks skip the `Msg` loop; art scan per event. |
-| concat (Slint) | 7 | 6 | – | – | 7 | 7 | One `Modal` pattern and one token file; accessibility 2, i18n 6. |
-| concat-api | 8 | 8 | 5 | – | 8 | 9 | Every verb JSON, docs match; security 3: any path, any size. |
-| concat-server | 7 | 8 | 4 | – | 8 | 9 | Constant-time token, loopback default; no limits, one Hub thread; security 5. |
-| concat-cli | 7 | 8 | – | – | 7 | – | `serve` parks forever, so a Unix socket file is left behind. |
-| concat-android | 8 | 7 | – | – | – | – | Justified `unsafe`; no tests. |
-| concat-speech | 7 | 5 | 6 | 7 | 5 | 7 | Whisper on Metal; models listed in five places; only the window can reach it; security 3. |
-| concat-vision | 8 | 7 | 5 | 6 | 9 | 8 | ORT fully wrapped; batch-1 serial inference under one lock; security 6. |
-| concat-text | 8 | 8 | 5 | 4 | 9 | 8 | No engine deps; copies the whole font file per render, draws on a frame-sized canvas. |
-
----
-
-## 4. Scores per feature
-
-Status: **done**, **partial** (works with a known gap), **thin** (present, weak).
-
-| Feature | Where | Status | Score | Why |
-|---|---|---|:-:|---|
-| Launcher, new project, recents | `ui/start.slint` | done | 8 | Three shapes by width; the one screen with a phone layout. |
-| Import, bin, drag and drop | `ui/workspace/media-pane.slint` | done | 8 | Filters, sort, probe on a worker. |
-| Timeline editing | `ui/timeline/`, `panes/timeline.rs` | done | 8 | Split, ripple, snap, several timelines, virtualised clips. |
-| Undo and redo | `concat-project/src/editor.rs` | done | 8 | Arc-shared snapshots, gestures as one step; sharing lost on ripple (finding 8). |
-| Transitions | `concat-export/src/lib.rs:467` | partial | 6 | Fades and wipes baked into chains; absent from the preview. |
-| Effects and filters | `concat-effects/packages/` | partial | 7 | 173 packages; CPU export renders them through FFmpeg chains never pixel-compared with the WGSL. |
-| Titles and text | `concat-text` | done | 7 | Correct; CPU-heavy per animated frame. |
-| Keyframes, animations | `concat-project/src/model.rs:1573` | done | 7 | Keys survive split, trim and merge; offset and rotation keys unclamped. |
-| Crop, transform, blend, masks | `concat-render/src/plan.rs` | partial | 6 | Darken and Lighten wrong at partial opacity on both paths (finding 10). |
-| Cutout, smart brush | `concat-vision/src/segment.rs` | done | 7 | Compiled-in fallback model; serial batch-1 inference. |
-| Enhance | `concat-host/src/enhance.rs` | thin | 6 | One Real-ESRGAN model on every platform; native scalers not used. |
-| Auto-captions | `ui/dialogs/captions.slint` | done | 7 | Whisper on Metal, with progress; results can land in the wrong project (finding 3). |
-| Text-to-speech, voice from a sample | `concat-speech/src/tts.rs` | done | 7 | Kokoro, Pocket, Chatterbox; downloads unverified (finding 6). |
-| Templates | `concat-host/src/templates.rs` | done | 7 | |
-| LUT import | `concat-effects/src/cube.rs` | done | 7 | Tables over 65 a side refused. |
-| Custom and community packages | `concat-effects/src/catalogue.rs` | thin | 5 | Multi-pass parsed and never run; the trial never runs before enabling (finding 7). |
-| Proxies | `concat-host/src/proxy.rs` | thin | 4 | No UI, no sweep of `cache/proxy`. |
-| Export | `ui/dialogs/export.slint`, `concat-export` | done | 7 | Quality tiers, 10-bit, CBR and VBR; serial loop; one social preset. |
-| Playback and audio | `concat-host/src/playback.rs` | partial | 6 | A speed-curved clip becomes up to 400 decodes; the mixer scans every clip per sample. |
-| Hardware decode | `concat-media/src/hardware.rs` | done | 7 | Sound, falls back once; frames still round-trip through memory. |
-| Remote API | `concat-api`, `concat-server` | partial | 6 | Clean contract; unsafe defaults (finding 4); the window does not use it (finding 5). |
-| Relink missing media | `ui/dialogs/relink.slint` | done | 8 | |
-| Themes and accent colours | `ui/theme/` | done | 8 | Two lane colours hard-coded to dark. |
-| Languages | `concat/locales/` | partial | 6 | 14 files, 13 at about 71 % of the 1 015 keys; Persian ships without right-to-left. |
-| Accessibility | the whole UI | thin | 2 | Two `accessible-*` properties in 23 636 lines; 12 px base type. |
-
----
-
-## 5. Findings, by severity
-
-Each carries the location and, where I re-verified it by hand, **(verified)**.
-
-### Critical
-
-1. **A corrupt document opens as an empty project, and the next save
- overwrites it.** (verified) `Session::open` matches `Err(_)` from
- `projects::read_document` and builds an empty editor
- (`concat-host/src/session.rs:95`); that function returns the same `Err` for
- "file not found" and for a JSON parse failure (`projects.rs:211-215`). The
- comment nine lines above says a corrupt document is an error. Saves are
- atomic with fsync (`projects.rs:168-190`), so the window is narrow: a power
- cut, a sync tool or an external edit. When it happens it is silent data
- loss.
-
-2. **A panicking scheduler job wedges the scheduler.** (verified)
- `prefetch.rs:355` runs `job()` with no unwind guard. A panic skips the
- decrements at 357-363, `running` never drops, `drain` waits forever
- (`prefetch.rs:307`) and background lanes stay blocked once the counters sit
- at their limit. Failed thread spawns are also ignored (`prefetch.rs:194`).
-
-3. **Worker results land in whichever project is open.** `Studio::handle`
- (`concat/src/studio.rs:5376-5429`) checks no project identity, and
- `close_project` (`studio.rs:5308`) does not reset the export, captions or
- speech panes. A probe, a caption run or a speech job started in project A
- delivers into project B (`panes/media_bin.rs:153`, `panes/captions.rs:132`,
- `panes/speech.rs:339`). ARCHITECTURE §5 claims the opposite.
-
-4. **The API lets a token holder write anywhere and send anything.**
- (verified) `preview.frame` creates folders and truncates any path
- (`concat-api/src/lib.rs:707-714`); `export.run`, `project.create` and
- `template.instantiate` take any path. Width and height are only checked
- for zero (`lib.rs:499-502`), so 65535×65535 asks for 17 GB on the Hub
- thread. The JSON transport reads lines of unbounded length before
- authentication, spawns two threads per connection with no cap and no
- timeout, and queues output in an unbounded channel
- (`concat-server/src/json.rs:63,117,193`). The token itself is right:
- 128 bits, constant-time compare, required on loopback.
-
-5. **The window and the Remote page are two engines on one folder.**
- (verified) Settings › Remote starts `Server::start(config, Api::new)`
- (`concat/src/panes/settings.rs:403`): its own session registry and its own
- exporter. An edit from either side is overwritten by whoever saves last,
- and "one export at a time" no longer holds across the two.
- `Api::exporter()` exists for sharing and nothing calls it. The §7 diagram
- showing the window over the API is wrong.
-
-6. **Model downloads are unverified and unpinned.** (verified) 20 of 23 rows
- in `models/manifest.toml` have an empty `sha256`; `verify` accepts an empty
- digest (`concat-host/src/models.rs:289-291`); 18 upstream URLs point at
- `resolve/main`. The source list includes a mirror and a user-set base URL
- (`models.rs:149-180`), and a dropped transfer resumes with a Range request
- against the *next* source (`models.rs:216-219`), relying on the digest to
- catch a splice.
-
-### High
-
-7. **A hostile package is not really refused.** (verified that `trial` has
- no production caller) `WgpuCompositor::trial` is only called from tests.
- The loop check asks only whether a `break` token exists
- (`concat-effects/src/shader.rs:507-549`); binding types are never checked;
- validation uses `Capabilities::all()` (`shader.rs:614`). A pipeline that
- fails to build does so outside any error scope (`gpu.rs:1091`), which in
- wgpu is a panic.
-
-8. **Undo's structural sharing is defeated by every ripple.** (verified)
- `clips_mut()` calls `Arc::make_mut` on every clip before the caller filters
- (`concat-project/src/model.rs:1966`); ripple trim, ripple delete, freeze
- and fill-slot use it (`commands/clips.rs:245,468,596,653`,
- `commands/media.rs:105`). The only sharing test covers `UpdateClip`.
-
-9. **The stream's start time is ignored.** (verified: no reference in the
- crate) Timestamps are used raw (`decode.rs:743`) and turned straight into
- frame indexes (`pool.rs:364`). MTS and other camera files that start at a
- non-zero time will be offset in the pool, the paced export and the
- duration clamp. Inferred from the code, not reproduced.
-
-10. **Darken and Lighten are wrong at partial opacity, on both paths.**
- (verified) Both compute `min(colour·α, ground)`
- (`compositor.rs:258-259`, `gpu.rs:421-430`): a white Darken layer at 30 %
- turns a mid-grey ground dark. Parity cannot catch it because both agree.
-
-11. **The UI thread scans every media item on every pointer event.**
- (verified) The `handler!` macro calls `refresh_art()` after each callback
- (`concat/src/lib.rs:373`), including the lanes' hover and drag stream.
- That loops over all media calling `proxy::ensure` (two `stat` calls each,
- `studio.rs:2010-2030`) and decodes cached JPEGs synchronously. Separately,
- playback runs a full `publish_chrome` about 30 times a second
- (`host.rs:197`, `studio.rs:1905`), rebuilding every bin row, both menus and
- the recents list each time.
-
-12. **`main` is red, and CI never compiles macOS or Windows until a tag.**
- Today's four CI runs on `main` fail at `cargo fmt --check` (8 files),
- which hides six failing tests behind it (§1.2). `build-app.yml` is
- `workflow_call` only; `ci.yml` runs on `ubuntu-latest`; `release.yml`
- does not require CI to pass. The GPU parity suite returns early with no
- adapter (`gpu/tests.rs:20-23`) and the runner installs none, so the
- parity row in ARCHITECTURE §8 is enforced only on developers' machines.
- No workflow runs `concat-perf --check`; 9 of its 13 scenarios never run
- in CI.
-
-### Medium
-
-13. **Clamps live in four places and disagree.** Commands re-implement
- `Clip::tidy` (`commands/properties.rs:28-37,340-355`); `TextStyle::tidy`
- runs only on load (`model.rs:1418`); the minimum transition length is
- applied only in tidy; `SetClipKey` leaves offset and rotation unclamped
- (`properties.rs:198`). A document can change on save and reopen. The
- speed range is written three times (`model.rs:1090`,
- `concat-core/src/retime.rs:27`, `concat-media/src/audio.rs:35`).
-
-14. **The end-to-end suite checks timing, not looks.** It samples the centre
- pixel and maps colour to a source second (`tests/export.rs:305-323`),
- which proves cuts, trims, speed and reverse. Transitions, effects, keys,
- transform, crop, blend, fades, titles and adjustment layers assert only
- length and sound (`export.rs:915-1064`).
-
-15. **Caches with no bound.** `cache/proxy` and artwork are never swept
- (`proxy.rs:16` says otherwise; `media.rs:323`); frames pinned by the
- prefetcher are neither deduplicated nor capped and `Monitor::prefetch`
- keeps adding while paused (`prefetch.rs:243,266`, `preview.rs:255`); the
- 8 warm readers each decode with every core (`pool.rs:531`,
- `decode.rs:651`) outside the byte budget; on the GPU, `run_passes` claims a
- fresh full-size target per pass (`gpu.rs:1453`, about 133 MB each at 8K)
- and the LUT and reveal caches are never evicted (`gpu.rs:198-201`).
-
-16. **Per-frame copies and graph rebuilds.** Every paced frame is deep-copied
- before filtering (`decode.rs:934,944,994`), then filtered, then copied row
- by row into a `Frame` (`decode.rs:873`); a hardware frame adds a download.
- On the monitor path every treated frame builds, parses and validates a
- fresh filter graph (`treat.rs:85-126`).
-
-17. **The Hub thread does heavy work in line, and event fan-out holds a
- mutex.** Preview decode and composite, probes and title painting run on
- the one thread that serves every caller (`hub.rs:71`, `lib.rs:505,396`).
- The gRPC subscriber does a `blocking_send` into a channel of 64 while the
- subscribers mutex is held (`grpc.rs:70-74`, `hub.rs:52`): one slow client
- stalls the export thread.
-
-18. **ONNX Runtime is probably linked twice** (sherpa-onnx's static copy plus
- `ort` 2.0.0-rc.13 through vision and Chatterbox), and CoreML is registered
- with default options and an aspect-dependent input shape
- (`concat-vision/src/runtime.rs:141`, `segment.rs:115`). Needs profiling.
-
-19. **Speech is reachable only from the window.** Only `concat` depends on
- `concat-speech`, so captions and TTS have no API, CLI or server verb;
- "three ways in" does not hold for them. Also missing verbs: enhance,
- proxies, package install, cache clear, a jobs list, CBR export.
-
-20. **The title rasteriser copies the whole font file per render** and draws
- on a frame-sized canvas it then un-premultiplies pixel by pixel
- (`concat-text/src/lib.rs:268-279,529-539`). With a CJK `.ttc` that is tens
- of megabytes per animated frame.
-
-### Low
-
-21. **UI facts that are wrong or stale.** Settings says H is "Flip
- horizontally" (`ui/dialogs/settings.slint:671`) while H is the pan tool
- (`app.slint:418`); it lists 3 of about 27 bindings. Tooltips show ⌘ on
- every platform (`ui/timeline/tray.slint:127,134`). README says twelve
- languages; there are 14 files. Two lane colours ignore the theme
- (`lanes.slint:792,835`). The window minimum is 900×560 while Android is
- listed as supported.
-
-22. **Dead Slint still imported.** `ui/demo/` (495 lines) is unreachable, yet
- `EffectData` is imported from it by `app.slint:21`, `editor.slint:4` and
- `workspace/config-pane.slint:9`.
-
-23. **Stale docs.** ARCHITECTURE §8 says 429 tests in 75 files (544 in 82);
- §1's crate map omits speech → media; §7's diagram puts the window over the
- API; `src/README.md:115` says `mobile.yml` runs on push (it does not);
- `release.yml:13` describes archives the build does not ship;
- `CONTRIBUTING.md:43` calls `test/` fixtures (it is Python prototypes).
- Newcomer setup omits ONNX Runtime, libclang, ALSA and GTK.
-
-24. **Small hygiene.** 8 files fail `cargo fmt --check`. `core::Project`
- (`timeline.rs:500`) is dead. `missing_media` does disk IO inside a crate
- that must build for wasm (`model.rs:1896`). The CLI's `serve` parks
- forever and never removes its Unix socket (`concat-cli/src/main.rs:226`).
- `lib.rs:23` in the window uses `slint::private_unstable_api`. Pipelines
- are keyed by `id@version`, so a WGSL edit without a version bump keeps the
- stale pipeline (`shader.rs:417`).
-
----
-
-## 6. What is good, and should stay that way
-
-- **The layering.** Every arrow points one way. FFmpeg is only in
- concat-media; ONNX Runtime is only behind `concat-vision/src/runtime.rs`;
- concat-core has no dependencies and builds for wasm.
-- **`FramePlan`.** Geometry and shading are computed once for both
- compositors (`plan.rs:140-416`), so placement and weighting cannot drift.
-- **The document.** Commands are the only write path; snapshots share through
- Arc; gestures fold into one step; no-ops record nothing; batches are
- all-or-nothing; unknown fields survive a round trip; migrations are a table.
-- **Hardware decode.** Reference-counted device, `get_format` fallback, a
- failed download reopens in software and seeks back to the exact frame,
- with tests that the frame count and timing match.
-- **Package hygiene at load.** `deny_unknown_fields`, every chain rendered at
- its defaults, minimum and maximum, an FFmpeg filter allowlist, LUT paths
- held to the folder, the `concat.` namespace reserved.
-- **The API contract.** Every verb is JSON named `area.verb`; gRPC wraps the
- same JSON; the docs match the code on every method spot-checked.
-- **The window's pane pattern.** State + `Msg` + `update` + `data`, with the
- pane taken out during `update`; the monitor keeps one request in flight and
- builds its texture on the event loop on purpose.
-- **Model fetching.** `.part` files, staging, rename into place, a
- compiled-in fallback cutout model, licences recorded for every model.
-- **Tests that check behaviour.** Synthetic media coloured by source second;
- SSIM parity; migration tests against v1 documents; token handshake tests
- over TCP and gRPC.
-
----
-
-## 7. Where we stand
-
-The 16 September proposal had eleven steps. All eleven landed on `main`
-within the week: the bug-fix pass, structural-sharing undo, `Clip::blank` and
-`Clip::tidy`, `resolve.rs`, the tolerant reader with migrations, keys that
-stay on their instant, ten panes behind one `Msg` loop, the source cache and
-scheduler, `FramePlan` with parity, hardware decode, timeline virtualisation
-and the loopback token. The tree is now the shape the proposal drew.
-
-What that week did not buy is enforcement. Almost every serious finding here
-is a stated rule the code does not yet hold: a corrupt document is an error
-(it is not), a hostile package is refused (it is not tried), clamps live in
-one place (four), stale results are dropped in one place (nowhere), publish
-by diff (rows are rebuilt), one engine handle (two). The debt has moved from
-structure to bounds and guarantees, which is the right direction, and is
-cheaper to pay.
-
-**Recommended order.** Each line is a day or a few, and each closes a
-finding above.
-
-1. Correctness trio: `Session::open` distinguishes not-found from parse
- error; `catch_unwind` with a counter guard around scheduler jobs; a
- session epoch captured in `spawn` and checked in `deliver`. (1, 2, 3)
-2. Server limits and a root allowlist: line length cap, pre-auth timeout,
- connection cap, bounded outbox, bounded preview and export sizes,
- canonical-path check against configured roots. Hand the Remote server the
- window's exporter and sessions. (4, 5)
-3. Fill every `sha256`, pin upstreams to commits, reject an empty digest in
- release builds, generate the five model tables from the manifest. (6)
-4. Run `trial` at install at a realistic size, wrap pipeline creation in an
- error scope that disables the package, check binding types, key pipelines
- by source hash. (7)
-5. `clips_where(pred)` in place of `clips_mut`, with `ptr_eq` tests for
- ripple and freeze; one clamp pass after every command; subtract
- `start_time` in one place; fix Darken and Lighten as
- `mix(ground, min(colour, ground), α)`. (8, 9, 10, 13)
-6. Art refresh only on revision change; targeted publish from `deliver`;
- cache treat graphs by spec and size; `av_frame_ref` instead of deep
- clone; sweep proxy and artwork. (11, 15, 16)
-7. CI: macOS and Windows `cargo check` plus engine tests; lavapipe on Linux
- with parity required; `concat-perf --check` as a step; release depends on
- CI. Fix the stale docs in the same pass. (12, 23)
-
-After those, the architectural follow-ups in ARCHITECTURE §10 stay in their
-order: the plan filled by the export, source textures pooled, zero-copy
-hardware frames, one texture per track, native scalers behind Enhance,
-multi-pass packages.
-
----
-
-## 8. Caveats
-
-Scores are judgements over code, calibrated across reviewers but not
-measured against users or against another editor. Performance findings are
-from reading hot paths plus the perf table; the window's repaint was not
-measured. The OS-side claims in §10 of ARCHITECTURE (macOS 26 scalers, the
-Windows super-resolution API) were not checked. Findings marked (verified)
-were re-read by hand; the rest rest on one reviewer's reading and a line
-number you can open.
diff --git a/audit/AUDIT-2026-09-28-QUICK.md b/audit/AUDIT-2026-09-28-QUICK.md
deleted file mode 100644
index 27063aa8..00000000
--- a/audit/AUDIT-2026-09-28-QUICK.md
+++ /dev/null
@@ -1,136 +0,0 @@
-# Concat audit, 28 Sep 2026 — the quick version
-
-Same audit as `AUDIT-2026-09-28.md`, one screen. Each bar has 10 slots:
-🟢 filled = good, ⚪ empty = missing. Severity: 🔴 critical · 🟠 high · 🟡 medium · ⚪ low.
-Arrows are against the 23 Sep audit. *prov.* = that crate's review was still
-running when this was written; score carried from 23 Sep and adjusted only by hand checks.
-
-## 🎯 Verdict
-
-**The pipeline is one thing now, and the guarantees are half paid.** Five days and 143 commits after the last audit, the picture path is one GPU compositor in linear light with HDR in and out, 9 of the 24 findings are closed (4 of the 6 criticals), 10 are part-closed, and `main` is green. No new critical. What came in with the speed: a self-update whose trust root is the same GitHub release it downloads from, a Remote API rooted at the home folder, and three seams where pre-0.2.5 documents open changed and say nothing.
-
-## 📊 The whole app
-
-| Aspect | Bar | 23 Sep | Today |
-|---|---|:-:|:-:|
-| 🧹 Code quality | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 |
-| 🏛️ Architecture | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7 | 8 ↑ |
-| 🧭 Design philosophy | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6 | 7 ↑ |
-| 🔧 Maintainability | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 😊 User likeability | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 |
-| ⚡ Performance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5 | 6 ↑ |
-| 📈 Scalability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 4 | 5 ↑ |
-| 🔐 Security | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 4 | 5 ↑ |
-| 🧪 Testing | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6 | 7 ↑ |
-| 📚 Docs | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 |
-| 🤖 CI and release | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 4 | 6 ↑ |
-
-## 🧱 Per crate (average of its six scales)
-
-| Crate | Bar | Avg | One thing |
-|---|---|:-:|---|
-| concat-effects *prov.* | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7.7 ↑ | 🟢 format 2, probes on all 98 · 🟡 `[[replaces]]` drops keys |
-| concat-core | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.3 | 🟢 no deps, rational time · ⚪ dead `Project`, linear lookup |
-| concat-cli | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.3 | 🟢 token from env · ⚪ `serve` parks, socket left |
-| concat-api | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.2 | 🟢 real roots check · 🟠 rooted at the home folder |
-| concat-server | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.2 ↑ | 🟢 line, seat, outbox caps · 🟠 fd leak per connection |
-| concat-vision *prov.* | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.2 | 🟢 unchanged, ORT wrapped · ⚪ serial batch-1 |
-| concat-text *prov.* | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.2 | 🟢 any font via fontdb · ⚪ per-render copy to confirm |
-| concat-project | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.8 ↑ | 🟢 `clips_where` + one tidy pass · 🟡 no frame ceiling, bin cloned per snapshot |
-| concat-media *prov.* | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.8 ↑ | 🟢 `catch_unwind`, `start_of` · ⚪ copies per frame to confirm |
-| concat (Slint) *prov.* | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.8 | 🟢 one Modal pattern · ⚪ two `accessible-*` in 27 346 lines |
-| concat-render *prov.* | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.7 ↑ | 🟢 one compositor, oracle in tests · ⚪ `gpu.rs` 3 599 lines |
-| concat-speech *prov.* | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.2 | 🟢 CoreML with CPU fallback · 🟡 only the window can reach it |
-| concat-export | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 | 🟢 device lent in · ⚪ serial loop, 40-field `ExportClip` |
-| concat-host | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 ↑ | 🟢 corrupt doc refused, with tests · 🟠 update trust chain |
-| concat (Rust) *prov.* | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 | 🟢 project epoch exists · ⚪ `studio.rs` 9 461 lines |
-
-## 🎛️ Per feature
-
-| Feature | Bar | 23 Sep | Today | State |
-|---|---|:-:|:-:|---|
-| Launcher | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ exact NTSC rates |
-| Import and bin | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Timeline editing | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ trim-follow, Rotate |
-| Undo / redo | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ sharing survives ripple |
-| Transitions | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 6 | 8 ↑ | ✅ in the monitor, pinned by pixel |
-| Effects and filters | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7 | 8 ↑ | ✅ 98 GPU packages in light |
-| LUTs | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7 | 8 ↑ | ✅ BOM, Resolve range, float |
-| Languages | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 6 | 8 ↑ | ✅ 14 locales, all 100 % |
-| Relink | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Themes | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| ★ Colour and HDR | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | – | 7 | ✅ HDR in and out; pre-0.2.5 imports keep SDR proxies |
-| ★ Scopes | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | – | 7 | ✅ GPU-counted (render review pending) |
-| ★ Colour grading | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | – | 7 | ✅ wheels keyable, curves not |
-| Custom packages | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 5 | 7 ↑ | ✅ trialled at install, error scopes |
-| Crop, blend, masks | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6 | 7 ↑ | ✅ Darken/Lighten fixed |
-| Titles | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ any font; missing font silent |
-| Keyframes | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ old preset animations dropped silently |
-| Cutout | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Captions | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Text-to-speech | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ verified downloads |
-| Templates | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Export | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ window's device, CBR, HDR; one thread |
-| Hardware decode | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ per-stream choice; not zero-copy |
-| Enhance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ one model; copies in `cache/` |
-| Playback and audio | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ 400 pieces per curved clip |
-| Remote API | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ home root, fd leak, second GPU device |
-| ★ Phone shell | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ real shell; no phone build on push |
-| Proxies | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 4 | 5 ↑ | ⚠️ swept at open; still no UI |
-| ★ Self-update | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | – | 5 | ⚠️ no signature; downgrade one click |
-| Accessibility | 🟢🟢⚪⚪⚪⚪⚪⚪⚪⚪ | 2 | 2 | ❌ two `accessible-*` properties |
-
-## 🔁 The 23 Sep findings
-
-| Closed (9) | Partly closed (10) | Open (2) | Pending a review (3) |
-|---|---|---|---|
-| 1 corrupt doc · 2 scheduler panic · 5 two engines · 6 model digests · 7 hostile packages · 8 undo sharing · 9 start time · 10 Darken/Lighten · 12 CI | 3 wrong-project results · 4 API paths and sizes · 11 art scan · 13 clamps · 14 E2E pixels · 15 cache bounds · 17 Hub thread · 21 UI facts · 23 docs · 24 hygiene | 19 speech only in the window · 22 dead `ui/demo/` | 16 per-frame copies · 18 ORT twice · 20 font copied per render |
-
-## 🚨 Top new findings
-
-| # | Sev | What | Where |
-|:-:|:-:|---|---|
-| 1 | 🟠 | Self-update: digest published beside the package, no signature, quarantine stripped, Windows installer unsigned | `concat-host/src/updates.rs:45,381,430,539` · `release.yml:100` |
-| 2 | 🟠 | Remote page's API is rooted at the home folder: a token holder can truncate `~/.ssh/authorized_keys`; no `HOME` = anywhere; open-then-save escapes | `concat-api/src/lib.rs:648,950` · `concat-server/src/lib.rs:80` |
-| 3 | 🟠 | Every JSON connection leaks a file descriptor; seats counted before auth | `concat-server/src/json.rs:106,161,102` |
-| 4 | 🟠 | Clear cache deletes the media behind reversed and enhanced clips; never swept | `concat-host/src/reverse.rs:71` · `projects.rs:330` |
-| 5 | 🟠 | 0.2.4 documents with reversed clips or preset animations open changed, no notice | `concat-project/src/doc.rs:62` |
-| 6 | 🟡 | Embedded API still opens a second GPU device (the #202 path) | `concat-api/src/lib.rs:179,761` |
-| 7 | 🟡 | `[[replaces]]` drops keyframes, is silent, skips user packages and templates | `concat-project/src/editor.rs:89` · `catalogue.rs:585` |
-| 8 | 🟡 | HDR proxies not regenerated for pre-0.2.5 imports | `concat/src/studio.rs:6123` · `model.rs:171` |
-| 9 | 🟡 | Hub: heavy work in line, no `catch_unwind`, slow event reader silently dropped | `concat-server/src/hub.rs:58,83` |
-| 10 | 🟡 | Downgrade is one click with no gate | `updates.rs:141` · `settings.rs:447` |
-| 11 | 🟡 | No ceiling on frame size or rate in the model | `commands/timelines.rs:54` |
-| 12 | 🟡 | Every undo snapshot deep-copies the bin | `concat-project/src/editor.rs:141` |
-| 13 | 🟡 | Export still one thread with a readback per frame; 400-piece curves; mixer scans every clip | `concat-export/src/lib.rs:1204,1005` · `playback.rs:1032` |
-| 14 | ⚪ | Font by absolute path, missing font silent | `model.rs:1797` · `titles.rs:211` |
-| 15 | ⚪ | `has_non_finite` misses 3 fields; no-op undo steps; HDR re-flip; no `extra` on `video`/`text`; `"concat": "0.1.0"` | `commands/mod.rs:878` · `doc.rs:317` |
-| 16 | ⚪ | Socket umask mode; `connections()` over-counts; `export.progress.stage` names differ from docs | `concat-server/src/lib.rs:196` · `message.rs:591` |
-| 17 | ⚪ | `rpm -q` on the UI thread; predictable export temporaries; two TLS stacks | `settings.rs:796` · `lib.rs:887` |
-| 18 | ⚪ | Carried: 5 ⌘ glyphs, `ui/demo/`, dead `core::Project`, `Path::exists` in the wasm crate, speed range ×5 | various |
-
-## 🧾 Today's checks
-
-| Check | Result |
-|---|---|
-| fmt | ✅ 0 files would change (8 on 23 Sep) |
-| clippy | ✅ 0 warnings |
-| tests | ⚠️ **647 passed, 2 failed**, 1 ignored (650 in all, 544 on 23 Sep). Both failures are window tests that saw German where they expected English (`format::tests::phrases_are_coarse`: "gestern" for "yesterday"; `panes::speech::tests::a_voice_name_reads_as_a_person_would_say_it`): the process-wide locale leaking between tests in one binary (confirmed: both pass alone, and the suite passes with `--test-threads=1`), not a product bug. CI on `main` is green. |
-| perf --check | ✅ 22 of 22 within budget: GPU composite 2.3 ms (3.2), 720p export 204 fps (115), HDR export 111 fps, 4K HLG deep-for-GPU 140 fps vs 29 tone-mapped on the CPU, 4K HEVC 10-bit hardware 112 fps and software 43, three 4K streams 61 fps, 4K Gaussian blur 6.1 ms at radius 10 and 9.3 at 50 |
-| CI on `main` | ✅ green at head; 3 of today's 12 runs were red between 10:12 and 11:02 UTC, fixed by `d1d3265` |
-
-## 🗺️ Do next, in order
-
-1. 🟠 **Sign the update** — a key compiled into the app, a signed manifest, keep quarantine, confirm downgrade; sign the Windows installer and notarise every macOS build.
-2. 🟠 **Remote page defaults** — root at project and export folders; refuse to start with no root; check `writable` on open or save; seats after auth; drop the connection clone on close; `catch_unwind` round the Hub; lend the window's device to the embedded API.
-3. 🟠 **Processed media out of `cache/`** — reverse and enhance copies under `media/`; sweep art and peaks; re-probe pre-0.2.5 media at open.
-4. 🟠 **Say it when a document changes** — a migration for `reverse` and the animations; surface `upgrade_links` and carry its keys; run it for templates and user packages.
-5. 🟡 **Finish the 23 Sep list** — epoch on every worker path; a tidy for the frame; one speed range; the CLI socket; `ui/demo/`; the ⌘ glyphs; the two flaky-by-locale tests isolated.
-6. 🟡 **Pipeline the export**, index the mixer, `Arc` the bin in snapshots.
-7. 🟡 **Speech, enhance, proxies and CBR as API verbs.**
-
-Then HDR phase 4 (a true HDR preview) and phase 5 (zero-copy, hardware encoders).
-
-## ✅ Keep
-
-Layering · one compositor with the oracle in tests · `Session::open`'s three-way split · `clips_where` + `tidy_touched` · `follow_first_hdr` as an editor diff · mandatory digests with pinned upstreams · the updater's input hygiene · `Api::writable` · `OpenProjects` · `SingleFlight` · the lent compositor · Reverse's bounded two-pass design · tolerant `wire` readers · one contract, two transports · a hand-written CHANGELOG · CI that gates on fmt, clippy, perf and wasm.
diff --git a/audit/AUDIT-2026-09-28.md b/audit/AUDIT-2026-09-28.md
deleted file mode 100644
index 99057d7e..00000000
--- a/audit/AUDIT-2026-09-28.md
+++ /dev/null
@@ -1,596 +0,0 @@
-# Concat audit, 28 September 2026
-
-Where the codebase stands five days after the 23 September audit, on the day
-0.2.5 shipped. Same method, same seven scales, same reviewers' split, and
-every one of the 24 findings from last time re-verified against today's
-tree. Every finding names a file and a line. The sibling
-`AUDIT-2026-09-28-QUICK.md` is this audit as a one-screen scorecard.
-
-**How it was done.** Nine reviewers each read one part of the tree in full
-(core and project; media; render and effects; export and host; the window's
-Rust; the Slint UI; the API, server and CLI; speech, vision, text, Android
-and the model manifest; tests, CI, build, docs and release), re-verified the
-23 September findings assigned to them, scored their part and reported new
-findings with locations. The headline items were re-checked by hand against
-the code before they went in here. Line counts, test counts and the unsafe,
-unwrap and panic tallies come from the tree; `cargo fmt --check`,
-`cargo clippy --workspace --all-targets`, `cargo test --workspace` and
-`cargo run -p concat-perf --release -- --check` were run on this machine
-(Apple Silicon, macOS 26). The app itself was not launched.
-
-**Status of this document.** Written the same morning. The core/project,
-export/host and API/server/CLI reviews are in and folded in below; the
-media, render/effects, window, Slint, models and CI/docs reviews were still
-running when this was written, so those crates' scores are marked
-*provisional* (carried from 23 September, adjusted only where a hand check
-justified it). Each of those sections is replaced as its review lands. The
-check suite, the perf table and the CI history are complete.
-
-**Scale.** 0 to 10. 8 and up means "keep doing this"; 5 to 7 means "works,
-with known debt"; 4 and under means "will bite at the next order of
-magnitude, or already bites".
-
----
-
-## 1. The numbers
-
-### 1.1 Size, tests and hygiene per crate
-
-Deltas are against the 23 September table. The unwrap column gives the total
-and, in brackets, the count outside test modules.
-
-| Crate | Rust lines | Δ | Slint / WGSL | Tests (files) | `unsafe` | `unwrap`/`expect` (non-test) | `panic!` family |
-|---|---:|---:|---:|---:|---:|---:|---:|
-| concat (window) | 19 564 | +3 506 | 27 346 slint | 59 (14) | 1 | 29 (5) | 1 |
-| concat-host | 11 540 | +2 813 | | 76 (17) | 1 | 227 (2) | 15 |
-| concat-project | 9 192 | −719 | | 98 (1) | 0 | 310 (11) | 14 |
-| concat-media | 9 041 | +1 003 | | 86 (11) | 37 | 188 (42) | 5 |
-| concat-render | 8 044 | +3 171 | shaders inline in `gpu.rs` | 85 (6) | 1 | 87 (8) | 4 |
-| concat-effects | 7 090 | +2 426 | 1 744 wgsl | 69 (9) | 0 | 129 (10) | 2 |
-| concat-export | 3 498 | +212 | | 27 (3) | 0 | 23 (0) | 0 |
-| concat-speech | 3 094 | +7 | | 19 (3) | 3 | 23 (2) | 0 |
-| concat-core | 2 829 | +138 | | 47 (7) | 0 | 24 (4) | 2 |
-| concat-vision | 2 760 | 0 | | 31 (8) | 0 | 22 (2) | 0 |
-| concat-api | 2 350 | +357 | | 21 (2) | 0 | 23 (0) | 8 |
-| concat-server | 1 376 | +193 | | 12 (4) | 0 | 30 (5) | 2 |
-| concat-text | 1 240 | +139 | | 15 (1) | 0 | 36 (1) | 1 |
-| concat-perf | 1 092 | +288 | | 2 (1) | 0 | 34 (34) | 0 |
-| concat-cli | 514 | +65 | | 3 (1) | 0 | 2 (2) | 0 |
-| concat-android | 335 | 0 | | 0 (0) | 10 | 9 (9) | 2 |
-| **Total** | **~83 600** | **+12 600** | **~29 100** | **650 (88)** | **53** | **~1 200 (137)** | **56** |
-
-The tree grew by an eighth in five days and lost weight in the right
-places: concat-project shed its 800-line preset-animation module, and
-concat-effects shed 59 format 1 picture packages (WGSL lines fell from
-2 205 to 1 744 while the crate's Rust grew by half for the format 2
-contract, multi-pass and probes). concat-render doubled: the CPU compositor
-became a test oracle and the GPU one took on linear light, HDR, scopes and
-the texture cache. The window's Rust grew by a fifth, most of it in
-`studio.rs`, which is now 9 461 lines (7 484 on the 23rd).
-
-Outside test modules the workspace holds 137 `unwrap`/`expect` calls, 34 of
-them in the perf harness and 42 in concat-media's FFI, and 13 `panic!`
-family calls, 7 of which are `unreachable!` routing arms in concat-project.
-
-Other facts: 123 effect packages (98 on format 2 with WGSL: 21 effects,
-57 looks, 20 transitions; 25 audio packages still on FFmpeg chains, by
-design); 14 locale files, every one at 969 lines with 0 missing and 0 stale
-keys per `scripts/locales.py --check`; 730 commits since 25 August, 143 of
-them since the last audit (810 files, +41 000 / −34 000 lines; 136 by the
-maintainer, 5 from four outside contributors through three merged PRs);
-29 release tags, latest `v0.2.5`, released today; largest files
-`concat/src/studio.rs` 9 461, `concat-project/src/lib.rs` 3 791 (3 755 of
-them tests), `concat-render/src/gpu.rs` 3 599, `concat-export/src/lib.rs`
-2 672, `concat-effects/src/shader.rs` 2 406; largest Slint files
-`workspace/media-pane.slint` 2 156 and `inspector/clip-inspector.slint` 1 954.
-
-### 1.2 The checks, run today
-
-| Check | Result |
-|---|---|
-| `cargo fmt --all -- --check` | 0 files would change (8 on the 23rd) |
-| `cargo clippy --workspace --all-targets` | 0 warnings |
-| `cargo test --workspace --no-fail-fast` | 647 passed, 2 failed, 1 ignored (650 in all; 544 on the 23rd) |
-| `cargo run -p concat-perf --release -- --check` | 22 of 22 scenarios within budget |
-| CI on `main` | green at head (`b78ee56`, CI and Nix both pass) |
-
-**Two window tests fail locally; CI is green.** Both failures are tests
-that translate a string and got German where they expected English:
-`format::tests::phrases_are_coarse` sees "gestern" for "yesterday"
-(`concat/src/format.rs:502`) and
-`panes::speech::tests::a_voice_name_reads_as_a_person_would_say_it` finds no
-"American" in a voice's detail (`panes/speech.rs:672`). Since `fc09a61` the
-locale is process-wide and the locales test in the same binary loads every
-shipped file in turn, `de` first; the two tests read whichever locale the
-other thread left behind. Confirmed: both pass when run alone, and the whole
-window suite passes with `--test-threads=1`. A test-isolation bug, not a
-product one. The end-to-end export suite (11 tests, 21 s), the parity suite
-(85 tests, 39 s) and every engine crate pass.
-
-**The perf table**, release build, every scenario inside its budget. New
-rows since the 23rd are marked ★.
-
-| Scenario | Measured | Budget |
-|---|---:|---:|
-| plan one frame of 200 clips on 4 tracks | 0.96 µs | ≤ 200 µs |
-| 200 clip additions and 200 undos | 0.78 ms | ≤ 150 ms |
-| ripple delete of 200 clips out of 400 | 0.20 ms | ≤ 20 ms |
-| save and open a 200-clip document | 1.98 ms | ≤ 60 ms |
-| choose the visible clips of 200 for one publish | 0.21 µs | ≤ 50 µs |
-| read an hour's waveform across 2048 columns | 4.41 µs | ≤ 2000 µs |
-| decode 1080p h264 in software | 1 626 fps | ≥ 120 |
-| decode 1080p h264 on the hardware (VideoToolbox) | 316 fps | ≥ 120 |
-| ★ decode 4K h264, the app's choice (software) | 151 fps | ≥ 30 |
-| ★ decode 4K HEVC 10-bit, the app's choice (VideoToolbox) | 112 fps | ≥ 30 |
-| ★ decode 4K HEVC 10-bit in software | 43 fps | ≥ 24 |
-| ★ decode 3 streams of 4K HEVC 10-bit at once | 61 fps | ≥ 24 |
-| ★ decode 4K HLG tone-mapped on the CPU, 960×540 | 29 fps | ≥ 15 |
-| ★ decode 4K HLG deep for the GPU, 960×540 | 140 fps | ≥ 30 |
-| scrub forward over 90 frames, cold | 0.72 ms/frame | ≤ 25 |
-| scrub back over covered ground | 0.25 µs/frame | ≤ 200 |
-| decodes when an effect knob is turned | 0 | ≤ 0 |
-| compose 1080p, 3 layers, a look, on the GPU | 2.31 ms | ≤ 25 |
-| ★ Gaussian blur over 4K, radius 10, on the GPU | 6.11 ms | ≤ 33 |
-| ★ Gaussian blur over 4K, radius 50, on the GPU | 9.34 ms | ≤ 33 |
-| export a 4 s 720p cut of two overlapping clips | 204 fps | ≥ 30 |
-| ★ export the same cut as HDR: HLG, HEVC 10-bit | 111 fps | ≥ 30 |
-
-Three readings worth keeping: the 720p export went from 115 to 204 fps
-with the CPU composite gone; the deep HDR decode that hands the GPU the
-conversion runs at 140 fps where the CPU tone map ran at 29, the "five
-times faster" of the CHANGELOG; and hardware decode of 1080p h264 is still
-five times slower than software (316 against 1 626) because the frame is
-downloaded and converted on the CPU after the chip decodes it, the
-zero-copy item of ARCHITECTURE §10.
-
-**`main` is green.** The last 12 runs on `main` are 7 successes, 3
-failures and 2 cancellations. The three failures were this morning's
-commits between 10:12 and 11:02 UTC (the roadmap move, the export-device
-fix, the Version page); `d1d3265` fixed the Windows build and every run
-since passes. CI now runs `cargo fmt --check` (`ci.yml:93`), clippy with
-`-D warnings` (`ci.yml:108`), the quick perf scenarios against their budgets
-(`ci.yml:120`), the engine on wasm32 (`ci.yml:134`) and a macOS 15 plus
-Windows 2022 matrix (`ci.yml:162-168`), and the release job needs the check
-and engine jobs (`release.yml:67`). Four of the seven CI items on the 23rd
-are done.
-
----
-
-## 2. Scorecard for the whole app
-
-| Aspect | 23 Sept | Today | One line |
-|---|:---:|:---:|---|
-| Code quality | 7 | 7 | Clippy-clean under `-D warnings`, fmt-clean, typed errors; `studio.rs` grew 2 000 lines and `ExportClip` has 40 fields. |
-| Design architecture | 7 | 8 | One compositor, `FramePlan` draws transitions for monitor and export alike, the device is lent to the export, the API and window share one register and one export slot. |
-| Design philosophy | 6 | 7 | Rules that were stated and not held on the 23rd now hold: a corrupt document is an error, one clamp pass, digests mandatory, packages trialled. "Writes only under roots" and "dropped in one place" are still half true. |
-| Maintainability | 6 | 6 | The CPU path, 59 packages and the preset catalogue are gone; `studio.rs` at 9 461 lines and `gpu.rs` at 3 599 pull the other way. |
-| User likeability | 7 | 7 | HDR end to end, scopes, colour wheels and curves, a phone shell, self-update, any font; accessibility is still two properties. |
-| Performance | 5 | 6 | Every picture on the GPU in half floats, a texture cache, ping-pong passes, 4K HLG five times faster; the export loop is still one thread with a readback per frame. |
-| Scalability | 4 | 5 | Transport limits, bounded LUT and pool caches, a swept proxy folder; one Hub thread, a file descriptor leaked per API connection, four cache folders never swept. |
-| Security | 4 | 5 | Every model verified and pinned, a canonicalising roots check, constant-time tokens; the default root is the home folder, and the self-update trusts a digest published beside the package. |
-| Testing | 6 | 7 | 650 tests; the end-to-end suite pins pixels for crop, flips, fades, wipes and HDR; session refusal tests; 98 packages hold probes. |
-| Documentation | 7 | 7 | A hand-written CHANGELOG and a current ARCHITECTURE; developer docs moved off the repo; two §7 claims overstate the code. |
-| CI and release | 4 | 6 | fmt, clippy, perf and wasm gates, a desktop matrix, release gated on CI, a `SHA256SUMS` file; nothing is signed and three runs went red today. |
-
----
-
-## 3. Scores per crate
-
-Columns: code quality, maintainability, scalability, performance,
-architecture, design philosophy, security, testing. A dash means the scale
-does not apply. Arrows are against 23 September. Rows marked *prov.* carry
-the 23 September score adjusted only by today's hand checks; their reviews
-were still running.
-
-| Crate | CQ | Maint | Scale | Perf | Arch | Phil | Sec | Test | Note |
-|---|:-:|:-:|:-:|:-:|:-:|:-:|:-:|:-:|---|
-| concat-core | 8 | 8 | 6 | 6 | 8 | 8 | – | 7 | Held. Still no dependencies; `Depth`, `Signal` and `Stage` landed cleanly; linear `clip_on_track_at` and dead `Project` unchanged. |
-| concat-project | 8 ↑ | 7 ↑ | 6 ↑ | 6 ↑ | 7 | 7 ↑ | – | 7 | `clips_where`, one `tidy_touched` after every command, HDR flip as an editor-level diff; the frame has no ceiling, snapshots clone the bin. |
-| concat-media *prov.* | 7 | 7 | 5 | 6 | 8 | 8 ↑ | – | – | `catch_unwind` round jobs and a `start_of(stream)` subtraction exist; the copy count per frame awaits the review. |
-| concat-render *prov.* | 7 | 5 | 6 ↑ | 7 ↑ | 8 ↑ | 7 ↑ | – | – | One compositor with the CPU one as oracle; Darken and Lighten fixed; trial and error scopes exist; `gpu.rs` doubled. |
-| concat-effects *prov.* | 8 | 8 | 6 | 8 ↑ | 8 ↑ | 8 ↑ | – | – | Format 2 contract, multi-pass, probes on every picture package; `[[replaces]]` drops keys (finding 7). |
-| concat-export | 6 | 4 | 6 | 5 | 8 ↑ | 7 | 6 | 7 | Device lent in, plan shared with the monitor; 2 672-line `lib.rs`, 40-field `ExportClip`, 12 side tables, serial loop. |
-| concat-host | 7 ↑ | 6 | 5 | 6 | 6 ↑ | 6 | 4 | 6 | Corrupt documents refused with tests, `SingleFlight`, sweeps; the update trust chain, unswept reverse and enhance copies. |
-| concat (Rust) *prov.* | 7 | 5 | 5 | 6 ↑ | 6 | 7 ↑ | – | – | A project epoch exists and three panes use it; `refresh_art` is one comparison on the hot path; `studio.rs` 9 461 lines. |
-| concat (Slint) *prov.* | 7 | 6 | – | – | 7 | 7 | – | – | 27 346 lines; still two `accessible-*` properties, five ⌘ glyphs, `ui/demo/` still imported. |
-| concat-api | 8 | 8 | 5 | – | 8 | 7 ↓ | 6 ↑ | 7 | A real roots check and bounded sizes; four write fields checked, the home folder the default root, open-then-save unchecked. |
-| concat-server | 7 | 8 | 5 ↑ | – | 8 | 8 ↓ | 6 ↑ | 7 | Line, seat, outbox and pre-auth limits; fd leak per connection, seats counted before auth, no `catch_unwind` on the Hub. |
-| concat-cli | 7 | 8 | – | – | 7 | 7 | 6 | 3 | Held; `serve` still parks forever; the token can come from `CONCAT_API_TOKEN`. |
-| concat-speech *prov.* | 7 | 5 | 6 | 7 | 5 | 7 | – | – | Only the window depends on it still. |
-| concat-vision *prov.* | 8 | 7 | 5 | 6 | 9 | 8 | – | – | Unchanged since the 23rd (0 lines of churn). |
-| concat-text *prov.* | 8 | 8 | 5 | 5 ↑ | 9 | 8 | – | – | Now on `fontdb`; the per-render copy awaits the review. |
-| concat-android *prov.* | 8 | 7 | – | – | – | – | – | – | Unchanged; still no tests. |
-
----
-
-## 4. Scores per feature
-
-Status: **done**, **partial** (works with a known gap), **thin** (present,
-weak). New rows since the 23rd are marked ★.
-
-| Feature | Where | Status | 23 Sept | Today | Why |
-|---|---|---|:-:|:-:|---|
-| Launcher, new project, recents | `ui/start.slint` | done | 8 | 8 | Eleven frame rates plus typed ones, NTSC as exact fractions end to end; a picked folder shows on Windows. |
-| Import, bin, drag and drop | `ui/workspace/media-pane.slint` | done | 8 | 8 | |
-| Timeline editing | `ui/timeline/`, `panes/timeline.rs` | done | 8 | 8 | Trim-follow (#194), Rotate, Freeze/Reverse/Mirror behind one mark. |
-| Undo and redo | `concat-project/src/editor.rs` | done | 8 | 8 | Sharing survives ripple now; every snapshot still clones the bin (finding 12). |
-| Transitions | `concat-render/src/plan.rs`, `transitions.rs` | done | 6 | 8 | Fades and wipes drawn by the plan in monitor and export alike, pinned by pixel in the E2E suite; packaged ones whole on the GPU. |
-| Effects and filters | `concat-effects/packages/` | done | 7 | 8 | 98 GPU packages in linear light with probes; the CPU chains are gone; retired ids open with a notice. `[[replaces]]` drops keys (finding 7). |
-| ★ Colour and HDR | `model.rs` `color_space`, `gpu.rs`, `encode.rs` | done | – | 7 | A timeline colour, deep decode, HDR HEVC/AV1 export with PQ metadata, HDR proxies; the monitor is tone-mapped until phase 4; pre-0.2.5 imports keep SDR proxies (finding 8). |
-| ★ Scopes | `concat-render/src/scopes.rs`, `scopes-pane.slint` | done | – | 7 | Waveform, parade, vectorscope, histogram counted on the GPU; provisional until the render review. |
-| ★ Colour grading | `packages/concat.color-wheels`, `curves`, `grading.slint` | done | – | 7 | Wheels keyable, curves not; widgets pointer-only (UI review pending). |
-| Titles and text | `concat-text`, `concat-host/src/titles.rs` | done | 7 | 7 | Any font on the machine or imported; a missing font falls back silently (finding 14). |
-| Keyframes, animations | `concat-project/src/model.rs` | done | 7 | 7 | Rotation keys wrap on purpose; preset animations were removed and old documents are not told (finding 5). |
-| Crop, transform, blend, masks | `concat-render/src/plan.rs` | done | 6 | 7 | Darken and Lighten fixed on both paths; crop and flips in the plan and pinned by pixel. |
-| Cutout, smart brush | `concat-vision/src/segment.rs` | done | 7 | 7 | Unchanged. |
-| Enhance | `concat-host/src/enhance.rs` | thin | 6 | 6 | One model on every platform; copies land in `cache/` and are deleted by Clear cache (finding 4). |
-| Auto-captions | `ui/dialogs/captions.slint` | done | 7 | 7 | Whether every result checks the project epoch awaits the window review. |
-| Text-to-speech | `concat-speech/src/tts.rs` | done | 7 | 7 | CoreML on every Mac with a CPU fallback; downloads now verified. |
-| Templates | `concat-host/src/templates.rs` | done | 7 | 7 | `delete` fenced by canonical prefix; the `[[replaces]]` pass is skipped for templates. |
-| LUT import | `concat-effects/src/cube.rs`, `looks.rs` | done | 7 | 8 | BOM and Resolve range (#208), tables in float, ACEScct in log. |
-| Custom and community packages | `concat-effects/src/catalogue.rs` | done | 5 | 7 | Trialled on the GPU at install with a timeout; error scopes at draw; multi-pass real. |
-| Proxies | `concat-host/src/proxy.rs` | partial | 4 | 5 | Swept at project open; HDR proxies; still no UI. |
-| Export | `ui/dialogs/export.slint`, `concat-export` | done | 7 | 7 | Draws on the window's device; CBR fixed; HDR row; still one thread, one readback per frame. |
-| Playback and audio | `concat-host/src/playback.rs` | partial | 6 | 6 | Audio cache bounded to 2 GB; a speed-curved clip is still up to 400 decodes; the mixer scans every clip per sample. |
-| Hardware decode | `concat-media/src/hardware.rs` | done | 7 | 7 | Per-stream CPU-or-chip choice; on by default on Windows and Android; still not zero-copy. |
-| Remote API | `concat-api`, `concat-server` | partial | 6 | 6 | Roots, limits and a shared register; the root is the home folder, fds leak, the embedded API opens a second GPU device. |
-| ★ Self-update | `concat-host/src/updates.rs` | done | – | 5 | Drafts and pre-releases filtered, tag validated, file names checked, no auto-check at start; digest from the same release, no signature, quarantine stripped, downgrade unconfirmed (findings 1, 10). |
-| ★ Phone shell | `ui/phone/`, `platform::phone` | partial | – | 6 | An editor of its own on Android and iOS, testable on a desk with `CONCAT_PHONE=1`; no phone build in CI on push. |
-| Relink missing media | `ui/dialogs/relink.slint` | done | 8 | 8 | |
-| Themes and accent colours | `ui/theme/` | done | 8 | 8 | |
-| Languages | `concat/locales/` | done | 6 | 8 | 14 locales, all complete (969 keys each, 0 missing), keyed by ids; RTL awaits the UI review. |
-| Accessibility | the whole UI | thin | 2 | 2 | Two `accessible-*` properties in 27 346 lines. |
-
----
-
-## 5. The 23 September findings, re-verified
-
-Nine of the 24 are closed, ten are partly closed, two are open, and three
-await a review still running. Of the six criticals, four are closed and two
-are partly closed.
-
-| # | Finding (23 Sept) | Status | Evidence today |
-|:-:|---|:-:|---|
-| 1 | Corrupt document opens empty, next save overwrites it | **closed** | `read_document` returns `Ok(None)` only for `NotFound` and `Err` for a parse failure (`concat-host/src/projects.rs:213-222`); `Session::open` returns it and refuses a newer version too (`session.rs:73-100`); tests `session.rs:341-369`. |
-| 2 | Panicking scheduler job wedges the scheduler | **closed** | `catch_unwind(AssertUnwindSafe(job))` at `concat-media/src/prefetch.rs:363`; the counter path awaits the media review. |
-| 3 | Worker results land in whichever project is open | partial | `host::project_epoch()` captured at spawn and checked by `on_ui_in_project` (`concat/src/studio.rs:5136, 5372, 5461`); 3 of 11 pane files reference it; whether every worker path does awaits the window review. |
-| 4 | API writes any path, sizes unbounded | partial | Roots check refuses `..`, canonicalises the deepest existing ancestor and compares component-wise (`concat-api/src/lib.rs:243-275, 910-929`); `MAX_SIDE` 8192, `MAX_CRF` 63, `MAX_RATE` 240; line cap 4 MiB, auth line 4 KiB, 10 s pre-auth, 64 seats, 256-line outbox (`concat-server/src/json.rs`). Left: the home folder as the root, unconfined with no `HOME`, open-then-save, socket mode (findings 2, 3). |
-| 5 | Window and Remote page are two engines on one folder | **closed** | `OpenProjects` register claimed on every open path (`concat-api/src/lib.rs:99-142, 410-416`), the window claims `Holder::Window` (`studio.rs:6098, 6195, 6239`), one `Arc<SingleFlight>` export slot (`concat-host/src/export.rs:129-132`). Per process only; the embedded API still opens its own GPU device (finding 6). |
-| 6 | Model downloads unverified and unpinned | **closed** | 23 of 23 manifest rows carry a 64-hex `sha256`, 0 at `resolve/main`, 18 pinned to a commit; `verify` refuses an empty digest (`concat-host/src/models.rs:295-301`, test 472-485). |
-| 7 | Hostile package not really refused | **closed** | `Catalogue::install_with` at `studio.rs:6447` trials every user package on the GPU (`concat-host/src/preview.rs:120` → `trial_at`); pipelines are built inside `push_error_scope` (`gpu.rs:1820`). The loop and binding checks await the render review. |
-| 8 | Undo's sharing defeated by every ripple | **closed** | `Timeline::clips_where` copies only the picked clips (`concat-project/src/model.rs:1972-1980`); `clips_mut` is gone; used by ripple trim, freeze, ripple delete and fill slot (`commands/clips.rs:239, 459, 584, 651`, `media.rs:113`). A `ptr_eq` test covers ripple delete only (`lib.rs:1316-1353`). |
-| 9 | Stream start time ignored | **closed** | `start_of(stream)` (`concat-media/src/ffi.rs:164`) and `Decoder::starting_at` (`decode.rs:146`); the pool's use awaits the media review. |
-| 10 | Darken and Lighten wrong at partial opacity | **closed** | `min`/`max(colour, ground) · α` premultiplied (`concat-render/src/gpu.rs:152-160`), the oracle the same (`reference.rs:189`). |
-| 11 | Art scan on every pointer event; full publish 30×/s | partial | `refresh_art` still runs after every callback (`concat/src/lib.rs:117, 420, 1185`) but is one comparison unless the document, the screen or the pending set changed (`studio.rs:8397-8403`). The publish rate awaits the window review. |
-| 12 | `main` red; CI Linux-only until a tag | **closed** | fmt, clippy `-D warnings`, quick perf, wasm32, macOS and Windows matrix in `ci.yml:93-168`; release needs CI (`release.yml:67`); `main` green at head. Parity on the Linux runner awaits the CI review. |
-| 13 | Clamps in four places | partial | `tidy_touched` runs `Clip::tidy` over every clip a command wrote (`concat-project/src/editor.rs:170-197`) and on load (`doc.rs:275`). The speed range still has 3 definitions and 2 literal copies (`model.rs:1076`, `concat-core/src/retime.rs:27`, `concat-media/src/audio.rs:35`, `studio.rs:3980`, `concat-host/src/cutout.rs:178`); `properties.rs:156-159` is a stale comment. |
-| 14 | E2E suite checks timing, not looks | partial | `colour_at_point` (`concat-host/tests/export.rs:358`) pins crop and flips (753), fades and wipes (887), HLG to HDR and SDR (979, 1036), export on a lent device (701), reverse (1862), hardware decode (1788). Effects, keys, transform, blend and titles await the tests review. |
-| 15 | Caches with no bound | partial | `cache/proxy` swept at open (`proxy.rs:166-186`), audio cache 2 GB (`playback.rs:62-64`), titles 400 PNGs (`titles.rs:355-380`), GPU pool 512 MB / 1024 textures. Still unswept: `cache/art`, peaks, reverse and enhance copies (`media.rs:337-364`; finding 4). |
-| 16 | Per-frame copies; treat graph rebuilt per frame | pending | The treat chains went with the CPU path; the decode copies await the media review. |
-| 17 | Hub does heavy work in line; fan-out under a mutex | partial | Fan-out clones the list and gRPC uses `try_send` (`hub.rs:54-66`, `grpc.rs:70-76`). The Hub still composites previews, paints titles, probes and copies templates on its one thread (finding 9). |
-| 18 | ONNX Runtime linked twice | pending | Awaits the models review; two HTTP stacks are linked (ureq 2 with rustls, ureq 3 with native-tls through `ort-sys`). |
-| 19 | Speech only reachable from the window | **open** | Only `concat` depends on `concat-speech`; 21 verbs, none for speech, enhance, proxies, packages, cache or jobs; the API's export hard-codes VBR (`concat-api/src/lib.rs:568-569`), so the CBR fix of #104 is the window's alone. |
-| 20 | Title rasteriser copies the font per render | pending | `concat-text` is on `fontdb` now (`lib.rs:198`); the copy awaits the models review. |
-| 21 | Stale UI facts | partial | "Flip horizontally" gone with the shortcuts section; README says 14 languages; the window minimum is 320×480 on a phone (`app.slint:92-93`). ⌘ still appears in 5 places; lane colours await the UI review. |
-| 22 | Dead `ui/demo/` still imported | **open** | `ui/demo/editor-chrome.slint` and `effects-panel.slint` remain, imported by `app.slint`, `editor.slint` and `workspace/config-pane.slint`. |
-| 23 | Stale docs | partial | ARCHITECTURE §8 no longer states a count; `test/` is gone; `src/README.md:114` still says `mobile.yml` runs on every pull request; §7's "writes only under its roots" holds for four fields of nine verbs. |
-| 24 | Small hygiene | partial | fmt clean. Still: dead `core::Project` (`concat-core/src/timeline.rs:487-504`), `Path::exists` in the wasm crate (`model.rs:1900-1910`), the CLI's `serve` parks forever and leaves its socket (`concat-cli/src/main.rs:247-249`). |
-
----
-
-## 6. New findings, by severity
-
-No new critical. Each carries a location and whether the reviewer read the
-whole path **(verified)** or reasoned from part of it.
-
-### High
-
-1. **The self-update trusts a digest published beside the package, with no
- signature, and takes Gatekeeper out of the loop.** (verified) The release
- list comes from `api.github.com` (`concat-host/src/updates.rs:45`), the
- package URL and its SHA-256 from `manifest.json` in the same GitHub
- release (`updates.rs:340-383`), written by `scripts/models.py
- release-manifest` in the release job (`.github/workflows/release.yml:100-108`)
- with `contents: write`. The check at `updates.rs:430` proves only that
- the bytes match what the release says they are: whoever can write the
- release (a leaked token, a compromised maintainer account, the release
- action) replaces the package and its digest together, and every person
- who presses Install runs it. On macOS the swapped bundle then has its
- quarantine stripped (`updates.rs:539-541`); the Windows installer is
- unsigned (no `signtool` in `build-app.yml`; Inno with
- `PrivilegesRequired=lowest`, `assets/windows/concat.iss:59`); the macOS
- bundle is notarised only when the secrets are set (`build-app.yml:463-498`).
- The input hygiene around it is good: drafts and pre-releases filtered,
- the tag validated before it enters a URL, a file name with a folder
- refused, no check at start (only when the page opens). A compiled-in
- public key and a signed manifest would close this.
-
-2. **From the window's Remote page a token holder can truncate
- `~/.ssh/authorized_keys`.** (verified) `Config::roots` empty means the
- home folder (`concat-server/src/lib.rs:80-86, 145-149`), and the Remote
- page passes no roots (`concat/src/panes/settings.rs:561-565`).
- `preview.frame` with `output: "~/.ssh/authorized_keys"` (or `~/.zshrc`,
- a keychain, an autostart entry) passes `writable` (`concat-api/src/lib.rs:648`),
- then `create_dir_all` and `File::create` truncate it and fill it with
- PNG bytes (`lib.rs:950-964`); `export.run` does the same for anything
- FFmpeg can name a container for. Not code execution, but silent
- destruction of credentials and shell profiles by anyone with the token.
- Two neighbours: with neither `HOME` nor `USERPROFILE` set the root list
- is empty and `writable` treats empty as unconfined (`lib.rs:244`), the
- opposite of the comment at `concat-server/src/lib.rs:79`; and
- `project.open` takes any readable folder, after which `project.save`
- and the mask cache write into it (`lib.rs:454-456, 576-580`), so an
- existing project on a shared volume can be rewritten from the socket.
-
-3. **Every JSON connection leaks a file descriptor, and unauthenticated
- callers hold the seats.** (verified) `json.rs:106-122` (TCP) and
- `161-177` (Unix) clone the stream into a closure pushed onto
- `connections`; nothing removes it when the connection ends (`json.rs:135`
- shuts the socket, the clone stays) and the list is drained only in
- `Server::shut` (`concat-server/src/lib.rs:284-288`). Any local process,
- token or not, that connects and disconnects 256 times exhausts the
- editor's descriptors: no media opens, no export writes, `accept` spins
- on `continue` (`json.rs:100`). A well-behaved script that connects per
- request leaks the same way. Separately, `Seat::take` runs before the
- handshake (`json.rs:102, 157`), so 64 idle connects refuse the token
- holder with `busy` for each 10 s window, renewably.
-
-4. **Clear cache deletes the media behind reversed and enhanced clips, and
- those copies are never swept.** (verified for reverse) The reversed copy
- lands in `project/cache/reverse-<hash>-<start>-<dur>.mp4`
- (`concat-host/src/reverse.rs:71-93`) and `adopt_reversed` makes it the
- clip's media through `ReplaceClipMedia` (`concat/src/studio.rs:5495-5520`).
- `projects::clear_cache` removes the whole `cache/` folder
- (`projects.rs:330-339`), from Settings (`studio.rs:6202-6216`), so a
- reversed clip is missing media at the next open; the original import is
- untouched but the clip no longer points at it. Enhance copies are named
- into the cache the same way (`enhance.rs:57-76`). A new file is written
- for every different trim (the key holds start and duration), at CRF 16,
- and nothing sweeps them; `proxy::sweep` touches only `cache/proxy/*.mp4`.
- Processed media belongs under `media/`, or the sweep and Clear cache
- have to know what a clip points at.
-
-5. **0.2.4 documents with reversed clips or preset animations open silently
- changed.** (verified) `Clip` lost `reverse` and the four `animation_*`
- fields, `animation.rs` (800 lines) and `SetClipAnimation` went, and
- `MIGRATIONS` is still two steps (`concat-project/src/doc.rs:62-71`).
- `Session::open` surfaces nothing; the retired-effects path gets a
- notice, this does not, and CHANGELOG's "Changes to note" does not say
- it. A 0.2.4 project with a reversed shot opens, previews and exports
- forward with no message; the flag rides in `extra` and is written back
- forever (only a menu arm still reads `"reverse"`).
-
-### Medium
-
-6. **The Remote page's embedded API still opens a second GPU device inside
- the window.** (verified path; the crash itself is #202's) `Api::new`
- builds `Monitor::new()` (`concat-api/src/lib.rs:179`), and an
- `export.run` goes `ExportJob::run` → `export::run` → `run_on(request,
- None, …)` (`concat-host/src/export.rs:89-95`) → `best_compositor()` →
- `WgpuCompositor::new()` (`concat-export/src/lib.rs:1094-1130`); previews
- use the static `headless()` device. The window's own export and cards
- use `monitor.sibling()` (`concat/src/panes/export.rs:353`,
- `studio.rs:6393`). So a remote export or preview on the NVIDIA laptop of
- #202 takes the road the fix closed. The CLI and a headless API rightly
- use their own device or fail with `NO_RENDERER`.
-
-7. **`[[replaces]]` upgrades drop keyframes, are silent, and skip user
- packages and templates.** (verified) `Editor::upgrade_links` rewrites
- links before any edit and outside undo (`concat-project/src/editor.rs:89-107`);
- `Session::open` ignores its `bool` (`session.rs:104-105`): no dirty flag,
- no notice. `Package::replacing` returns `keys: BTreeMap::new()`
- (`concat-effects/src/catalogue.rs:585-591`, admitted at 553-556). An
- animated Green Screen opens as a constant Chroma Key; the file is
- intact until the next save, then the old link and its keys are gone.
- Only `Catalogue::builtin()` is consulted, so a user package's
- `[[replaces]]` never fires; `templates.rs:323` builds an editor without
- the pass. No test covers `upgrade_links`.
-
-8. **HDR proxies are not regenerated for media imported before 0.2.5.**
- (verified in the open path) The HDR flag in the proxy name comes from
- `item.color_space.is_hdr()` (`studio.rs:6123, 2481`, `proxy.rs:72-74`),
- and `MediaItem::color_space` is absent from a document written before it
- existed and reads SDR (`model.rs:171-176`). For an HDR file imported
- under 0.2.4 the old SDR proxy is kept by the sweep, still adopted, and no
- HDR proxy is queued; no re-probe of existing media was found. CHANGELOG
- says the tone-mapped proxies "are made again".
-
-9. **The Hub does heavy work in line, and a panic in any verb leaves a
- zombie server.** (verified) `hub.rs:83-87` has no `catch_unwind`; a
- panic in `dispatch` ends the API thread and every later call returns
- "the server is shutting down" while the listeners keep accepting and
- the Remote page says it is listening. The one thread still composites
- previews, paints titles, runs every `media.probe` and copies template
- files (`concat-api/src/lib.rs:326, 465, 485, 632, 689`,
- `templates.rs:368`). Also: a subscriber whose 256-line outbox fills is
- dropped on the first `false` with no signal (`hub.rs:58-66`,
- `json.rs:319-324`); progress goes out every 15 frames, so a client that
- reads nothing for about a minute never sees `export.done`.
-
-10. **Downgrade is one click with no gate.** (verified) The list keeps
- every stable release from 0.2.5 (`updates.rs:141-143`) and Install runs
- the chosen one (`settings.rs:447-497`, `ui/dialogs/settings.slint:1090-1105`)
- with no "older than this build" check. An older release refuses a
- document only when `DOCUMENT_VERSION` was bumped; a newer document at
- the same version opens with its features silently absent.
-
-11. **The model has no ceiling on frame size or rate.** (verified for the
- model) `SetTimelineVideo` accepts any positive value
- (`commands/timelines.rs:54-57`, `model.rs:1789-1791`); the ceilings live
- in the window (`FRAME_SIDES 16..8192`, `studio.rs:243-246`) and the API
- (`MAX_SIDE`). A hand-edited document with `"width": 100000` opens; the
- compositor's texture limit is 8192 (`concat/src/gpu.rs:89`). §6's
- "clamps live in `Clip::tidy`" has no counterpart for the timeline's
- frame.
-
-12. **Every undo snapshot deep-copies the bin.** (path verified) `editor.rs:141`
- clones the project; `Project.media: Vec<MediaItem>` is not behind `Arc`
- (`model.rs:1812-1813`); `Batch` clones again per nesting level
- (`commands/mod.rs:994`). 200 snapshots of a 2 000-item bin is 400 000
- `MediaItem` copies, against the module doc's "a snapshot copies
- pointers" (`editor.rs:6-10`).
-
-13. **The export is still one thread, and playback's two old costs
- remain.** (verified) Decode → composite with a GPU readback →
- `write_frame` on one thread (`concat-export/src/lib.rs:1204-1330`);
- `ExportClip` has 40 public fields and `BuiltTimeline` 12 side tables
- (`resolve.rs:88-118`). `audio_pieces` still cuts a curved clip into up
- to 400 pieces, each its own decode (`lib.rs:1005-1007`,
- `studio.rs:2124-2136`); the mixer loops every clip per output sample
- (`playback.rs:1032-1055`).
-
-### Low
-
-14. **A document names its font by absolute path into the app's own
- folder** (`CustomFont { family, path }`, `model.rs:1797-1806`, installed
- to `config/fonts/`, `concat/src/presets.rs:317-331`). On another machine
- `Fonts::add_file` fails silently (`titles.rs:211-214`) and the title is
- drawn in a fallback face with no notice, unlike missing media. No
- traversal: install uses `file_name()` and the document path is only read.
-
-15. **Small correctness in concat-project.** `has_non_finite` misses the
- three `TextStyle` fields added since the 23rd (`commands/mod.rs:878-889`
- vs `model.rs:885-892`), so a NaN is not refused, only replaced by tidy;
- `SetClipKey` and `AddCutoutStroke` can record an undo step that changes
- nothing (`properties.rs:169-173, 136-139`); `follow_first_hdr` flips a
- timeline back to HLG after the person set it to SDR, removed the HDR
- clip and dropped one again (`editor.rs:238`); `VideoSettings`,
- `TextStyle` and six other structs have no `extra` map, so an unknown
- field inside `video` or `text` is dropped on round trip and an unknown
- `colorSpace` reads SDR (`model.rs:1056-1059, 1717-1735`); `to_document`
- stamps `"concat": "0.1.0"` (`doc.rs:317`).
-
-16. **Small hygiene in the server and CLI.** The Unix socket is created at
- the umask's mode (`concat-server/src/lib.rs:196`); `connections()`
- over-counts hung-up callers; `export.progress.stage` is documented as
- "video", "audio", "mux" (`message.rs:591`) and emitted as "rendering",
- "mixing audio", "muxing" (`concat-export/src/lib.rs:909, 914, 1333`);
- `project.create`'s `video` is unbounded (`lib.rs:386-393`).
-
-17. **Small hygiene in the host.** `updates::installed_as()` runs `rpm -q`
- synchronously on the UI thread once per run (`settings.rs:796-797`,
- `updates.rs:248-267`); export temporaries are predictable siblings of
- the output with no exclusivity (`lib.rs:887-888`); two TLS stacks are
- linked (ureq 2 with rustls and webpki roots, so a TLS-inspecting proxy
- breaks updates and model downloads; ureq 3 with native-tls through
- `ort-sys`).
-
-18. **Carried from the 23rd.** Five ⌘ glyphs in the UI; `ui/demo/` still
- imported; `core::Project` dead; `Path::exists` in the wasm crate; the
- speed range in five places; `src/README.md:114` on `mobile.yml`.
-
----
-
-## 7. What is good, and should stay that way
-
-- **The layering, still.** Every arrow points one way; FFmpeg is only in
- concat-media; the CPU compositor is a test oracle and nothing else.
-- **`Session::open`'s three-way split** (none, corrupt, newer) with two
- refusal tests (`session.rs:73-100, 341-369`). The critical of the 23rd,
- closed with a test that pins it.
-- **`clips_where` + `tidy_touched`.** Sharing preserved by construction and
- one clamp function used both on load and after every command
- (`model.rs:1972`, `editor.rs:170`).
-- **`follow_first_hdr` as an editor-level diff.** One undo step for free,
- correct through `Batch`, gesture folding, undo and redo, with a test that
- undo takes the clip and the colour back together (`lib.rs:144-192`).
-- **Mandatory digests with pinned upstreams**, one `download`/`verify` pair
- shared by models and the updater (`models.rs:201-310`).
-- **The updater's input hygiene.** Drafts and pre-releases filtered, the tag
- validated before it becomes a URL, a file name with a folder refused,
- cancel and timeouts everywhere, and a fallback to "open the package" when
- the swap cannot be done (`updates.rs:119-161, 361-368, 483-511`).
-- **`Api::writable` + `resolved`.** Refuse `..`, canonicalise the deepest
- existing ancestor, compare component-wise, with a test that walks a
- not-yet-existing folder (`concat-api/src/lib.rs:243-275, 910-929, 1384-1435`).
-- **`OpenProjects`** as a shared, cheap-to-clone register with
- claim-or-name-the-holder semantics and release in `Drop`.
-- **`SingleFlight`.** One slot per long job, released on drop even through
- a panic, cancel reaches only the running job (`jobs.rs:18-111`).
-- **The lent compositor**, the adapter named in the log at the first frame,
- and an end-to-end test that exports on a sibling device
- (`concat-export/src/lib.rs:1129-1140`, `tests/export.rs:701`).
-- **Reverse's two-pass, bounded-memory design** and partial-then-rename
- landing (`reverse.rs:63-65, 167-177, 219-291`).
-- **The tolerant `wire` readers**, per-struct `extra` maps and the
- `WRITER_OWNED` list: a hand-edited file degrades instead of failing.
-- **One contract, two transports**, a constant-time token with the test
- that pins its promise, and `version.capabilities` extended by the server
- with what is actually listening.
-- **A hand-written CHANGELOG** grouped by where the person notices the
- change, and CI that gates on fmt, clippy, perf and wasm on every push.
-
----
-
-## 8. Where we stand
-
-The 23 September audit ended with seven recommended steps. Five days later:
-
-| Step (23 Sept) | Landed |
-|---|---|
-| 1. Correctness trio (open, scheduler, epoch) | open and scheduler yes; epoch exists, coverage to confirm |
-| 2. Server limits, root allowlist, shared exporter | yes, with the home folder as the root and two holes (findings 2, 3) |
-| 3. Fill every digest, pin, refuse empty | yes, in full |
-| 4. Trial at install, error scopes, source-hash keys | trial and error scopes yes; key scheme to confirm |
-| 5. `clips_where`, one clamp pass, `start_time`, Darken/Lighten | all four, with the speed range still copied |
-| 6. Art refresh on change, sweeps, graph caching | art yes; proxy swept; art, peaks, reverse, enhance not |
-| 7. CI matrix, perf gate, release on CI, docs | matrix, perf and release yes; parity and docs partly |
-
-And on top of that list, the HDR plan's phases 0 to 3 shipped: one GPU
-compositor in linear light, format 2 packages with probes and multi-pass,
-HDR in and out, scopes, and the retirement of the old effects. That is the
-week's real story: the picture pipeline is now one thing, and the audit's
-render and effects findings of the 23rd (CPU kernels for 147 shaders,
-unbounded GPU memory, chains never pixel-compared) are gone with the code
-that carried them.
-
-What the week bought in debt is of two kinds. The first is the feature
-that shipped fastest: the self-update is well built at every step except
-the one that matters, the trust root, and it should not sit in a release
-for long without a signature. The second is the seam between old and new:
-pre-0.2.5 documents with reversed or animated clips, pre-0.2.5 HDR imports,
-and `[[replaces]]` links with keys all open changed and say nothing. Both
-are cheap next to what was paid this week.
-
-**Recommended order.** Each line is a day or a few, and each closes a
-finding above.
-
-1. Sign the update manifest with a key compiled into the app and refuse an
- unsigned one; keep the quarantine bit; confirm a downgrade; sign the
- Windows installer and notarise every macOS build. (1, 10)
-2. Remote page defaults: root the API at the project and export folders,
- not the home folder; refuse to start with no root; check `writable` in
- `project.open` or `save`; count seats after auth; drop the connection
- clone when it ends; `catch_unwind` round the Hub's dispatch; hand the
- embedded API the window's device through `sibling()`. (2, 3, 6, 9)
-3. Processed media out of `cache/`: put reverse and enhance copies under
- `media/` (or teach Clear cache and the sweep what clips point at); sweep
- `cache/art` and peaks; re-probe pre-0.2.5 media for its signal at open.
- (4, 8)
-4. Tell the person: a migration step for `reverse` and the animations
- (turn the flag into the new Reverse, or name the clips in the open
- notice); surface `upgrade_links` and carry the keys; run it for
- templates and user packages. (5, 7)
-5. Finish the 23 September list: an epoch check on every worker path; a
- `tidy` for the timeline's frame; one speed range; the CLI's socket;
- `ui/demo/`; the five ⌘ glyphs; the `mobile.yml` sentence. (11, 15, 18)
-6. Pipeline the export (decode, composite, encode on three threads) and
- give the mixer an index; `Arc` the bin in snapshots. (12, 13)
-7. Speech, enhance, proxies and CBR as API verbs, so "three ways in" holds
- for everything the window does. (19 of the 23rd)
-
-After those, phase 4 of the HDR plan (a true HDR preview), then phase 5
-(zero-copy hardware frames, hardware encoders) in the plan's order.
-
----
-
-## 9. Caveats
-
-Scores are judgements over code, calibrated against the 23 September scores
-but not measured against users or another editor. Rows marked *provisional*
-in §3 and the three findings marked *pending* in §5 rest on hand checks of
-the specific lines named, not on a full read; they are replaced as those
-reviews land. Findings marked (verified) were read end to end by the reviewer
-and the headline ones re-read by hand; the rest rest on one reviewer's
-reading and a line number you can open. The app was not launched: "user
-likeability" is judged from what the UI code does, not from users.
diff --git a/audit/AUDIT-2026-10-04-QUICK.md b/audit/AUDIT-2026-10-04-QUICK.md
deleted file mode 100644
index 0cd59bdd..00000000
--- a/audit/AUDIT-2026-10-04-QUICK.md
+++ /dev/null
@@ -1,175 +0,0 @@
-# Concat audit, 4 Oct 2026 — the quick version
-
-Same audit as `AUDIT-2026-10-04.md`, one screen. Each bar has 10 slots:
-🟢 filled = good, ⚪ empty = missing. Severity: 🔴 critical · 🟠 high · 🟡 medium · ⚪ low.
-Arrows are against the 28 Sep audit. Every crate was read in full; nothing is provisional.
-
-## 🎯 Verdict
-
-**A good week for the window, a lost week for the gate.** Six days and 60 commits after 0.2.5, the app gained a razor, a scrollbar, a volume line, keyboard menus, Save Frame and Save Audio, shapes, an export sheet with a picture, hardware encoders, and an export that survives a dead GPU, each well made at its seam. But `main` has not been green since the release: fmt has been red on every push and nothing behind it has run for 71 commits; none of the seven steps recommended on the 28th was taken; and three of the old findings got worse while the engine sat still. No new critical. Four new highs: the ignored gate, a `[[replaces]]` that lets a user package hijack a built-in, preview sound off by the stream start on MTS files, and eight locales a fifth English.
-
-## 📏 Size
-
-| | Lines | Δ vs 28 Sep |
-|---|---:|---:|
-| Rust, all crates | 87 376 | +3 800 |
-| Slint UI | 28 796 | +1 450 |
-| WGSL, packages | 1 764 | +20 |
-| Tests (`#[test]`) | 676 | +26 |
-| Largest file `concat/src/studio.rs` | 10 029 | +568 |
-
-## 📊 The whole app
-
-| Aspect | Bar | 28 Sep | Today |
-|---|---|:-:|:-:|
-| 🧹 Code quality | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 7 | 6 ↓ |
-| 🏛️ Architecture | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 |
-| 🧭 Design philosophy | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 7 | 6 ↓ |
-| 🔧 Maintainability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 6 | 5 ↓ |
-| 😊 User likeability | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 |
-| ⚡ Performance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 📈 Scalability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 |
-| 🔐 Security | 🟢🟢🟢🟢⚪⚪⚪⚪⚪⚪ | 5 | 4 ↓ |
-| 🧪 Testing | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 7 | 6 ↓ |
-| 📚 Docs | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 7 | 6 ↓ |
-| 🤖 CI and release | 🟢🟢🟢⚪⚪⚪⚪⚪⚪⚪ | 6 | 3 ↓ |
-
-## 🧱 Per crate (average of its scales)
-
-| Crate | Lines | Δ | Bar | Avg | One thing |
-|---|---:|---:|---|:-:|---|
-| concat-vision | 2 760 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.2 | 🟢 untouched since the 23rd · ⚪ runtime wrapper untested |
-| concat-text | 1 974 | +734 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.1 | 🟢 faces mmapped once behind `Arc` · 🟡 emoji are boxes on a Mac |
-| concat-effects | 7 096 | +6 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.4 ↓ | 🟢 344 probes · 🟠 `[[replaces]]` can hijack a built-in |
-| concat-core | 2 854 | +25 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.3 | 🟢 `hold` is one flag in two places · ⚪ dead `Project`, i64 rate panics |
-| concat-server | 1 376 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.0 | 🟢 constant-time token · 🟠 fd leak, zero churn |
-| concat-api | 2 351 | +1 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.9 ↓ | 🟢 real roots check · 🟡 four window-only operations now |
-| concat-project | 9 465 | +273 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.9 | 🟢 shapes tidied in the one right place · 🟡 0.2.5 deletes shapes |
-| concat-render | 8 124 | +80 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.6 | 🟢 parity for 10 plan features · 🟡 a timed-out trial kills the monitor |
-| concat-media | 9 494 | +453 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.8 | 🟢 37 `unsafe`, 37 SAFETY lines · 🟠 playback audio ignores the stream start |
-| concat-speech | 3 109 | +15 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.4 ↑ | 🟢 one accelerator switch, one fallback · ⚪ `DirectML.dll` not staged |
-| concat (Slint) | 28 796 | +1 450 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.5 | 🟢 0 ⌘ glyphs now · 🟡 no dialog a keyboard can press |
-| concat-cli | 514 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.2 ↓ | 🟢 token from env · 🟡 no logger, so export diagnostics vanish |
-| concat-export | 3 702 | +204 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.1 | 🟢 device loss recovered, no partial file · 🟡 hold plays keyed effects early |
-| concat-host | 11 869 | +329 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.8 | 🟢 logs with a level of their own · 🟠 update trust chain unchanged |
-| concat-android | 335 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 | 🟢 portrait locked · ⚪ 0 tests, ships Chatterbox |
-| concat (Rust) | 21 261 | +1 697 | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5.4 ↓ | 🟢 epoch gate on 10 of 12 workers · 🟡 `studio.rs` 10 029 lines, a full publish per frame |
-
-## 🎛️ Per feature
-
-| Feature | Bar | 28 Sep | Today | State |
-|---|---|:-:|:-:|---|
-| Launcher | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Import and bin | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ Wayland drop |
-| Timeline editing | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ ★ razor, ★ scrollbar |
-| Undo / redo | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Effects and filters | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ shapes round, glows past white |
-| LUTs | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Relink | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Themes | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ lime default |
-| Transitions | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 8 | 7 ↓ | ⚠️ ★ dissolve hold; keyed effects play early |
-| Colour and HDR | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Scopes | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ verified this time |
-| Colour grading | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Titles | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ ★ wraps at the frame; emoji boxes on macOS |
-| Keyframes | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Crop, blend, masks | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Cutout | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Captions | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ words off by the stream start on MTS |
-| Text-to-speech | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ ★ DirectML with CPU fallback |
-| Templates | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Export | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ ★ survives a dead device; rate ladder fixed 24/30/60 |
-| Hardware decode | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| ★ Logs | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | – | 7 | ✅ a file per run, bursts folded |
-| ★ Save Frame / Save Audio | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | – | 7 | ✅ epoch-gated; Save Audio window-only |
-| Custom packages | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 7 | 6 ↓ | ⚠️ trial at defaults only; `[[replaces]]` unfenced |
-| ★ Shapes / Stickers | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ not editable; 0.2.5 deletes them |
-| ★ Volume line | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ can drop typed title words |
-| ★ Hardware encode | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ never exercised on a chip in CI |
-| ★ Keyboard menus | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ stalls past four dead rows |
-| Languages | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 8 | 6 ↓ | ⚠️ 8 locales a fifth English; no RTL |
-| Remote API | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ one line changed; fell behind the window |
-| Enhance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ copies in `cache/` |
-| Phone shell | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ ★ swipe to close; its own 44 px rule broken |
-| ★ Windows data folders | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ a failed move is silent |
-| Proxies | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ HDR copy under an SDR name |
-| Playback and audio | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 6 | 5 ↓ | ⚠️ sound ahead of picture on MTS |
-| Self-update | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ ★ `.msi` path; same trust chain |
-| Accessibility | 🟢🟢⚪⚪⚪⚪⚪⚪⚪⚪ | 2 | 2 | ❌ two `accessible-*` in 28 796 lines |
-
-## 🔁 The 28 Sep findings today
-
-**0 closed · 4 partly closed · 14 open**, three of them worse than described.
-
-| # | Was | Now |
-|:-:|---|---|
-| 1 | update trust chain | 🟠 open; Windows signed only when a secret is set |
-| 2 | API rooted at the home folder | 🟠 open, byte-identical |
-| 3 | fd leak per connection | 🟠 open, byte-identical |
-| 4 | Clear cache deletes reversed media | 🟠 open |
-| 5 | 0.2.4 documents open changed | 🟠 open |
-| 6 | embedded API's second GPU device | 🟡 open |
-| 7 | `[[replaces]]` drops keys | 🟠 **worse**: user packages' tables do fire, and can name a live built-in |
-| 8 | HDR proxies not regenerated | 🟡 **worse**: an HDR copy written under the SDR name |
-| 9 | Hub in line, no `catch_unwind` | 🟡 open |
-| 10 | downgrade one click | 🟡 open |
-| 11 | no frame ceiling | 🟡 **sharper**: an i64 rate panics the export thread |
-| 12 | bin cloned per snapshot | 🟡 open |
-| 13 | export one thread; 400 pieces | 🟡 partial: the cut moved to one place in concat-export |
-| 14 | font by path, silent | ⚪ open |
-| 15 | small project correctness | ⚪ partial: no-op undo steps fixed; `ShapeStyle` joins the no-`extra` list |
-| 16 | server hygiene | ⚪ open |
-| 17 | host hygiene | ⚪ partial: `rpm -q` memoised |
-| 18 | carried hygiene | ⚪ partial: ⌘ glyphs gone, `mobile.yml` sentence was true; `ui/demo/`, `core::Project`, speed range remain |
-| – | two locale-flaky tests | ⚪ open, one failed again today |
-
-## 🆕 New findings
-
-| # | Sev | What | Where |
-|:-:|:-:|---|---|
-| 1 | 🟠 | `main` red at `cargo fmt --check` since 28 Sep; clippy, tests and perf have not run on CI for 71 commits; CI builds on 1.99, the pin says 1.93 | `ci.yml:93` · 10 files, 22 sites |
-| 2 | 🟠 | A user package's `[[replaces]]` can name a live built-in and rewrite every link to it at open, keys dropped, no notice | `catalogue.rs:1050,986` · `manifest.rs:91` · `session.rs:105` |
-| 3 | 🟠 | Preview sound and the transcript ignore the stream start; the export does not: sound runs ahead on MTS files | `samples.rs:91,176` vs `audio.rs:424` |
-| 4 | 🟠 | Eight locales have 128–197 values that are the English string; the check compares keys only | `scripts/locales.py:134` · `de.json` |
-| 5 | 🟡 | Shapes: 0.2.5 opens them as Video, drops them, deletes them on save at the same `DOCUMENT_VERSION`; unknown kinds become squares; not editable after placement; 17 of 44 commands outside the round-trip test | `model.rs:1159,1184` · `doc.rs:41,266` · `commands/mod.rs:143` |
-| 6 | 🟡 | Install trial runs at the knobs' defaults; a timed-out trial sets `dead` on the monitor's own compositor | `catalogue.rs:489` · `gpu.rs:1869` · `preview.rs:120` |
-| 7 | 🟡 | Dissolve hold plays keyed effects over the pre-roll; keys on a clip with a handle play early | `concat-export/src/lib.rs:776,803,1478` |
-| 8 | 🟡 | Export rate is a fixed 24/30/60 ladder; a 25 fps cut exports at 30 by default | `studio.rs:154` · `export.slint:363` |
-| 9 | 🟡 | A volume-line drag within 120 ms of a title blur drops the typed words | `studio.rs:4446,6213,4634` |
-| 10 | 🟡 | Art caches never released across projects; `spawn_art`, `spawn_strip` and `save` not epoch-gated; `close_project` never cancels an export | `studio.rs:6626,6739,2822,2310` |
-| 11 | 🟡 | A full publish per monitor frame during playback | `host.rs:243` · `monitor.rs:257` |
-| 12 | 🟡 | No dialog button or card is keyboard-reachable; Escape closes no modal | `modal.slint:17,168` |
-| 13 | 🟡 | Six strings and the effects category bar bypass `I18n` | `media-pane.slint:1332,2197,2239` · `tray.slint:182` |
-| 14 | 🟡 | Emoji paint as boxes in titles on macOS | `concat-text/src/lib.rs:432,530` |
-| 15 | 🟡 | No Unreleased section; version still 0.2.5; ~20 user-visible changes unrecorded | `CHANGELOG.md` · `src/Cargo.toml:6` |
-| 16 | 🟡 | +3 800 lines, 26 tests; `tests/export.rs` untouched | `git log --numstat` |
-| 17 | 🟡 | README sells an MCP API that does not exist | `README.md:62,79,114` |
-| 18 | 🟡 | The CLI installs no logger; device-loss diagnostics vanish through `serve` | `concat-cli/src/main.rs` |
-| 19 | 🟡 | API one line changed; Save Audio, log level, CBR and the hardware-encoder choice are window-only; `export.run` cannot refuse a chip | `concat-export/src/lib.rs:1277` · `concat-api/src/lib.rs:568` |
-| 20–28 | ⚪ | NaN check after the clone; scopes freeze on a failed map; `transitions.rs` oracle unchecked; no fallback once a chip encoder opened; "· hardware" for Media Foundation; greys-only colour tests; silent data-folder move; `msiexec` on a hand-unpacked zip; poster under the wrong name; extension swapped after the overwrite prompt; four copies of the inspector ranges; menu walking stalls past 4 dead rows; 28 colours outside `theme/`; phone 34/40 px buttons; `in-chain` capped at 8; gRPC without caps; Chatterbox on the phone; `DirectML.dll` not staged; AVX baseline CI-only; signtool password on the command line; `SECURITY.md` silent on the updater | see the full audit §6 |
-
-## 🧾 Today's checks
-
-| Check | Result |
-|---|---|
-| fmt | ❌ **10 files would change** (0 on 28 Sep): `studio.rs`, `concat-export/lib.rs`, `dirs.rs`, `titles.rs`, `encode.rs`, `concat-project/{lib,commands/mod,commands/clips}.rs`, `gpu.rs`, `concat-text/lib.rs` |
-| clippy | ✅ 0 warnings |
-| tests | ⚠️ **672 passed, 1 failed**, 1 ignored (674 in all; 650 on 28 Sep). The failure is `panes::speech::tests::a_voice_name_reads_as_a_person_would_say_it`, the locale race of the 28th, unfixed. E2E 11/11 in 31 s, parity 88/88 in 55 s. |
-| perf --check | ✅ 22 of 22 within budget. Model and GPU rows flat or better (composite 1.9 ms, blur 6.3 / 9.8 ms); every decode and export row 20–50 % below the 28th (software 1080p h264 809 fps vs 1 626; 720p export 157 fps vs 204; HDR export 82 vs 111) on a decode path that did not change, so most likely the machine: re-run on a quiet one before reading it as a trend. |
-| CI on `main` | ❌ **0 green runs since v0.2.5** (`b78ee56`, 28 Sep 14:19 UTC): 11 failed at fmt, 29 cancelled by the next push. Nix (build only) green 2 Oct. A release tag today would fail at CI and ship nothing. |
-
-## 🗺️ Do next, in order
-
-1. 🟠 **Go green** — `cargo fmt`, push, watch; put the toolchain pin into CI; isolate the two locale tests; write the Unreleased section; bump to 0.2.6.
-2. 🟠 **Fence `[[replaces]]`** — refuse one naming a live package at install; carry keys; tell the person when `upgrade_links` changed a document; trial at `params_at(Max)` on a `sibling()` compositor.
-3. 🟠 **Stream start in playback audio** — `AudioDecoder::open` takes `start_of`, seeks and trims at `start + offset`; an MTS fixture with sound.
-4. 🟡 **Shapes across versions** — bump `DOCUMENT_VERSION`; keep an unknown kind in `extra`; a `shape` on `ClipPatch`; all 44 commands in the round-trip test.
-5. 🟡 **Say it in the person's language** — `locales.py --check` flags English values; the six literals and the category bar through `I18n`; the export rate from the timeline; effect keys frozen at the hold's head.
-6. 🟡 **The window's seams** — flush before `clip_volume`; epoch on art, strips and save; clear art tables at close; lanes-only publish for monitor frames; `FocusScope` on `Button`, Escape on `Modal`.
-7. 🟠 **Then the 28 Sep list, in its order** — sign the update; the Remote page's root, seats and fd leak; processed media out of `cache/`; the migration notices; the export pipeline; the API verbs.
-
-Then HDR phase 4 (a true HDR preview) and phase 5 (zero-copy, Mac hardware encoders).
-
-## ✅ Keep
-
-Layering · `recovered` with no partial file · `hold` as one flag in two places · shapes tidied in the one right place · the scheduler's `Finished` guard · `hardware_wins` with its numbers written down · 37 `unsafe`, 37 SAFETY lines · `wayland_drop.rs` as a careful guest · one epoch gate with a test · dialogs deferred to the next loop turn · monitor request coalescing · 344 probes in half floats · logs with a level of their own · faces mapped once · `.part` → staging → rename · `token::matches` · `locales.py` building `en.json` from the source · the release gate order · parity on lavapipe.
diff --git a/audit/AUDIT-2026-10-04.md b/audit/AUDIT-2026-10-04.md
deleted file mode 100644
index afe3b177..00000000
--- a/audit/AUDIT-2026-10-04.md
+++ /dev/null
@@ -1,806 +0,0 @@
-# Concat audit, 4 October 2026
-
-Where the codebase stands six days after the 28 September audit, at
-`a1e8f11` (3 October), 60 commits past the 0.2.5 release. Same method, same
-seven scales, same nine reviewers' split, and every one of the 18 findings
-from the 28th (and the handful carried from the 23rd) re-verified against
-today's tree. Every finding names a file and a line. The sibling
-`AUDIT-2026-10-04-QUICK.md` is this audit as a one-screen scorecard.
-
-**How it was done.** Nine reviewers each read one part of the tree in full
-(core and project; media; render and effects; export and host; the window's
-Rust; the Slint UI and locales; the API, server and CLI; speech, vision,
-text, Android and the model manifest; tests, CI, build, docs and release),
-re-verified the 28 September findings assigned to them, scored their part
-and reported new findings with locations. The headline items (the
-`[[replaces]]` hijack, the playback audio offset, the red CI) were re-read
-by hand against the code before they went in here. Line counts, test counts
-and the unsafe, unwrap and panic tallies come from the tree;
-`cargo fmt --check`, `cargo clippy --workspace --all-targets`,
-`cargo test --workspace` and `cargo run -p concat-perf --release -- --check`
-were run on this machine (Apple Silicon, macOS 26). The app itself was not
-launched.
-
-**Status of this document.** Complete: all nine reviews are in and folded
-in below. No row is provisional.
-
-**Scale.** 0 to 10. 8 and up means "keep doing this"; 5 to 7 means "works,
-with known debt"; 4 and under means "will bite at the next order of
-magnitude, or already bites".
-
----
-
-## 1. The numbers
-
-### 1.1 Size, tests and hygiene per crate
-
-Deltas are against the 28 September table. The unwrap column gives the
-total and, in brackets, the count outside test modules and test files; the
-panic column is the non-test count.
-
-| Crate | Rust lines | Δ | Slint / WGSL | Tests (files) | `unsafe` | `unwrap`/`expect` (non-test) | `panic!` family (non-test) |
-|---|---:|---:|---:|---:|---:|---:|---:|
-| concat (window) | 21 261 | +1 697 | 28 796 slint | 64 (16) | 12 | 30 (6) | 0 |
-| concat-host | 11 869 | +329 | | 79 (17) | 1 | 230 (2) | 0 |
-| concat-project | 9 465 | +273 | | 99 (1) | 0 | 319 (11) | 6 |
-| concat-media | 9 494 | +453 | | 90 (11) | 37 | 181 (23) | 0 |
-| concat-render | 8 124 | +80 | shaders inline in `gpu.rs` | 88 (6) | 1 | 90 (8) | 4 |
-| concat-effects | 7 096 | +6 | 1 764 wgsl | 69 (9) | 0 | 129 (10) | 2 |
-| concat-export | 3 702 | +204 | | 29 (3) | 0 | 24 (0) | 0 |
-| concat-speech | 3 109 | +15 | | 19 (3) | 3 | 23 (2) | 0 |
-| concat-core | 2 854 | +25 | | 48 (7) | 0 | 26 (4) | 2 |
-| concat-vision | 2 760 | 0 | | 31 (8) | 0 | 22 (2) | 0 |
-| concat-api | 2 351 | +1 | | 21 (2) | 0 | 23 (0) | 0 |
-| concat-server | 1 376 | 0 | | 12 (4) | 0 | 29 (2) | 0 |
-| concat-text | 1 974 | +734 | | 22 (1) | 1 | 53 (1) | 0 |
-| concat-perf | 1 092 | 0 | | 2 (1) | 0 | 34 (34) | 0 |
-| concat-cli | 514 | 0 | | 3 (1) | 0 | 2 (2) | 0 |
-| concat-android | 335 | 0 | | 0 (0) | 10 | 9 (9) | 2 |
-| **Total** | **87 376** | **+3 800** | **~30 600** | **676 (90)** | **65** | **~1 200 (116)** | **16** |
-
-The tree grew by a twentieth in six days, almost all of it in the window
-(`concat/src` +4 148 / −917 and the UI +1 291 / −168) and in concat-text,
-which took on shapes and grew by three fifths. `studio.rs` is now 10 029
-lines (9 461 on the 28th; 47 % of the window's Rust) and was not split.
-The engine crates barely moved: concat-api gained one line, concat-server,
-concat-cli, concat-vision and concat-perf none. Twelve new `unsafe` sites
-are the Wayland drop target (`concat/src/wayland_drop.rs`, 9) and the Linux
-window edges, every one with a SAFETY line.
-
-Other facts: 123 effect packages (98 on format 2 with WGSL: 21 effects, 57
-looks, 20 transitions, with 344 probe tables between them; 25 audio
-packages on FFmpeg chains, by design); 14 locale files at 1 001 keys each,
-0 missing and 0 stale per `scripts/locales.py --check` (but see finding 4);
-801 commits since 25 August, 60 of them since the last audit (509 files,
-+10 700 / −1 600; all 60 by the maintainer, no merges); 29 tags, latest
-`v0.2.5`, 71 commits behind `main`; largest files `concat/src/studio.rs`
-10 029, `concat-project/src/lib.rs` 3 851 (3 800 of them tests),
-`concat-render/src/gpu.rs` 3 630, `concat-export/src/lib.rs` 2 873,
-`concat-effects/src/shader.rs` 2 412; largest Slint files
-`workspace/media-pane.slint` 2 425 and `inspector/clip-inspector.slint`
-1 959. `Cargo.lock` holds 817 packages, 63 crate names at two or more
-versions.
-
-### 1.2 The checks, run today
-
-| Check | Result |
-|---|---|
-| `cargo fmt --all -- --check` | **10 files would change, 22 sites** (0 on the 28th) |
-| `cargo clippy --workspace --all-targets` | 0 warnings |
-| `cargo test --workspace --no-fail-fast` | 672 passed, 1 failed, 1 ignored (674 in all; 650 on the 28th) |
-| `cargo run -p concat-perf --release -- --check` | 22 of 22 scenarios within budget |
-| CI on `main` | **red at head, and at every run since the 28th** |
-
-**fmt is red, and so is CI.** The ten files are `concat/src/studio.rs`,
-`concat-export/src/lib.rs`, `concat-host/src/dirs.rs` and `titles.rs`,
-`concat-media/src/encode.rs`, `concat-project/src/lib.rs`,
-`commands/mod.rs` and `commands/clips.rs`, `concat-render/src/gpu.rs` and
-`concat-text/src/lib.rs`. The first three sites date from the afternoon of
-the 28th (`35b80d1`, `8d3bf19`, `dc12b64`), after the last audit was
-written; the rest came with `f97c745`, `0699b44` and `21855a7`. The CI
-engine job runs `cargo fmt --check` first (`.github/workflows/ci.yml:93`)
-and has failed there on every completed run since `a262553` (28 September
-22:16 UTC): 11 failures, 29 cancelled by the next push, 0 successes. The
-last green CI run is `b78ee56`, the 0.2.5 release commit. Clippy, the
-locale and model inventories, the 674 tests, the quick perf gate and the
-gRPC build (`ci.yml:98-126`) have not executed on `main` for 71 commits.
-The Nix job, which only builds, is green as of 2 October. The release
-workflow needs CI (`release.yml:66-68`), so a 0.2.6 tag today would produce
-no bundles. Locally, clippy is clean.
-
-**One window test fails locally, the same way as on the 28th.**
-`panes::speech::tests::a_voice_name_reads_as_a_person_would_say_it`
-(`concat/src/panes/speech.rs:678`) read a translated voice detail while the
-i18n tests in the same binary had the process-wide locale on `de`. The
-other test of the pair, `format::tests::phrases_are_coarse`, happened to
-win its race this time. Both pass alone and the window suite passes with
-`--test-threads=1`: the test-isolation bug of the 28th, unfixed. The
-end-to-end export suite (11 tests, 31 s), the parity suite (88 tests,
-55 s) and every engine crate pass.
-
-**The perf table**, release build, every scenario inside its budget. The
-28 September reading is in the fourth column.
-
-| Scenario | Measured | Budget | 28 Sept |
-|---|---:|---:|---:|
-| plan one frame of 200 clips on 4 tracks | 0.54 µs | ≤ 200 µs | 0.96 µs |
-| 200 clip additions and 200 undos | 0.71 ms | ≤ 150 ms | 0.78 ms |
-| ripple delete of 200 clips out of 400 | 0.13 ms | ≤ 20 ms | 0.20 ms |
-| save and open a 200-clip document | 1.67 ms | ≤ 60 ms | 1.98 ms |
-| choose the visible clips of 200 for one publish | 0.21 µs | ≤ 50 µs | 0.21 µs |
-| read an hour's waveform across 2048 columns | 4.23 µs | ≤ 2000 µs | 4.41 µs |
-| decode 1080p h264 in software | 809 fps | ≥ 120 | 1 626 fps |
-| decode 1080p h264 on the hardware (VideoToolbox) | 175 fps | ≥ 120 | 316 fps |
-| decode 4K h264, the app's choice (software) | 115 fps | ≥ 30 | 151 fps |
-| decode 4K HEVC 10-bit, the app's choice (VideoToolbox) | 92 fps | ≥ 30 | 112 fps |
-| decode 4K HEVC 10-bit in software | 31 fps | ≥ 24 | 43 fps |
-| decode 3 streams of 4K HEVC 10-bit at once | 45 fps | ≥ 24 | 61 fps |
-| decode 4K HLG tone-mapped on the CPU, 960×540 | 22 fps | ≥ 15 | 29 fps |
-| decode 4K HLG deep for the GPU, 960×540 | 127 fps | ≥ 30 | 140 fps |
-| scrub forward over 90 frames, cold | 0.95 ms/frame | ≤ 25 | 0.72 |
-| scrub back over covered ground | 0.43 µs/frame | ≤ 200 | 0.25 |
-| decodes when an effect knob is turned | 0 | ≤ 0 | 0 |
-| compose 1080p, 3 layers, a look, on the GPU | 1.92 ms | ≤ 25 | 2.31 ms |
-| Gaussian blur over 4K, radius 10, on the GPU | 6.34 ms | ≤ 33 | 6.11 ms |
-| Gaussian blur over 4K, radius 50, on the GPU | 9.77 ms | ≤ 33 | 9.34 ms |
-| export a 4 s 720p cut of two overlapping clips | 157 fps | ≥ 30 | 204 fps |
-| export the same cut as HDR: HLG, HEVC 10-bit | 82 fps | ≥ 30 | 111 fps |
-
-Two readings worth keeping. The model and GPU rows are flat or a little
-better. Every decode row and both export rows read 20 to 50 % below the
-28th, with the software h264 decode at half. The decode path did not
-change in six days (the only commit in concat-media touched `encode.rs`,
-and `hardware_wins`, the pool and the pacer are byte-identical), so the
-likeliest cause is the machine, not the code: the run came at the end of
-a session that had built the workspace three times. The budgets are set
-far enough from the numbers that nothing tripped, but a row that halves
-for a reason the code cannot explain is a row to re-run on a quiet
-machine before the next audit takes it as a trend. The 1080p hardware
-row still measures a path the app never takes for 8-bit H.264 (the
-"app's choice" rows are the real routing, `hardware.rs:209`).
-
----
-
-## 2. Scorecard for the whole app
-
-| Aspect | 28 Sept | Today | One line |
-|---|:-:|:-:|---|
-| Code quality | 7 | 6 ↓ | Clippy-clean; but fmt is red in 10 files, 6 UI strings bypass i18n, 28 colours are named outside `theme/`, and `studio.rs` grew another 568 lines. |
-| Design architecture | 8 | 8 | Held: one compositor, one plan, the lent device now survives a loss and recovers on a fresh or software adapter; the hold for a dissolve is one flag the renderers agree on. |
-| Design philosophy | 7 | 6 ↓ | Six stated rules not held: "every operation the window can perform is one an API caller can", "nothing outside theme/ names a colour", "every string goes through I18n", "no button below 44 px", "one definition of source time", and ARCHITECTURE's hardware-decode switch that no longer exists. |
-| Maintainability | 6 | 5 ↓ | `studio.rs` 10 029 lines, `concat-export/src/lib.rs` 2 873 with a 41-field `ExportClip`, `ui/demo/` still compiled in for one struct, `tts.rs` 1 244 lines across three model families. |
-| User likeability | 7 | 7 | A razor, a scrollbar, a volume line, keyboard menus, Save Frame As, a scrub strip in the export sheet; against that an export that defaults to 30 fps on a 25 fps cut, eight locales a fifth English, and no dialog a keyboard can press. |
-| Performance | 6 | 6 | Decoded frames go through two CPU copies (three on the chip) before upload; a full publish per monitor frame during playback; the export loop unchanged. |
-| Scalability | 5 | 5 | Art caches never released across projects; the fd leak per API connection, the one Hub thread and the four unswept cache kinds all as they were. |
-| Security | 5 | 4 ↓ | The update trust chain unchanged; new: a user package's `[[replaces]]` can hijack every link to a live built-in at open, and a timed-out install trial kills the monitor's compositor. |
-| Testing | 7 | 6 ↓ | 676 tests, +26 on +3 800 lines; no CI test run for 71 commits; the E2E suite untouched while transitions, titles, shapes and logs gained code; the two locale-flaky tests still race. |
-| Documentation | 7 | 6 ↓ | No Unreleased section for ~20 user-visible changes; the README sells an MCP transport that does not exist; three ARCHITECTURE sentences describe code that changed. |
-| CI and release | 6 | 3 ↓ | Zero green runs since 0.2.5; the gate fired on every push and was ignored; CI builds on `stable` (1.99) while `rust-toolchain.toml` pins 1.93; signing still only when secrets are set. |
-
----
-
-## 3. Scores per crate
-
-Columns: code quality, maintainability, scalability, performance,
-architecture, design philosophy, security, testing. A dash means the scale
-does not apply. Arrows are against 28 September. Every row rests on a
-full read this time; nothing is provisional.
-
-| Crate | CQ | Maint | Scale | Perf | Arch | Phil | Sec | Test | Note |
-|---|:-:|:-:|:-:|:-:|:-:|:-:|:-:|:-:|---|
-| concat-core | 8 | 8 | 6 | 6 | 8 | 8 | – | 7 | Held. `hold` landed as one flag read in two places with a test; dead `Project`, linear `clip_on_track_at` and a caller-less `source_duration` remain. |
-| concat-project | 8 | 7 | 6 | 6 | 7 | 7 | – | 7 | Held. Shapes copied the title pattern cleanly through settle, tidy and `has_non_finite`; but a shape cannot be edited once placed, an unknown kind is rewritten as a square, and 0.2.5 deletes them. |
-| concat-media | 7 | 7 | 5 | 7 ↑ | 8 | 7 ↓ | 6 | 7 | Copy count known now (2 CPU copies, 3 on the chip); every `unsafe` carries SAFETY; playback audio ignores the stream start while the export mix honours it (finding 3). |
-| concat-render | 7 | 5 | 6 | 7 | 8 | 7 | 6 | 7 | Parity for 10 plan features, 344 probes, device loss recovered; the install trial runs on the monitor's compositor and kills it on a timeout. |
-| concat-effects | 8 | 8 | 6 | 8 | 8 | 7 ↓ | 6 | 8 | `matches_category` hard-codes nine ids against "effects are data"; `[[replaces]]` unfenced (finding 2); `Package::replacing` still drops keys. |
-| concat-export | 6 | 4 | 6 | 5 | 8 | 7 | 6 | 7 | Device loss handled in three steps with the frame redrawn; `lib.rs` 2 873 lines, 41-field `ExportClip`, serial loop; the hold plays keyed effects early. |
-| concat-host | 7 | 6 | 5 | 6 | 6 | 6 | 4 | 6 | Logs as files with a level of their own; the data-folder move; the update chain, the cache deletion and the HDR proxy name are as they were. |
-| concat (Rust) | 7 | 4 ↓ | 5 | 6 | 6 | 6 ↓ | – | 5 | `studio.rs` 10 029; ten of twelve worker paths epoch-gated, art and save not; Save Audio As is window-only; a full publish per frame. |
-| concat (Slint) | 7 | 6 | 6 | 7 | 7 | 6 ↓ | – | – | Seeded-not-bound held everywhere; 0 ⌘ glyphs now; still 2 `accessible-*` properties, 6 literal strings, 28 colours outside `theme/`, no keyboard path into any dialog. |
-| concat-api | 8 | 8 | 5 | – | 8 | 6 ↓ | 6 | 7 | One line changed in six days while the window gained five operations the API cannot name; home-folder root and open-then-save unchanged. |
-| concat-server | 7 | 8 | 5 | – | 8 | 8 | 6 | 7 | Zero churn: fd leak, pre-auth seats, no `catch_unwind` on the Hub, gRPC without the JSON caps. |
-| concat-cli | 7 | 8 | – | – | 7 | 6 ↓ | 6 | 3 | Installs no logger, so the week's export diagnostics never reach a `serve` user; `serve` still parks forever. |
-| concat-speech | 7 | 5 | 6 | 7 | 6 ↑ | 7 | 7 | 6 | One `accelerated()` switch and one fallback serve CoreML and DirectML alike; nothing logs which provider a session got; `tts.rs` doc says whisper is a child process. |
-| concat-vision | 8 | 7 | 5 | 6 | 9 | 8 | – | 7 | Zero churn since the 23rd. |
-| concat-text | 8 | 7 ↓ | 5 | 6 ↑ | 9 | 8 | – | 7 | Faces mmapped once behind `Arc` (28 Sept #20 closed); shapes appended to the one 1 974-line file; emoji paint as boxes on a Mac. |
-| concat-android | 8 | 7 | – | – | 7 | 7 | – | 1 | Portrait locked; 10 `unsafe` each argued; still no tests; ships Chatterbox against the speech crate's own words. |
-
----
-
-## 4. Scores per feature
-
-Status: **done**, **partial** (works with a known gap), **thin** (present,
-weak). New rows since the 28th are marked ★.
-
-| Feature | Where | Status | 28 Sept | Today | Why |
-|---|---|---|:-:|:-:|---|
-| Launcher, new project, recents | `ui/start.slint` | done | 8 | 8 | Defaults to the desktop the system names; chip rows have arrow keys and a focus ring. |
-| Import, bin, drag and drop | `ui/workspace/media-pane.slint` | done | 8 | 8 | Drop works on native Wayland; "N selected" and the empty-bin sentence are untranslated. |
-| Timeline editing | `ui/timeline/`, `panes/timeline.rs` | done | 8 | 8 | ★ Razor B / Ctrl+B snaps to the frame and refuses within `MIN_DURATION`; ★ scrollbar whose extent never shrinks under a drag. |
-| ★ Volume line | `lanes.slint:587-675`, `studio.rs:6205` | partial | – | 6 | One drag is one undo step; it does not flush a pending title commit first (finding 9); `Colors.white` not a token. |
-| Undo and redo | `concat-project/src/editor.rs` | done | 8 | 8 | Sharing survives split, ripple and edit (two `ptr_eq` tests now); every snapshot still clones the bin. |
-| Transitions | `concat-render/src/plan.rs`, `concat-export/src/lib.rs:746` | done | 8 | 7 ↓ | ★ A dissolve into a clip with no handle holds its first frame; but the hold runs the clip's keyed effects early, and keys on a clip with a handle play early by the transition's length (finding 7). "Keep alpha" is unobservable. |
-| Effects and filters | `concat-effects/packages/` | done | 8 | 8 | 98 GPU packages, 344 probes, shapes round and glows past white; `[[replaces]]` drops keys and is now a hijack vector (finding 2). |
-| Colour and HDR | `model.rs`, `gpu.rs`, `encode.rs` | done | 7 | 7 | HLG and PQ round trips pinned; hardware HDR encodes carry static metadata in the container only; pre-0.2.5 imports still never re-probed. |
-| Scopes | `concat-render/src/scopes.rs` | done | 7 | 7 | Verified this time: constants agree between shader and drawer, pinned by test; one failed readback freezes them for good (low). |
-| Colour grading | `packages/concat.color-wheels`, `grading.slint` | done | 7 | 7 | Unchanged: wheels keyable, curves not, pointer-only. |
-| Titles and text | `concat-text`, `concat-host/src/titles.rs` | done | 7 | 7 | ★ Wraps at the frame edge with no box, circular plate corners, ★ several titles styled together; emoji are boxes on macOS (finding 14); missing font silent. |
-| ★ Shapes / Stickers | `media-pane.slint:1415`, `clips.rs:122`, `concat-text/src/lib.rs:1173` | partial | – | 6 | Seven figures placed in one command, round-trip and tidy; cannot be edited after placement, deleted by 0.2.5 on save, unknown kinds become squares (finding 5). |
-| Keyframes, animations | `concat-project/src/model.rs` | done | 7 | 7 | `SetClipKey` no longer records a no-op; a drag on a keyed position keeps its key; preset animations of 0.2.4 still dropped silently. |
-| Crop, transform, blend, masks | `concat-render/src/plan.rs` | done | 7 | 7 | Darken and Lighten parity at 0.99; Screen and Add never held to the oracle. |
-| Cutout, smart brush | `concat-vision/src/segment.rs` | done | 7 | 7 | Unchanged. |
-| Enhance | `concat-host/src/enhance.rs` | thin | 6 | 6 | Copies still land in `cache/`. |
-| Auto-captions | `ui/dialogs/captions.slint` | done | 7 | 7 | Epoch-gated end to end; the transcript's words sit off by the stream start on MTS files (finding 3). |
-| Text-to-speech | `concat-speech/src/tts.rs` | done | 7 | 7 | ★ Chatterbox on DirectML with a CPU fallback; nothing says which provider ran; `DirectML.dll` is not staged beside the binary. |
-| Templates | `concat-host/src/templates.rs` | done | 7 | 7 | Poster copied under a name nothing reads (low). |
-| LUT import | `concat-effects/src/cube.rs` | done | 8 | 8 | Unchanged. |
-| Custom and community packages | `concat-effects/src/catalogue.rs` | done | 7 | 6 ↓ | Trial at 512 px / 3 s exists, at the knobs' defaults only, on the monitor's own compositor; `[[replaces]]` unfenced (findings 2, 6). |
-| Proxies | `concat-host/src/proxy.rs` | partial | 5 | 5 | ★ Full quality plays the files themselves; a pre-0.2.5 HDR import with no proxy gets an HDR copy under its SDR name. |
-| Export | `ui/dialogs/export.slint`, `concat-export` | done | 7 | 7 | ★ Survives a dead device on a fresh one, then the software adapter; ★ resolution is the timeline's own; ★ picture and scrub strip in the sheet (coalesced, not per pixel); rate ladder is a fixed 24/30/60 (finding 8); still one thread. |
-| ★ Hardware encode | `concat-media/src/encode.rs:325` | done | – | 6 | NVENC, Quick Sync and AMF lead for HEVC and AV1; fallback at open only; nothing on this machine or in CI exercises a successful chip open; "· hardware" claimed for Media Foundation's software transforms. |
-| Playback and audio | `concat-host/src/playback.rs`, `concat-media/src/samples.rs` | partial | 6 | 5 ↓ | The preview's sound runs ahead of the picture by the stream start on MPEG-TS files (finding 3); the mixer still scans every clip per sample. |
-| Hardware decode | `concat-media/src/hardware.rs` | done | 7 | 7 | On by default everywhere with a device, not only Windows and Android as the 28th said; the "5× slower" row measures a path the app never takes for 8-bit H.264. |
-| Remote API | `concat-api`, `concat-server` | partial | 6 | 6 | One line changed; the window's new operations are five reachable, four not; `export.run` cannot refuse a hardware encoder. |
-| Self-update | `concat-host/src/updates.rs` | done | 5 | 5 | ★ `.msi` and Store detection; same trust chain, downgrade still one click; any executable under Program Files is taken for an `.msi` install. |
-| ★ Logs | `concat-host/src/logs.rs`, `studio.rs:2136` | done | – | 7 | A file per run, ten kept, 16 MiB cap, Concat's own level separate from the crates underneath; bursts folded in the window; a burst followed by quitting is never counted. |
-| ★ Keyboard menus, platform keys | `app.slint:418`, `platform.rs:473` | partial | – | 6 | F10 / Alt+letter open and walk; arrow walking stalls past four dead rows; Exit on Windows, Quit on Linux. |
-| ★ Save Video Frame As, Save Audio As | `studio.rs:6105`, `5488` | done | – | 7 | Both on a worker, epoch-gated, dialog on the next loop turn; the extension is substituted after the overwrite prompt; Save Audio As is window-only. |
-| Phone shell | `ui/phone/`, `concat-android` | partial | 6 | 6 | ★ Sheets close on a swipe; portrait locked; the shell's own 44 px rule broken by its chips and transport buttons; no preview on the phone's export sheet. |
-| Relink missing media | `ui/dialogs/relink.slint` | done | 8 | 8 | |
-| Themes and accent colours | `ui/theme/` | done | 8 | 8 | ★ Lime is the default; light and dark token sets identical. |
-| Languages | `concat/locales/` | partial | 8 | 6 ↓ | 1 001 keys × 14 locales, 0 missing; but 128 to 197 values per Latin and CJK locale are the English string verbatim and the check cannot see it (finding 4); no RTL. |
-| Accessibility | the whole UI | thin | 2 | 2 | Still two `accessible-*` properties in 28 796 lines; 149 `TouchArea`s, 14 `FocusScope`s. |
-| ★ Windows data folders | `concat-host/src/dirs.rs:95` | done | – | 6 | Models and caches in `%LOCALAPPDATA%`, moved once by rename; a failed move is silent and orphans gigabytes in the roaming profile. |
-
----
-
-## 5. The 28 September findings, re-verified
-
-Of the 18, none is closed outright, four are partly closed, and fourteen
-are open. Two of the open ones are worse than described (7 and 8) and one
-is sharper (11). The engine crates the findings live in barely changed in
-six days; the work went into the window.
-
-| # | Finding (28 Sept) | Status | Evidence today |
-|:-:|---|:-:|---|
-| 1 | Self-update trusts a digest beside the package; no signature; quarantine stripped; Windows unsigned | **open** | Releases from `api.github.com` (`concat-host/src/updates.rs:45`), package and digest from the same release's `manifest.json` (`updates.rs:440-456`), `xattr -dr com.apple.quarantine` (`596-598`). `a387328` signs the Windows installer only when `WINDOWS_CERTIFICATE_PFX` is set (`build-app.yml:666-684`, "unsigned otherwise"); `0d5bda5` adds the `msiexec /i` path (`updates.rs:533-539`). |
-| 2 | Remote page rooted at the home folder; no `HOME` = unconfined; open-then-save | **open** | `concat-server/src/lib.rs:145-149`, `80-86`; `concat-api/src/lib.rs:244` returns `Ok(())` on empty roots; Remote page passes no roots (`concat/src/panes/settings.rs:573-577`); `project.save` unchecked (`lib.rs:454-456`). Byte-identical since the 28th. |
-| 3 | Every JSON connection leaks an fd; seats counted before auth | **open** | `concat-server/src/json.rs:102` `Seat::take` before any read; `106-122` closer pushed into `connections`, drained only in `shut` (`lib.rs:284-288`); `json.rs:100` spins on EMFILE. |
-| 4 | Clear cache deletes the media behind reversed and enhanced clips; never swept | **open** | `reverse.rs:88-93`, `enhance.rs:71-74` name copies under `project/cache/`; `projects::clear_cache` removes the folder (`projects.rs:330-338`) from `studio.rs:6703`; `proxy::sweep` touches `cache/proxy/*.mp4` only. |
-| 5 | 0.2.4 documents with reversed clips or preset animations open silently changed | **open** | `MIGRATIONS` still two steps (`concat-project/src/doc.rs:62-71`); nothing reads `"reverse"` or `animation*` but a menu arm (`studio.rs:9627`); CHANGELOG 0.2.5 "Changes to note" does not say it. |
-| 6 | Embedded API opens a second GPU device | **open** | `settings.rs:581-586` builds `Api::new(events)` and shares only the exporter and the register; `export.run` → `run_on(request, None)` → `best_compositor()` → `WgpuCompositor::new()` (`concat-export/src/lib.rs:1240-1243`, `1165`). The new `recovered()` path does cover API exports. |
-| 7 | `[[replaces]]` drops keys, is silent, skips user packages and templates | **open, worse** | `Package::replacing` returns `keys: BTreeMap::new()` (`concat-effects/src/catalogue.rs:591`); `session.rs:105` discards the bool; no test. The 28th's "a user package's `[[replaces]]` never fires" is wrong today: `Catalogue::builtin()` returns `CURRENT`, which `install_with` fills with user packages (`catalogue.rs:901-903, 949-950`), so it fires at every open, and nothing checks what it replaces (finding 2). |
-| 8 | HDR proxies not regenerated for pre-0.2.5 imports | **open, worse** | No probe at open (`studio.rs:6608-6620` keys by the document's `color_space`); `proxy::write` decides HDR from the probe (`proxy.rs:208`) while the name comes from the document (`proxy.rs:114`), so an HDR import from 0.2.4 with no proxy yet gets an HEVC/HLG copy under its SDR name. |
-| 9 | Hub: heavy work in line, no `catch_unwind`, slow subscriber dropped silently | **open** | `concat-server/src/hub.rs:83-87`; probes and composites on the API thread (`concat-api/src/lib.rs:326, 465, 485, 632, 690-704`); outbox `sync_channel(256)` and the silent drop (`json.rs:304, 319-325`, `hub.rs:58-66`). |
-| 10 | Downgrade is one click with no gate | **open** | `install_version` takes `releases.get(version_choice)` with no comparison against `Version::current()` (`settings.rs:461-473`); the list keeps every release ≥ 0.2.5 (`updates.rs:141-143`). |
-| 11 | No ceiling on frame size or rate in the model | **open, sharper** | `is_sane` is `> 0` (`model.rs:1917-1919`); `rate_num` is `i64` and the export's `FrameRate::new(Rational::new(rate_num, rate_den))` goes through `reduce`, which is `.expect("rational overflowed i64")` (`concat-core/src/time.rs:100`, `234`). A hand-edited `"rateNum": 9000000000000000000` loads and the first `duration × rate` panics the export thread. |
-| 12 | Every undo snapshot deep-copies the bin | **open** | `editor.rs:141` clones the project; `media: Vec<MediaItem>` (`model.rs:1941`); `Batch` clones again per level (`commands/mod.rs:1027`). Batch title styling now sends one `Batch` of N `UpdateClip`s per gesture, so the pattern is exercised more. |
-| 13 | Export one thread with a readback per frame; 400-piece curves; mixer scans every clip | partial | Loop unchanged (`concat-export/src/lib.rs:1301-1468`); `ExportClip` 41 fields; the window no longer owns the audio cut (`studio.rs:2339, 5538` call `concat_export::audio_pieces`), the `clamp(8, 400)` lives in one place now (`lib.rs:1077`); `playback.rs:1032` unchanged. |
-| 14 | Font by absolute path; missing font silent | **open** | `fonts.add_file(...)` result dropped (`titles.rs:339-342`); `pick` returns whatever answers (`concat-text/src/lib.rs:332-339`); `Rendered` carries no "face used". |
-| 15 | Small correctness in concat-project | partial, grown | (b) closed: `SetClipKey` compares before and after (`properties.rs:169-173`). Open: `has_non_finite` lists 8 of 11 `TextStyle` f64 fields (`commands/mod.rs:912-920`); `follow_first_hdr` re-flip (`editor.rs:238-244`); 5 of 20 structs carry `extra`, `ShapeStyle` joins the ones without (`model.rs:901`); `"concat": "0.1.0"` (`doc.rs:320`). |
-| 16 | Server hygiene: socket umask; `connections()` over-counts; `stage` names; `project.create` unbounded | **open** | `lib.rs:195-197` no `set_permissions`; `lib.rs:257-259`; `message.rs:591` vs `concat-export:979, 984, 1466`; `lib.rs:385-394`. |
-| 17 | Host hygiene: `rpm -q` on the UI thread; predictable temporaries; two TLS stacks | partial | `installed_as` memoised in a `OnceLock` (`updates.rs:253-255`), once per process: closed. Temporaries unchanged (`concat-export/src/lib.rs:957-958`); `Cargo.lock` still has ureq 2.12.1 on rustls and ureq 3.4.1 on native-tls, `webpki-roots` twice. |
-| 18 | Carried: 5 ⌘ glyphs; `ui/demo/`; dead `core::Project`; `Path::exists` in the wasm crate; speed range ×5; `mobile.yml` sentence | partial | ⌘ **closed**: 0 in strings, labels from `@keys` and `platform::keys` (`tray.slint:138`, `platform.rs:473-477`). `mobile.yml` **closed**: the sentence is true, `mobile.yml:36` has `pull_request:`. `ui/demo/` partial: `editor-chrome.slint` no longer imported by the live tree; `effects-panel.slint` still is, for one 3-field struct (`app.slint:21`, `editor.slint:4`, `config-pane.slint:9`), 4 files / 738 lines. Open: `core::Project` (`concat-core/src/timeline.rs:499-514`); `Path::exists` (`model.rs:2028-2031`); speed range 3 definitions + 3 literals (`studio.rs:4305, 8573`, `cutout.rs:178`). |
-| – | Two window tests flaky by locale | **open** | `format.rs:496-504` and `panes/speech.rs:668-679` read `t()` with no lock; the i18n tests select `de` under a lock private to their module (`i18n.rs:377-380, 516, 549`). CI runs the window binary with default test threads (`ci.yml:112` is build jobs), so it can flake there too. |
-
----
-
-## 6. New findings, by severity
-
-No new critical. Each carries a location and whether the reviewer read the
-whole path **(verified)** or reasoned from part of it.
-
-### High
-
-1. **`main` has been red at `cargo fmt --check` for six days, and nothing
- after it has run.** (verified) Every completed CI run since `a262553`
- (28 September 22:16 UTC) fails at `ci.yml:93`; the gates behind it,
- clippy under `-D warnings`, `locales.py --check`, `models.py --check`,
- the whole test suite, the quick perf budgets and the gRPC build
- (`ci.yml:98-126`), have not executed on `main` for 71 commits. The first
- unformatted sites came in the afternoon of the 28th
- (`concat-text/src/lib.rs:436, 717` from `35b80d1`,
- `concat-media/src/encode.rs:454, 1531, 1546` from `8d3bf19`,
- `concat-host/src/dirs.rs:95` from `dc12b64`) and sixteen more followed
- (`f97c745`, `0699b44`, `21855a7`). The gate is on push to `main`
- (`ci.yml:5-7`), it fired on every push, and it was ignored. A 0.2.6 tag
- today runs `release.yml:66-68`, fails at fmt and produces no bundles;
- meanwhile a test regression in any of the 60 commits is invisible. A
- related gap: the engine job installs `stable` (rustc 1.99.0 in the log
- of run `37165478921`) and never reads `src/rust-toolchain.toml`'s 1.93
- pin, so CI and a local checkout format and lint with different
- toolchains. The fix is `cargo fmt` and one commit; the lesson is that a
- gate nobody reads is not a gate.
-
-2. **A user package can hijack every link to a live built-in.** (verified)
- `Replaces::names` matches any id (`concat-effects/src/manifest.rs:91-93`);
- neither `Manifest::validate` (`manifest.rs:489-504`) nor
- `Catalogue::add` (`catalogue.rs:986-1003`) checks that the replaced id
- is absent from the catalogue; `Catalogue::upgrade` takes the first
- package whose table matches (`catalogue.rs:1050-1058`);
- `Catalogue::builtin()` returns the installed catalogue once
- `install_with` has run (`catalogue.rs:901-903, 949-950`); and
- `Session::open` applies the upgrade before any edit, outside undo, with
- no notice (`concat-host/src/session.rs:104-105`,
- `concat-project/src/editor.rs:89-107`). A shared package `evil.glow`
- declaring `[[replaces]] id = "concat.glow"` is installed; every project
- opened afterwards has its Glow links rewritten to `evil.glow` with their
- keys dropped, and the next save writes that out. The trial at install
- vets the shader, not the manifest's claims. Fix: refuse a `[[replaces]]`
- whose id or bare id the catalogue already answers to, at `add` and
- `install_with`, and tell the person when `upgrade_links` changed
- anything.
-
-3. **The preview's sound and the transcript ignore the stream start; the
- export does not.** (verified) `concat-media/src/samples.rs:176-182`
- seeks to `start × AV_TIME_BASE` with no `start_of`, and the graph trims
- on raw timestamps (`atrim=start={start}`, `samples.rs:91, 95`). For an
- MPEG-TS or MTS file whose first sample is at 1.4 s, `start = 0.5` lands
- at the file head and `atrim` keeps every frame, so the cached PCM begins
- at the sound's first sample while the picture begins 0.5 s in: the
- preview's sound runs ahead of the picture by `min(start, 1.4)` s
- (`concat-host/src/playback.rs:734-745`), and the transcript's words sit
- the same amount off (`concat-speech/src/transcribe.rs:339-350`). The
- export mix adds the start back on the seek and measures its lead off the
- first frame (`concat-media/src/audio.rs:422-434`), so the file is right
- and the preview is wrong: two definitions of source time, which
- `audio.rs:11-14` says do not exist. The 28th's "stream start time
- ignored" was closed for video only. `fixtures/starts-at-1.4s.ts` has no
- audio stream, so the test needs a new fixture.
-
-4. **Eight locales are a fifth English, and the check cannot see it.**
- (verified) `locales.py --check` compares key sets
- (`scripts/locales.py:134-148`), so a value equal to the English string
- passes. Today, ignoring symbol-only strings: de 195, fr 197, pt-BR 170,
- es 168, it 168, tr 150, ja 137, ko 128 values equal to `en.json`; hr 53,
- fa 5, zh-TW 5, zh-Hans 4, ru 3. In `de.json`, `captions.addCaptions` is
- "Add captions", `captions.oneTitlePerSentence` the full English
- sentence, and every `effects.*.description`. A German user opens the
- Captions sheet and reads English. The 33 keys added since the 28th are
- all translated; the debt is older (193 in `de` at `73e2176`). A
- `--check` that flags a value equal to English on a key of two or more
- words would make the "all 100 %" of the README mean what it says.
-
-### Medium
-
-5. **Shape clips do not survive 0.2.5, and an unknown shape kind is
- rewritten as a square.** (verified) `wire::clip_kind` reads `"shape"`
- as `Video` (`concat-project/src/model.rs:1159-1162`), `settle_clip` drops
- a Video clip with no media (`doc.rs:266-269`), and the flat `clips`
- mirror carries them too (`doc.rs:345-348`): 0.2.5 opens a 0.2.6 project
- with every shape gone and the next save removes them, at the same
- `DOCUMENT_VERSION` of 1 (`doc.rs:41`), with no "newer" refusal. In the
- other direction `wire::shape_kind` falls back to `ShapeKind::default()`
- (`model.rs:1184-1187`) and serialises it as `"square"`, so a later
- build's `"star"` is a square here and `"star"` is gone on save.
- `TextAlign` and `ColorSpace` have the same lossy fallback. Also: no
- command edits `Clip::shape` after placement (`ClipPatch` has `text`, no
- `shape`, `commands/mod.rs:143`), so fill, stroke and figure are fixed
- at `AddShapeClip` and an API caller has delete-and-re-add; and
- `AddShapeClip` is one of 17 of 44 commands absent from the serde
- round-trip test (`lib.rs:3440-3565`).
-
-6. **The install trial runs at the knobs' defaults only, and a timed-out
- trial kills the monitor.** (verified) `trial_pass` is
- `pass(&BTreeMap::new(), None)` (`concat-effects/src/catalogue.rs:489-491`)
- and `bounded` asks only that a `break` exist in the loop
- (`shader.rs:1108`): a loop whose count is `params.n × 1e6` with default
- 1 passes the trial and hangs the device when the slider is raised. And
- `trial_at` on timeout sets `self.dead = true` (`concat-render/src/gpu.rs:1869`)
- on the monitor's own compositor (`concat-host/src/preview.rs:120-126`);
- from then every frame answers "the GPU device was lost"
- (`preview.rs:234-237`), the pool stays allocated, and nothing in
- `concat/src` rebuilds the monitor. A hostile package is refused and
- still ends the session's preview. `Package::params_at(At::Max)` exists
- (`catalogue.rs:597`) and the vet does not use it; `monitor.sibling()`
- exists and the trial does not run on one.
-
-7. **The dissolve hold plays the clip's keyed effects early, and keys on a
- clip with a handle play early by the transition's length.** (verified
- for the hold; partial for the handle case) `pre_roll` clones the
- incoming clip (`concat-export/src/lib.rs:776`), clears its placement
- keys (`803`) but keeps `effects` with their knob keys, and `passes_at`
- resolves them at the hold's own fraction (`lib.rs:1478-1495`): a film
- grain keyed 0→100 over a 4 s clip, cross-faded in over 1 s with no
- handle, runs its ramp inside the one-second hold and snaps back to 0 at
- the cut. With a handle, `pre_roll` extends the clip backwards
- (`lib.rs:762-767`) and keys are fractions of the clip, so a key at
- fraction *a* lands *d·(1−a)* earlier than the clip's own span says.
- Monitor and export agree with each other, not with the keyframe markers.
- The holds are unit-tested (`lib.rs:2444-2512`) but not pixel-pinned;
- `concat-host/tests/export.rs` has not changed since the 28th.
-
-8. **The export's frame rate is a fixed 24/30/60 ladder, not the
- timeline's.** (verified) `EXPORT_RATES` (`concat/src/studio.rs:154`),
- `rate: 1` default (`panes/export.rs:113`), options hard-coded
- (`ui/dialogs/export.slint:363`). `e4d3507` made the resolution the
- timeline's own and greys the other rungs (`export.slint:350-356`); the
- rate was left. A 25 fps PAL or 23.976 cut exports at 30 fps unless the
- person notices, with a frame conversion the sheet never names.
-
-9. **A volume-line drag can drop typed title words.** (partial: the UI half
- read) `clip_set_text` leaves the words on the echo with
- `commit_pending = true` and `commit_targets = [title]`
- (`studio.rs:4446-4448`); the volume `TouchArea`'s press does not go
- through `clip_pressed`, so nothing flushes; `clip_volume` then
- overwrites `commit_targets` with the sound clip (`6213-6215`), and when
- the 120 ms timer fires `commit_now` diffs only that clip
- (`4634-4644`). Type into a caption's Content field, then press and drag
- a sound clip's volume line within 120 ms of the blur: the words go with
- the echo. Every other gesture entry flushes first (`3926`, `4883`,
- `4967`, `lib.rs:927`).
-
-10. **Art caches are never released across projects, and the art workers
- are not epoch-gated.** (verified) `open_project` and `close_project`
- clear `art_failed`, `art_stamp` and `proxied` only
- (`studio.rs:6626-6628`, `6739-6741`); `peaks`, `strips`, `windows`,
- `media.thumbs` and `waves` live for the process. `spawn_art` and
- `spawn_strip` go through `deliver(None, …)` (`host.rs:253-275`;
- `studio.rs:2822`, `2925`), so a strip decoded for project A lands in
- B's tables after the switch; `save` uses the ungated `spawn`
- (`studio.rs:2310-2320`), so a failed save of A sets `dirty` on B;
- `close_project` cancels cutouts, enhancers and reversers but never a
- running export (`6712-6762`). Ten of twelve worker paths are gated; the
- 28th's finding 3 is closed for those and open for these.
-
-11. **A full publish per monitor frame.** (verified) Every `deliver` and
- `on_ui` ends in `publish` (`host.rs:243`, `303`) and monitor frames
- arrive through `spawn_in_project` (`panes/monitor.rs:257`), so playback
- is one full publish per frame, about 30 a second, on top of the
- transport timer's `publish_lanes` every 33 ms (`studio.rs:2530-2566`).
- One full publish rebuilds every visible `ClipData` with wave and strip
- lookups (`7199-7248`), the key editor rows (`7325`), the font list
- collected and sorted (`4456-4480`), the bin rows (`8785`), the export
- ladder three times (`panes/export.rs:470-499`), the settings data and
- both menus. `refresh_art` itself is one comparison now
- (`8910-8920`); the publish behind it is not.
-
-12. **No dialog button, card or sheet is keyboard-reachable, and Escape
- closes nothing.** (verified) `Button` (`ui/primitives/modal.slint:168-251`)
- has a `TouchArea` and no `FocusScope`; `QualityCard`, `PillButton`,
- `SheetButton`, `ToolItem`, every `MediaCard`, `CatalogueCard` and
- `ShapeCard` likewise. `Key.Escape` is handled only in `select.slint:308`,
- `context-menu.slint:192`, `number-field.slint:143` and
- `tabs.slint:164`; `Modal` (`modal.slint:17-160`) and `ExportDialog`
- have no key handling. Open Export with Ctrl+E: Tab reaches the Select
- rows and nothing else, Escape does nothing, Return cannot press
- Export. Together with the two `accessible-*` properties this is the
- accessibility row's 2.
-
-13. **Six user-visible strings and the whole effects category bar bypass
- `I18n`.** (verified) `tray.slint:182` "Merge selected",
- `media-pane.slint:2197` "{n} selected", `media-pane.slint:2239` the
- empty bin's only sentence, `media-pane.slint:576` "still",
- `text-panel.slint:179` "Empty", `controls.slint:100` "% of height";
- and `CategoryPillBar`'s "Featured", "Retro & Film", "Optical & Lens"…
- (`media-pane.slint:1332-1340`), English on every locale and compared by
- string to Rust (`studio.rs:312`). `locales.py` extracts `I18n.t(...)`
- calls only, so none of these can reach a translator.
-
-14. **Emoji paint as boxes in titles on macOS.** (verified) `draws` counts
- a face only when the glyph has an outline
- (`concat-text/src/lib.rs:530-545`); Apple Color Emoji has none, so no
- face covers "🎉"; `assign` falls through to index 0 (`lib.rs:432-452`)
- and the style's face paints its `.notdef` box, in the monitor and in
- the export. Segoe UI Emoji has outlines, so Windows is fine; Linux
- depends on the fonts installed.
-
-15. **No Unreleased section, and the version is still 0.2.5.** (verified)
- `CHANGELOG.md` has one `##` section; nothing touched it in 60 commits;
- `src/Cargo.toml:6` says 0.2.5. Not recorded: hardware encode, Chatterbox
- on DirectML, **models and caches moved to `%LOCALAPPDATA%`** (a
- migration on first launch), keyboard menus and Exit/Quit, the razor,
- the scrollbar, the volume line, Save Video Frame As, the export sheet,
- the activity-trail log, multi-title styling, Stickers, the lime accent,
- Wayland drop, export surviving a lost device, the dissolve hold, the
- Store `.msix`, the AppImage rename. The release script's fallback is the
- commit list, which the project's own rule says is not wanted.
-
-16. **The code grew without tests.** (verified by `git log --numstat`)
- Since the 28th: `concat` +2 111 Rust and +1 291 Slint lines, 5 new
- tests (`platform.rs` 2, `wayland_drop.rs` 2, `studio.rs` 1);
- `concat-media` +126, 0; `concat-speech` +18, 0; twelve packages' WGSL
- changed, covered by probes only; the logs rewrite has one test
- (`level_names_round_trip`). 26 tests workspace-wide for 3 800 lines,
- none of them end to end. Nothing covers `commit_now`/`flush_commit`
- ordering, `clip_volume`, `save_frame_as`, `split_clip_at`, a real
- device loss, or a successful NVENC/QSV/AMF open.
-
-17. **The README sells an MCP transport that does not exist.** (verified)
- `README.md:62` "a JSON-RPC, gRPC and MCP API", `:79`, `:114`. No crate,
- feature or binary serves MCP; the only hit is a doc comment
- (`concat-api/src/lib.rs:7`); `version.capabilities` can name `json-rpc`,
- `unix-socket` and `grpc` only (`concat-server/src/lib.rs:137-141`).
- An agent builder reading the README looks for an endpoint that is not
- there.
-
-18. **The CLI installs no logger, so the week's export diagnostics are
- dropped.** (verified) `concat-cli/Cargo.toml:14-27` depends on neither
- `log` nor `concat-host::logs`; `main.rs` has no `set_logger`. `0699b44`
- and `d04a34c` put the device-loss reason, the adapter name and the
- recovery on `log::warn!`/`log::info!`
- (`concat-export/src/lib.rs:1206-1211, 1247-1252`); through `serve` or
- `api` they go to the no-op logger and the caller sees only
- `export.failed` with the last string. The Settings log level has no
- CLI or API counterpart either.
-
-19. **The API fell six days behind the window.** (verified) One line
- changed in `concat-api` (`proxy: false` on `FrameSpec`, `lib.rs:700`),
- none in `concat-server` or `concat-cli`, while `message.rs:25-26` still
- promises "every operation the window can perform is one an API caller
- can". Reachable: split (`SplitClips`), clip level (`UpdateClip.volume`),
- shapes (`AddShapeClip`, though `name` is a required string the caller
- must invent), Save Frame (`preview.frame`), the resolution ladder. Not
- reachable: Save Audio As, the log level, the activity trail, CBR or a
- bitrate (`lib.rs:568-569` hard-codes VBR), and the hardware encoder,
- which `export.run` cannot refuse because `EncodeOptions` gets
- `hardware: true` unconditionally (`concat-export/src/lib.rs:1277`) and
- neither `ExportRequest` nor `ExportSpec` carries the field; the window
- only labels it (`panes/export.rs:545-546`).
-
-### Low
-
-20. **concat-project.** `has_non_finite` runs after the snapshot clone
- (`editor.rs:141` vs `commands/mod.rs:1021-1023`), so a refused NaN
- still paid for a project clone; `AddShapeClip` repeats `AddTextClip`'s
- lane-minting block verbatim (`commands/clips.rs:131-150` = `81-100`);
- `core::Clip::source_duration` ignores `hold` and has no caller
- (`concat-core/src/timeline.rs:314-320`).
-
-21. **concat-render and concat-effects.** A scope readback that fails
- freezes the scopes for good (`gpu.rs:1777-1781`, `1623-1629`,
- `1790-1802`); `transitions.rs` is a 601-line `#[cfg(test)]` oracle
- nothing compares the GPU against (`lib.rs:28`, `reference.rs:41-43`);
- the built-in pipelines in `with_device` and the scopes compute pipeline
- are built outside any error scope (`gpu.rs:977, 1014, 1059, 1122,
- 1683-1692`), so ARCHITECTURE's "never an uncaptured error" holds for
- package pipelines only; `read_back` polls with no timeout
- (`gpu.rs:3441`); "transitions keep alpha" cannot be observed because
- both stages clear opaque black and `combine` reads back opaque frames
- (`gpu.rs:2932, 2891, 3483-3485`); the window now takes
- `adapter.limits()` (`concat/src/gpu.rs:94-98`) and concat-render never
- checks a plan against `max_texture_dimension_2d`; `matches_category`
- hard-codes nine package ids (`catalogue.rs:437-475`).
-
-22. **concat-media.** No fallback once a hardware encoder has opened: a
- `send_frame` error ends the export with no trailer
- (`encode.rs:1235-1237`, `concat-export/src/lib.rs:1457-1459`);
- "· hardware" is claimed for Media Foundation, whose `h264_mf` opens on
- any Windows machine through software transforms (`encode.rs:189-193,
- 371-373`); hardware HDR encodes carry the mastering display in the
- container atoms only, no SEI (`encode.rs:522-541, 667-696`); the
- colour round-trip tests fill frames with greys, which any matrix
- survives, and `sws_setColorspaceDetails`'s return is ignored
- (`encode.rs:1119-1128`, tests `1687`, `1830-1884`); a fresh reader
- seeks twice (`pool.rs:367`, `918-921`); AMF is handed `qp_b` on HEVC
- and AV1 (`encode.rs:656-662`), silently dropped; ARCHITECTURE.md:260
- describes a hardware-decode Settings switch that `settings.rs:206-211`
- says is "Not a choice".
-
-23. **concat-host and concat-export.** A failed data-folder move is
- silent and leaves gigabytes in the roaming profile for good
- (`dirs.rs:119-122`); any executable under Program Files without an
- uninstaller is taken for an `.msi` install and `msiexec /i` installs a
- second copy (`updates.rs:287-309`); a template's poster is copied as
- `cache/preview.jpg` while `poster_cache` reads `poster.jpg` and deletes
- `preview.jpg` (`templates.rs:349-353`, `media.rs:651-653, 705`); a
- burst's count line is written only when the next message arrives
- (`studio.rs:2136-2169`), so a burst followed by quitting is never
- counted; `dirs.rs:149-154` is a unit test that on Windows runs
- `move_data` against the real `%APPDATA%`; ARCHITECTURE.md:240 says
- proxies are "quarter-size H.264" while an HDR source's is ten-bit HEVC
- (`proxy.rs:223-247`); `updates.rs:777`'s fixture still names a
- `-linux-` AppImage after `c75b09c`.
-
-24. **The window.** `save_frame_as` and `export_clip_audio` substitute the
- extension after the OS's overwrite prompt (`studio.rs:6148-6155`,
- `5577-5584`) and on a phone write unconditionally; inspector ranges
- live in four literal copies (`studio.rs:546, 4265, 5199, 7723`);
- `freeze_at_playhead` decodes on the UI thread (`6342-6349`); the Remote
- page mints its token from two `RandomState` hashers
- (`settings.rs:856-864`) while `token::mint` one crate away uses
- `getrandom`; the token's doc comment now sits above `LOG_LEVELS`
- (`settings.rs:843-848`); `looks_dir`'s doc says it packs a template
- (`studio.rs:6858-6861`); `export_clip_audio` carries three superseded
- doc blocks (`5488-5509`).
-
-25. **The UI.** Keyboard menu walking looks at most four rows ahead
- (`context-menu.slint:87-93`), so a fresh project's Edit menu (undo,
- redo, rule, split, delete all dead) cannot be walked; 28 colours named
- outside `theme/` (the transition seam chrome at `lanes.slint:766-946`
- is palette-blind on the light theme; `Colors.white/black` at
- `lanes.slint:615, 664, 670` are new); the phone shell's own 44 px rule
- is broken by `LaneChip` at 34 px and the transport buttons at 40
- (`phone/editor.slint:175-177, 455-489`), and a sheet full of 28 px rows
- is mostly unscrollable surface (`pane.slint:124-133`); the inspector's
- `in-chain` scans are hand-unrolled eight deep, so a ninth link draws as
- "not applied" (`clip-inspector.slint:282-289, 708-716, 968-976`);
- `ui/demo/` is now reachable through one 3-field struct and could go in
- one commit; the phone's export sheet has no preview
- (`export.slint:823-849`); `IconButton.label` and `RoundButton.label`
- are documented as the accessible name and never bound to
- `accessible-label` (`panel.slint:193`, `phone/controls.slint:20`).
-
-26. **The server.** gRPC has none of the JSON transport's caps: no
- connection limit, no pre-auth timeout, nothing beyond tonic's 64-event
- channel (`grpc.rs:70, 129-175`); `Seat` and `AUTH_TIMEOUT` are
- JSON-only (`json.rs:48-70, 125`).
-
-27. **Speech, models and dependencies.** The phone build carries Chatterbox
- (`concat/Cargo.toml:121` enables it for every target) against
- `concat-speech/Cargo.toml:25-27`'s "a gigabyte of model is not for a
- phone"; `with_execution_providers` does not fail when DirectML cannot
- be created, so a machine without D3D12 runs on the CPU under a log line
- saying DirectML (`chatterbox.rs:303-313`), and the Windows stage copies
- no `DirectML.dll` (`build-app.yml:705-710`); the HF-mirror preference
- does nothing for the seven GitHub-hosted models (`models.rs:158-163`);
- the AVX baseline is a CI env only (`build-app.yml:85-89`), a
- contributor's `--profile app` is AVX2 again, and no runtime CPU check
- exists; `whisper-rs`'s `set_progress_callback_safe` leaks a box per
- transcription (`transcribe.rs:413`); `tts.rs:7` still says whisper
- runs as a child process. Two ONNX runtimes on every platform (pyke's
- through `ort-sys`, k2-fsa's inside `sherpa-onnx`) and two TLS stacks
- remain; `bzip2` is linked three times, `tiny-skia`, `fontdb`, `png`
- and `jni` twice each.
-
-28. **CI and docs.** `WINDOWS_CERTIFICATE_PFX` and `_PASSWORD` are not
- declared under `workflow_call.secrets` (`build-app.yml:37-53`) and
- reach the job only through `secrets: inherit` (`release.yml:73`), and
- signtool is given the password on the command line (`:684`);
- `README.md:56, 80` list four platforms while `:97` adds iOS and iPadOS;
- `src/README.md:129-130` says the quick perf set runs "on every push",
- which has not been so since the 28th; `SECURITY.md:39-40` covers model
- digests and says nothing about the self-update's trust root;
- `THIRD_PARTY_NOTICES.md` names none of the eight direct dependencies
- added since the 28th (all permissive; `unicode-script` carries the
- Unicode data licence).
-
----
-
-## 7. What is good, and should stay that way
-
-- **The layering, still.** FFmpeg is only in concat-media (`grep` finds no
- other crate naming `ffmpeg-the-third`); `FrameSource`/`FrameSink` the
- only seam; the CPU compositor a test oracle and nothing else.
-- **`recovered`: three deaths, two kinds of successor, the black frame
- redrawn, the reason in the log** (`concat-export/src/lib.rs:1189-1213`,
- `concat-render/src/gpu.rs:3439-3455`), and the proof that it cannot
- write a partial output: the silent file is renamed or muxed only after
- success and removed on error (`lib.rs:960-982`).
-- **`hold` as one boolean read in exactly the two places every renderer
- agrees on** (`concat-core/src/timeline.rs:295-297, 325-327`), with the
- export building the held copy rather than the model learning about
- transitions.
-- **`settle_clip` and `Clip::tidy` extended for shapes in the right two
- places and nowhere else** (`doc.rs:265, 275-277`; `model.rs:1511`): the
- "one place clamps live" rule held through a new clip kind.
-- **The scheduler's `Finished` guard**: counters as a `Drop`, so a panic
- cannot wedge the reserve, with the test that proves it
- (`concat-media/src/prefetch.rs:374-392, 408-427`).
-- **`hardware_wins` as a measured policy with its numbers in the doc
- comment**, and perf rows for "the app's choice" beside the raw chip
- (`hardware.rs:193-210`, `concat-perf/src/main.rs:655-668`).
-- **Every `unsafe` block in concat-media says why it is sound** (37 sites,
- 0 without a SAFETY line), and the new `wayland_drop.rs` is a careful
- guest on winit's display: own queue, `prepare_read`/poll/cancel loop,
- CLOEXEC pipe ends as `OwnedFd`, offers declined and destroyed, device
- released on stop (`wayland_drop.rs:133-194, 338-380`).
-- **One epoch gate in one place, with a test** (`concat/src/host.rs:197-246,
- 688-705`), and every long job's progress through `on_ui_in_project`.
-- **Dialogs from a menu row deferred to the next loop turn, with the reason
- written down** (`studio.rs:5556-5565, 6102-6103`): the Save Audio As
- abort was a native dialog opened inside a Slint timer callback, and
- both Save verbs now hop through `on_ui` first.
-- **Monitor request coalescing** (`panes/monitor.rs:234-237`): one frame in
- flight, latest wins, so the export sheet's scrub strip cannot decode per
- pixel; and no timer runs in an idle window.
-- **Probes read back in half floats with nothing between shader and
- number**, 344 of them across 98 packages on every `cargo test` with a GPU
- (`gpu.rs:1885-1901`, `gpu/tests.rs:1091-1114`); the pool that is also the
- cache, LRU within two bounds (`gpu.rs:2509-2581`).
-- **Logs as a file per run, ten kept, 16 MiB cap, Concat's own level
- separate from the crates underneath** (`concat-host/src/logs.rs:38-41,
- 228-268`).
-- **Faces mapped once and shared behind `Arc`**
- (`concat-text/src/lib.rs:475-491`); per-character coverage learnt once
- (`393-420`).
-- **The archive family's install**: `.part` → staging → `unpack_in` →
- single rename, so nothing half-unpacked ever looks installed
- (`concat-speech/src/tts.rs:529-580`); resumable downloads by `Range` with
- 416 handled and a real socket test (`concat-host/src/models.rs:201-284,
- 413-469`).
-- **`token::matches` with the test naming each promise**, used by both
- transports (`concat-server/src/token.rs:29-36, 52-64`); `Api::writable`
- + `resolved`; `OpenProjects` with `Drop`; `SingleFlight`.
-- **`locales.py` builds `en.json` from the source and the package
- manifests** (`scripts/locales.py:46-96`), so a key cannot be used
- without English and a stale key cannot survive; and the seeded-not-bound
- discipline for every text field, with the reason at each site.
-- **The release gate order**: tag must equal the version, CI must pass
- before a bundle is built, notes from CHANGELOG with a fallback
- (`release.yml:46-73`); parity real on the Linux runner through lavapipe
- and `CONCAT_REQUIRE_GPU` (`ci.yml:44-46, 116`).
-
----
-
-## 8. Where we stand
-
-The 28 September audit ended with seven recommended steps. Six days later:
-
-| Step (28 Sept) | Landed |
-|---|---|
-| 1. Sign the update; keep quarantine; confirm downgrade; sign Windows; notarise every Mac build | Windows signing exists when a secret is set; nothing else |
-| 2. Remote page defaults, seats after auth, drop the connection clone, `catch_unwind` the Hub, lend the device | nothing (one line changed in the API, none in the server) |
-| 3. Processed media out of `cache/`; sweep art and peaks; re-probe pre-0.2.5 media | nothing |
-| 4. Say it when a document changes: migrate `reverse` and animations; surface `upgrade_links` | nothing, and a new silent seam (shapes under 0.2.5) |
-| 5. Finish the 23 Sept list | ⌘ glyphs gone; `mobile.yml` sentence was true; `rpm -q` memoised; `audio_pieces` moved to one place; the rest open |
-| 6. Pipeline the export, index the mixer, `Arc` the bin | nothing |
-| 7. Speech, enhance, proxies and CBR as API verbs | nothing; the gap grew by four window-only operations |
-
-None of the seven was taken up. What landed instead is a week of the
-window: a razor, a scrollbar, a volume line, keyboard menus on every
-platform, Save Frame and Save Audio, an export sheet with a picture, logs
-that read as what the person did, shapes, batch title styling, Wayland
-drop, Linux window edges, Windows data folders and Store packaging,
-hardware encoders, Chatterbox on DirectML, and an export that survives a
-dead GPU. Each of those is well made at the seam it touches: the new
-gestures go echo-then-command, the new workers are epoch-gated, the new
-`unsafe` is argued, the device recovery cannot write a partial file. The
-app is more pleasant to use than it was.
-
-What the week cost is of three kinds. The first is the gate: `main` has
-not been green since the release, fmt has been the only check to run for
-71 commits, and the version and CHANGELOG do not know the week happened.
-The second is the engine debt of the 28th, untouched and in three places
-worse: `[[replaces]]` went from "drops keys" to "a package can hijack a
-built-in", the HDR proxy from "not regenerated" to "written under the wrong
-name", and the frame-rate ceiling from "no ceiling" to "a panic in the
-export". The third is a new correctness seam in the two paths the audit of
-the 28th had not read in full: the preview's sound is off by the stream
-start on MPEG-TS files, and shapes vanish under 0.2.5.
-
-**Recommended order.** Each line is an hour to a few days, and each closes
-a finding above.
-
-1. `cargo fmt`, push, watch it go green; put `rust-toolchain.toml` into the
- CI job; isolate the two locale tests; then write the Unreleased section
- and bump to 0.2.6. (1, 15, flaky tests)
-2. Refuse a `[[replaces]]` naming a live package at install; carry keys
- through `replacing`; tell the person when `upgrade_links` changed a
- document; run the trial at `params_at(Max)` on a `sibling()` compositor.
- (2, 6, 28 Sept #7)
-3. `AudioDecoder::open` takes `start_of`, seeks and trims at
- `start + offset`, with an MTS fixture that has sound. (3)
-4. Bump `DOCUMENT_VERSION` for shapes so 0.2.5 refuses rather than deletes;
- keep an unknown kind's name in `extra`; give `ClipPatch` a `shape`;
- grow the round-trip test to all 44 commands. (5)
-5. A `locales.py --check` that flags values equal to English; the six
- literals and the category bar through `I18n`; the export rate from the
- timeline; freeze effect keys at the hold's head. (4, 8, 13, 7)
-6. Flush the pending commit in `clip_volume`; gate `spawn_art`,
- `spawn_strip` and `save` by epoch and clear the art tables at close; a
- `publish_lanes`-only delivery for monitor frames. (9, 10, 11)
-7. A `FocusScope` on `Button` and Escape on `Modal`; bind
- `accessible-label` where the doc already promises it. (12)
-8. Then the 28 September list, in its order: sign the update; the Remote
- page's root, seats and fd leak; processed media out of `cache/`; the
- migration notices; the export pipeline; the API verbs. (28 Sept 1-7,
- today 17-19)
-
-After those, phase 4 of the HDR plan (a true HDR preview), then phase 5
-(zero-copy hardware frames, hardware encoders on the Mac) in the plan's
-order.
-
----
-
-## 9. Caveats
-
-Scores are judgements over code, calibrated against the 28 September
-scores but not measured against users or another editor. Every crate was
-read in full this time; nothing is provisional. Findings marked (verified)
-were read end to end by the reviewer and the three headline ones re-read
-by hand; the rest rest on one reviewer's reading and a line number you can
-open. The hardware encoders and DirectML could not be exercised on this
-machine; what is said of them is read from the code and the workflows. The
-app was not launched: "user likeability" is judged from what the UI code
-does, not from users.
diff --git a/audit/AUDIT-2026-10-09-QUICK.md b/audit/AUDIT-2026-10-09-QUICK.md
deleted file mode 100644
index 326b54a9..00000000
--- a/audit/AUDIT-2026-10-09-QUICK.md
+++ /dev/null
@@ -1,133 +0,0 @@
-# Concat audit, 9 Oct 2026 — the quick version
-
-Same audit as `AUDIT-2026-10-09.md`, one screen. Each bar has 10 slots:
-🟢 filled = good, ⚪ empty = missing. Severity: 🔴 critical · 🟠 high · 🟡 medium · ⚪ low.
-Arrows are against the 4 Oct audit. Every crate was read in full by one of four reviewers.
-
-## 🎯 Verdict
-
-**A good week for the code, a bad week for the pipeline.** Five days and 54 commits after the last audit, two of its three headline findings are closed (`[[replaces]]` fenced with tests, the preview's sound at the stream start), fmt is green, the window suite no longer races, and the day's six fixes each land with a test that fails without them: phones export and can find the file, stills keep their shape, Detach means detached, Chinese and Korean have a face to fall back on. But the release pipeline has three holes it did not have on the 4th's radar: **0.2.6 was built from `main`, 35 commits past its tag**; `main` has been red for five days on a Windows test that hangs; and the Android release key is readable by every pull-request build. In the engine, an export with sound can leave a broken file at the final path. 🟠 ×5 new, no 🔴.
-
-## 📏 Size
-
-| | Lines | Δ vs 4 Oct |
-|---|---:|---:|
-| Rust, all crates | 88 879 | +1 503 |
-| Slint UI | 29 717 | +921 |
-| WGSL, packages | 1 764 | 0 |
-| Tests (`#[test]`) | 713 | +37 |
-| Largest file `concat/src/studio.rs` | 10 435 | +406 |
-
-Checks today: fmt clean · clippy 0 warnings · 711 tests pass, 1 ignored · locales 0 missing · 22/22 perf scenarios in budget (on a loaded machine) · CI on `main` red.
-
-## 📊 The whole app
-
-| Aspect | Bar | 4 Oct | Today |
-|---|---|:-:|:-:|
-| 🧹 Code quality | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 🏛️ Architecture | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 |
-| 🧭 Design philosophy | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 🔧 Maintainability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 |
-| 😊 User likeability | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 |
-| ⚡ Performance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 📈 Scalability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 |
-| 🔐 Security | 🟢🟢🟢🟢⚪⚪⚪⚪⚪⚪ | 4 | 4 |
-| 🧪 Testing | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 📚 Docs | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 🤖 CI and release | 🟢🟢🟢⚪⚪⚪⚪⚪⚪⚪ | 3 | 3 |
-
-Flat across the board: what landed was good and what it cost was elsewhere. Security stays at 4 because the hijack closed and the model downloads are pinned, while the release key, the untagged release and the unverified FFmpeg opened.
-
-## 🧱 Per crate (average of its scales)
-
-| Crate | Lines | Δ | Bar | Avg | One thing |
-|---|---:|---:|---|:-:|---|
-| concat-vision | 2 760 | 0 | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7.5 ↑ | 🟢 untouched, pinned, digested · ⚪ runtime wrapper untested |
-| concat-core | 2 891 | +37 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.3 | 🟢 `FrameRate::checked` in lowest terms · ⚪ `reduce` panics on file rates |
-| concat-effects | 7 256 | +160 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.3 | 🟢 every take-over refused with a reason · 🟡 trials on the UI thread, a catalogue leaked per reload |
-| concat-project | 10 216 | +751 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.3 ↑ | 🟢 placement is a model rule, 10 tests · 🟡 a 1e300 start panics the export |
-| concat-server | 1 393 | +17 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.6 ↓ | 🟢 Hub panics are one caller's error · 🟠 fd leak, pre-auth seats, a reader that stops pins its seat |
-| concat-speech | 3 109 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.5 | 🟢 downloads pinned to commits · ⚪ no change, stale docs |
-| concat-media | 9 832 | +338 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.6 ↓ | 🟢 stream start fixed with a fixture; 38 `unsafe`, 38 SAFETY · 🟠 the mux writes the final path |
-| concat-text | 2 607 | +633 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.6 ↓ | 🟢 colour emoji probed, not trusted · 🟡 12 MB copied to the heap, PNG decoded per render |
-| concat-api | 2 351 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.4 ↓ | 🟢 confined writes · 🟡 reads unconfined, nine window operations it cannot name |
-| concat-render | 7 432 | −692 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.5 | 🟢 scopes gone cleanly · 🟡 no uncaptured-error handler, a 10 000-px frame panics |
-| concat-android | 398 | +63 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 | 🟢 publisher reports `[path, why]` · 🟡 a provider exception kills the process · ⚪ 0 tests |
-| concat-cli | 514 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 ↓ | ⚪ no logger, no signal handling, token on argv |
-| concat-host | 10 630 | −1 239 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.8 | 🟢 `catch_unwind`, memory-sized caches, voiceover · 🟡 `Server::stop` blocks the UI; crashed takes lost |
-| concat-export | 3 844 | +142 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.9 ↓ | 🟢 folder and still fixed with tests · 🟠 mux not atomic; silent frame loss on one path |
-| concat (Slint) | 29 717 | +921 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.5 | 🟢 Escape on the confirm sheet · 🟡 2 `accessible-*` in 29 717 lines |
-| concat (Rust) | 22 554 | +1 293 | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5.1 ↓ | 🟢 three `expect`s in 22.5k lines · 🟡 dialogs under a `RefCell` borrow; `studio.rs` +406; 71 tests |
-
-## 🎛️ Per feature
-
-| Feature | Bar | 4 Oct | Today | State |
-|---|---|:-:|:-:|---|
-| Launcher | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Import and bin | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ ★ click to preview |
-| Timeline editing | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ ★ never lands on top, ★ detach alone |
-| Titles and text | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7 | 8 ↑ | ✅ ★ colour emoji, ★ font picker |
-| Auto-captions | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7 | 8 ↑ | ✅ words at the right time on MTS |
-| Effects and filters | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ `[[replaces]]` fenced |
-| Undo / redo | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| LUTs | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Relink | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 8 | 6 ↓ | ⚠️ "Relink all" walks on the UI thread, follows symlink loops |
-| Themes | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Playback and audio | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 5 | 7 ↑ | ✅ stream start, one-word clock, window decoded ahead |
-| Transitions | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ hold plays keyed effects early |
-| Colour and HDR | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ AMF Main 10 |
-| Colour grading | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Keyframes | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Crop, blend, masks | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ shape masks still absent (#239) |
-| Cutout | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Text-to-speech | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Templates | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Custom packages | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6 | 7 ↑ | ✅ trials at Default, Min, Max on a sibling |
-| Export | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ ★ makes its folder; mux not atomic; overwrites without asking |
-| ★ Monitor stills | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | – | 7 | ✅ keeps its shape, pixel test |
-| ★ Export on a phone | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ into Movies/Concat; unseen on a phone |
-| Phone shell | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6 | 7 ↑ | ✅ exports, findable, stills right |
-| Languages | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6 | 7 ↑ | ✅ ★ CJK face held to the locales; ja real; de/fr/ko/pt-BR half English |
-| Volume line | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6 | 7 ↑ | ✅ flushes the title commit |
-| ★ Voiceover | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ a crashed take is lost |
-| Hardware encode | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ✅ |
-| Enhance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ copies in `cache/` |
-| Shapes / Stickers | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ vanish under 0.2.5 |
-| Keyboard menus | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ |
-| Windows data folders | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ✅ |
-| Remote API | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 6 | 5 ↓ | ⚠️ nine window operations it cannot name; `stop` blocks the UI |
-| Proxies | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ |
-| Self-update | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ no signature |
-| ★ VU meters | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | – | 5 | ⚠️ programme meter has no feed |
-| ★ Windows text | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | – | 5 | ⚠️ pixel-snapped at 1x; unseen on Windows |
-| Accessibility | 🟢🟢⚪⚪⚪⚪⚪⚪⚪⚪ | 2 | 2 | ❌ |
-| Scopes | – | 7 | – | removed |
-
-## 🚨 Findings
-
-🟠 **High**
-1. 0.2.6 was built from `main`, 35 commits past its tag: the build workflows check out with no `ref` (`build-app.yml:152`, `mobile.yml:58`).
-2. `main` red five days: the Windows test job hangs in `cards::tests::every_card_draws…` and dies at 90 min; the release gate reuses it.
-3. The Android release key and password reach pull-request builds and every build script (`mobile.yml:36, 113-130`).
-4. An export with sound muxes straight into the final path; a failure leaves a truncated file where the last export was (`audio.rs:720`, `concat-export/src/lib.rs:992`).
-5. The CJK face shipped with half of Big5 and no JIS; its test proved registration, not coverage. Fixed at the head with a coverage test.
-
-🟡 **Medium** — dialogs open under the studio's `RefCell` borrow (6) · the export sheet overwrites without asking and accepts `/` in a name (7) · a frame over 8192 px panics wgpu (8) · a 1e300 clip start panics the export (9) · a SAFETY line on `set_var` is untrue (10) · a group dragged past 0 collapses (11) · a speed curve grows over a neighbour (12) · trials on the UI thread, a catalogue leaked per reload (13) · "Relink all" on the UI thread (14) · fifteen English strings in Rust and Slint (15) · `Server::stop` blocks the window; jobs and decode workers die silently on a panic (16) · API reads unconfined, caller strings to FFmpeg's opener (17) · one export path drops failed frames silently (18) · pinned prefetch frames outside any budget (19) · a crashed voiceover take is lost (20) · unverified FFmpeg and sherpa downloads, mutable action tags (21) · three Android bridge crash paths (22).
-
-⚪ **Low** — doc drift in gpu.rs, studio.rs, pool.rs, ARCHITECTURE (23) · tests that skip in CI or prove the wrong thing (24) · Rational edges, negative starts, RIFF overflow, `DefaultHasher` proxy names (25) · non-ABI side-data sizes (26) · realtime audio frees and faults (27) · blocking IO on the UI thread (28) · hygiene (29).
-
-## 🧭 What the 4 Oct audit asked, and what landed
-
-✅ fmt, locale lock, 0.2.6 section · ✅ `[[replaces]]` fence, keys, notice, trials on a sibling · ✅ stream start with a fixture · ✅ export rate, volume flush, art caches · ❌ document version for shapes · ❌ `FocusScope`/Escape/`accessible-label` · ❌ the 28 Sept list (update signing, server seats and fds, API reach).
-
-## ▶️ Next, in order
-
-1. `ref:` the tag in both build workflows; Android signing on release runs only; skip or time out the cards test without a GPU; toolchain into CI; tag 0.2.7 from a green `main`.
-2. Mux into scratch and rename; one failure policy for the export; a size ceiling with an uncaptured-error handler; bound clip times and probed rates.
-3. Export sheet: refuse `/` and empty names, ask before overwriting; dialogs on the next loop turn.
-4. Trials, relink, cache clear and sweep off the UI thread.
-5. `Server::stop` cancels first; `catch_unwind` on jobs and workers; a protocol whitelist before FFmpeg; a write timeout.
-6. Pinned-frame budget; WAV placeholder header; group clamp; speed-curve ripple.
-7. The literals through `I18n`; a locale check that flags English.
-8. Digests for downloads; actions by SHA; `inputs.tag` through `env:`.
-9. See Windows text and the phone's export on real hardware. Then the 28 Sept list.
diff --git a/audit/AUDIT-2026-10-09.md b/audit/AUDIT-2026-10-09.md
deleted file mode 100644
index 9c8d9ed9..00000000
--- a/audit/AUDIT-2026-10-09.md
+++ /dev/null
@@ -1,766 +0,0 @@
-# Concat audit, 9 October 2026
-
-Where the codebase stands five days after the 4 October audit, at
-`99c3e26` (9 October), 62 commits past the 0.2.6 tag and 54 commits past
-the last audit. Same method, same scales, and every finding from the 4th
-(and what was carried from the 28th) re-verified against today's tree.
-Every finding names a file and a line. The sibling
-`AUDIT-2026-10-09-QUICK.md` is this audit as a one-screen scorecard.
-
-**How it was done.** Four reviewers each read one part of the tree in
-full: the window's Rust, Slint and locales; host, server, API and CLI;
-media, render, export and core; project, effects, text, vision, speech,
-Android, perf, CI and release. Each re-verified the 4 October findings in
-its part, scored it and reported new findings with locations. The four
-headline items (a release built from `main` instead of its tag, the
-Windows test hang that keeps `main` red, the Android key in pull-request
-builds, the non-atomic audio mux) were re-read by hand against the code
-and the workflow runs before they went in. Line, test and `unsafe`
-tallies come from the tree; `cargo fmt --check`,
-`cargo clippy --workspace --all-targets`, `cargo test --workspace`,
-`scripts/locales.py --check`, `scripts/models.py --check` and
-`cargo run -p concat-perf --release -- --check` were run on this machine
-(Apple Silicon, macOS 26) today. The app itself was not launched; the
-Android build was compile-checked for its target and the 0.2.6 APK was
-run on an emulator to the point where its window exits.
-
-**What also happened today.** Six fixes landed on `main` during the
-audit, after the reviewers had started: an export makes its folder
-(38e45a9), a phone's export goes into Movies (63f2633), a still keeps its
-shape on the monitor (89f5ddd), a CJK fallback face and five installer
-languages (eb18dbd, corrected by the commit at the head), a detached
-sound moves on its own (e11006e), and glyphs sit on whole pixels at 1x
-(3989e59, a vendored patch of Slint's Skia renderer). The reviewers judged
-them; two findings below are about them.
-
-**Scale.** 0 to 10. 8 and up means "keep doing this"; 5 to 7 means
-"works, with known debt"; 4 and under means "will bite at the next order
-of magnitude, or already bites".
-
----
-
-## 1. The numbers
-
-### 1.1 Size, tests and hygiene per crate
-
-Deltas are against the 4 October table. The unwrap column gives the
-total and, in brackets, the count outside test modules and test files
-(everything before the first `#[cfg(test)]` in each file; `tests/`
-directories excluded); the panic column is the non-test count of
-`panic!`, `unreachable!`, `unimplemented!` and `todo!`.
-
-| Crate | Rust lines | Δ | Slint / WGSL | Tests (files) | `unsafe` | `unwrap`/`expect` (non-test) | `panic!` family (non-test) |
-|---|---:|---:|---:|---:|---:|---:|---:|
-| concat (window) | 22 554 | +1 293 | 29 717 slint | 71 (18) | 12 | 28 (3) | 0 |
-| concat-host | 10 630 | −1 239 | | 85 (19) | 4 | 243 (2) | 0 |
-| concat-project | 10 216 | +751 | | 113 (1) | 0 | 355 (11) | 6 |
-| concat-media | 9 832 | +338 | | 96 (11) | 38 | 204 (23) | 0 |
-| concat-render | 7 432 | −692 | shaders inline in `gpu.rs` | 86 (5) | 1 | 86 (68) | 3 |
-| concat-effects | 7 256 | +160 | 1 764 wgsl | 72 (9) | 0 | 134 (5) | 2 |
-| concat-export | 3 844 | +142 | | 31 (3) | 0 | 28 (0) | 0 |
-| concat-speech | 3 109 | 0 | | 19 (3) | 3 | 23 (2) | 0 |
-| concat-core | 2 891 | +37 | | 49 (7) | 0 | 26 (4) | 2 |
-| concat-vision | 2 760 | 0 | | 31 (8) | 0 | 22 (2) | 0 |
-| concat-text | 2 607 | +633 | | 24 (1) | 1 | 67 (1) | 0 |
-| concat-api | 2 351 | 0 | | 21 (2) | 0 | 23 (0) | 0 |
-| concat-server | 1 393 | +17 | | 10 (3) | 0 | 28 (0) | 0 |
-| concat-perf | 1 092 | 0 | | 2 (1) | 0 | 34 (34) | 0 |
-| concat-cli | 514 | 0 | | 3 (1) | 0 | 2 (2) | 0 |
-| concat-android | 398 | +63 | +92 java | 0 (0) | 10 | 1 (1) | 0 |
-| **Total** | **88 879** | **+1 503** | **~31 500** | **713 (92)** | **69** | **1 304 (158)** | **13** |
-
-The tree grew by 1.7 % in five days, less than the 4.5 % of the week
-before, and for the first time a crate shrank: the Scopes pane and its
-GPU counting went (616d8fa, −1 239 lines in concat-host and −692 in
-concat-render). The growth is the window again (+1 293 Rust, +921 Slint:
-voiceover, VU meters, the font picker, click-to-preview, the publish
-hook, the CJK fallback), concat-project (+751: the placement rule and its
-tests), concat-text (+633: colour emoji, base fonts) and concat-android
-(+63 Rust, +92 Java: the media-store publisher). `studio.rs` is 10 435
-lines (10 029 on the 4th; 46 % of the window's Rust) and was not split;
-`concat-export/src/lib.rs` is 3 015 (2 873), `concat-render/src/gpu.rs`
-3 374 (3 630), `concat-effects/src/shader.rs` 2 412, `concat-text/src/lib.rs`
-2 607 (1 974). The four new `unsafe` sites are concat-host's memory
-sizing (`memory.rs:22, 37`, each argued) and concat-media's count moving
-from 37 to 38 with the audit's own recount; every one of the 69 carries a
-SAFETY line, and one of them is not true as written (finding 10 below).
-
-Other facts: 123 effect packages, unchanged (98 on format 2 with WGSL: 21
-effects, 57 looks, 20 transitions; 25 audio packages on FFmpeg chains);
-14 locale files at 1 011 keys each, 0 missing and 0 stale per
-`scripts/locales.py --check` (ja was translated for real this week,
-3aa6e5c; de, fr, ko and pt-BR still carry 137 to 204 English values, see
-finding 15); 873 commits since 25 August, 54 since the last audit (222
-files, +17 705 / −2 197; 37 by the maintainer, 14 by one contributor
-merged on the 6th, 3 by others); 30 tags, latest `v0.2.6`, 62 commits
-behind `main`; `Cargo.lock` holds 817 packages, one of them now a path
-dependency (`src/patches/i-slint-renderer-skia`, 1.18.1 with one
-function changed). The window carries 2.7 MB of Noto Sans CJK as a
-fallback face from today; the binaries grow by that much.
-
-### 1.2 The checks, run today
-
-| Check | Result |
-|---|---|
-| `cargo fmt --all -- --check` | **clean** (10 files would change on the 4th) |
-| `cargo clippy --workspace --all-targets` | 0 warnings |
-| `cargo test --workspace --no-fail-fast` | **711 passed, 0 failed, 1 ignored** (672 passed, 1 failed on the 4th) |
-| `scripts/locales.py --check` | 14 locales, 1 011 keys, 0 missing, 0 stale |
-| `scripts/models.py --check` | 23 models, 3.37 GB, mirrored on `models-v1` |
-| `cargo run -p concat-perf --release -- --check` | 22 of 22 scenarios within budget |
-| `cargo audit` | not installed on this machine; not run |
-| CI on `main` | **red at head, and on every completed push since 4 October 15:16** |
-
-**fmt is green and the window suite no longer races.** The ten fmt sites
-of the 4th were fixed the same day (3121107, 7b5cc44), and the two
-locale-dependent tests now take `i18n::language_lock()` (4c5055e,
-`concat/src/i18n.rs:363-366`), so the window suite passes with the
-default thread count. Every crate passes: the end-to-end export suite (11
-tests, 18 s) and the parity suite among them.
-
-**CI is red for a new reason.** The Linux engine job passes. The Windows
-test job added on the 6th (9440a67, `ci.yml:247-252`) hangs in
-`concat-host cards::tests::every_card_draws_and_shows_its_package_at_work`
-and is killed at 90 minutes, on every run that got that far (37521843321,
-37624855588, 37890980349). 14 of the 22 push runs since the audit were
-cancelled by the next push (`ci.yml:22-24`). `release.yml:67-72` reuses
-`ci.yml` as the gate, so a tag today would fail before it built anything.
-CI still installs `stable` (1.99) while `src/rust-toolchain.toml` pins
-1.93 (`ci.yml:38, 150, 191`).
-
-**The perf table**, release build, every scenario inside its budget. The
-4 October reading is in the fourth column.
-
-| Scenario | Measured | Budget | 4 Oct |
-|---|---:|---:|---:|
-| plan one frame of 200 clips on 4 tracks | 0.56 µs | ≤ 200 µs | 0.54 µs |
-| 200 clip additions and 200 undos | 0.97 ms | ≤ 150 ms | 0.71 ms |
-| ripple delete of 200 clips out of 400 | 0.12 ms | ≤ 20 ms | 0.13 ms |
-| save and open a 200-clip document | 1.80 ms | ≤ 60 ms | 1.67 ms |
-| choose the visible clips of 200 for one publish | 0.25 µs | ≤ 50 µs | 0.21 µs |
-| read an hour's waveform across 2048 columns | 4.92 µs | ≤ 2000 µs | 4.23 µs |
-| decode 1080p h264 in software | 568 fps | ≥ 120 | 809 fps |
-| decode 1080p h264 on the hardware (VideoToolbox) | 203 fps | ≥ 120 | 175 fps |
-| decode 4K h264, the app's choice (software) | 98 fps | ≥ 30 | 115 fps |
-| decode 4K HEVC 10-bit, the app's choice (VideoToolbox) | 71 fps | ≥ 30 | 92 fps |
-| decode 4K HEVC 10-bit in software | 26 fps | ≥ 24 | 31 fps |
-| decode 3 streams of 4K HEVC 10-bit at once | 32 fps | ≥ 24 | 45 fps |
-| decode 4K HLG tone-mapped on the CPU, 960×540 | 17 fps | ≥ 15 | 22 fps |
-| decode 4K HLG deep for the GPU, 960×540 | 103 fps | ≥ 30 | 127 fps |
-| scrub forward over 90 frames, cold | 1.36 ms/frame | ≤ 25 | 0.95 |
-| scrub back over covered ground | 0.51 µs/frame | ≤ 200 | 0.43 |
-| decodes when an effect knob is turned | 0 | ≤ 0 | 0 |
-| compose 1080p, 3 layers, a look, on the GPU | 3.33 ms | ≤ 25 | 1.92 ms |
-| Gaussian blur over 4K, radius 10, on the GPU | 6.42 ms | ≤ 33 | 6.34 ms |
-| Gaussian blur over 4K, radius 50, on the GPU | 11.12 ms | ≤ 33 | 9.77 ms |
-| export a 4 s 720p cut of two overlapping clips | 86 fps | ≥ 30 | 157 fps |
-| export the same cut as HDR: HLG, HEVC 10-bit | 64 fps | ≥ 30 | 82 fps |
-
-Read this table as a floor, not a trend. It ran while four reviewers
-read the tree and two cargo builds were in flight on the same machine,
-and the two rows that dropped most (software h264 decode, the 720p
-export) are the two that most want every core. The code on the decode
-path changed in one place this week (the still's cut size, 89f5ddd,
-which touches no video row). The software HEVC row at 26 fps against a
-budget of 24 is the one to watch: a quiet-machine run before the next
-audit should put it back near 31.
-
----
-
-## 2. Scorecard for the whole app
-
-| Aspect | 4 Oct | Today | One line |
-|---|:-:|:-:|---|
-| Code quality | 6 | 6 | fmt green, clippy clean, three `expect`s in 22.5k lines of window code; against that, more English in Rust strings, docs drifting onto the wrong items, and one SAFETY line that is not true. |
-| Design architecture | 8 | 8 | Held: the placement rule moved into the model where the window, API and CLI share it; the Hub's panics are one caller's error; the sibling compositor seam carries trials. |
-| Design philosophy | 6 | 6 | `[[replaces]]` fenced and `matches_category` made data-driven (two rules restored); the API fell further behind the window (voiceover, meters, fonts, preview); shapes still vanish under 0.2.5. |
-| Maintainability | 5 | 5 | `studio.rs` 10 435 lines in one `impl` on a 98-field struct; `run()` 1 600 lines; a vendored Slint crate to carry across upgrades; four Slint files over 1 800 lines. |
-| User likeability | 7 | 7 | A phone can export and find the file; stills keep their shape; Detach means detached; voiceover and meters; colour emoji. Against that, the Windows text fix is unseen on Windows, and the export sheet overwrites without asking. |
-| Performance | 6 | 6 | Playback publishes only the picture now and asks for frames on time; but prefetch pins frames outside any budget, package reloads trial shaders on the UI thread, and "Relink all" walks a folder on it. |
-| Scalability | 5 | 5 | Unchanged: the fd leak per API connection, the one Hub thread, the per-sample mixer scan, the per-edit bin clone. |
-| Security | 4 | 4 | The hijack is closed and model downloads are pinned and digested; but the Android release key reaches pull-request builds, 0.2.6 was built from `main` not its tag, FFmpeg is fetched from a moving release with no digest, and the update chain is as it was. |
-| Testing | 6 | 6 | 713 tests, +37 on +1 500 lines, and the suite is green and race-free locally; CI never finishes on Windows; the week's Android, record and publish code has no tests. |
-| Documentation | 6 | 6 | 0.2.6 has its section; the week since has none; four gpu.rs docs describe a CPU fallback that no longer exists; README still sells an MCP transport. |
-| CI and release | 3 | 3 | fmt is no longer the blocker, a hanging test is; the release workflow builds `main` on a hand start; the key and password reach every build step. |
-
----
-
-## 3. Scores per crate
-
-Columns: code quality, maintainability, scalability, performance,
-architecture, design philosophy, security, testing. A dash means the
-scale does not apply. Arrows are against 4 October.
-
-| Crate | CQ | Maint | Scale | Perf | Arch | Phil | Sec | Test | Note |
-|---|:-:|:-:|:-:|:-:|:-:|:-:|:-:|:-:|---|
-| concat-core | 8 | 8 | 6 | 7 ↑ | 8 | 8 | 6 | 7 | `FrameRate::checked` in lowest terms with a test; dead `Project` and `source_duration` remain; `reduce` still panics on file-derived rates and `Neg` on `i64::MIN`. |
-| concat-project | 8 | 7 | 6 | 6 | 8 ↑ | 7 | 6 | 8 ↑ | Placement is a model rule with 10 tests; 0.2.4 fields are dropped with a notice; a speed curve can still grow a clip over its neighbour, and a 1e300 start loads. |
-| concat-media | 7 | 6 ↓ | 5 | 7 | 8 | 7 | 6 | 7 | The stream-start fix landed with a fixture; every `unsafe` argued; the mux writes the final path directly; pinned prefetch frames have no budget; `cover` means two things for stills now. |
-| concat-render | 7 | 5 | 6 | 7 | 8 | 7 | 5 ↓ | 7 | Scopes gone (−692 lines); no uncaptured-error handler and no texture-limit check, so a 10 000-px frame panics the thread; four docs describe a CPU fallback that is gone. |
-| concat-effects | 8 | 7 ↓ | 6 | 5 ↓ | 8 | 8 ↑ | 6 | 8 | `Catalogue::add` refuses every take-over with a reason and tests; trials at Default, Min and Max; but they run on the UI thread's timer, the catalogue is leaked per reload, and `breaks()` miscounts a `switch` break. |
-| concat-export | 6 | 4 | 6 | 5 | 7 ↓ | 7 | 6 | 6 ↓ | Folder made, still fixed, both with tests; the audio mux is not atomic; frames that fail in the pool path vanish silently while the paced path aborts; `lib.rs` 3 015 lines. |
-| concat-host | 7 | 6 | 5 | 6 | 6 | 6 | 4 | 6 | Hub `catch_unwind` landed; memory-sized caches; voiceover; VU meters; the update chain, the cache deletion and the HDR proxy name as they were; `Server::stop` blocks the UI on a running API export. |
-| concat (Rust) | 6 ↓ | 4 | 5 | 6 | 5 ↓ | 6 | 6 | 3 ↓ | Rate ladder, volume flush, art caches, locale races fixed; the CJK face shipped once with half its coverage; dialogs open under a `RefCell` borrow; `studio.rs` grew 406 lines; 71 tests for 22.5k lines. |
-| concat (Slint) | 7 | 6 | 6 | 7 | 7 | 6 | – | – | Confirm sheet handles Escape; everything else of the 4th holds: 2 `accessible-*` properties, 158 `TouchArea`s, 15 `FocusScope`s, the category pills' English list on the phone. |
-| concat-api | 8 | 7 ↓ | 5 | 5 ↓ | 8 | 5 ↓ | 5 ↓ | 7 | One test line in the week while the window gained four operations; reads unconfined and caller strings reach FFmpeg's opener; export jobs have no panic guard. |
-| concat-server | 7 | 8 | 5 | 6 | 8 | 8 | 5 ↓ | 6 ↓ | `catch_unwind` on the Hub; the pre-auth fd leak, EMFILE spin and pre-auth seats as they were; a caller that stops reading pins its seat for good. |
-| concat-cli | 7 | 8 | – | – | 7 | 6 | 6 | 2 ↓ | No commits; no logger; `serve` parks with no signal handling, token on argv and stdout. |
-| concat-speech | 7 | 5 | 6 | 7 | 6 | 7 | 8 ↑ | 6 | No commits; downloads pinned to commit URLs with digests, staged and renamed once; the stale child-process doc and the silent DirectML fallback remain. |
-| concat-vision | 8 | 8 ↑ | 5 | 6 | 9 | 8 | 8 | 7 | Zero churn since 23 September; hashed cache names. |
-| concat-text | 7 ↓ | 6 ↓ | 5 | 6 | 8 ↓ | 8 | 7 | 6 ↓ | Colour emoji paint in their own colours, probed not trusted; the one file is 2 607 lines; 12 MB of embedded faces copied to the heap; PNG strikes decoded per render; the bitmap path never runs in CI. |
-| concat-android | 7 ↓ | 7 | – | 7 | 7 | 7 | 6 | 1 | The publisher: pending entry, copy, publish, original deleted, `[path, why]` back; a provider's runtime exception on import still kills the process; a restored fragment cannot be instantiated; 0 tests. |
-
----
-
-## 4. Scores per feature
-
-Status: **done**, **partial** (works with a known gap), **thin** (present,
-weak). Rows that changed since the 4th carry a reason; ★ marks a row new
-this week.
-
-| Feature | Where | Status | 4 Oct | Today | Why |
-|---|---|---|:-:|:-:|---|
-| Launcher, new project, recents | `ui/start.slint` | done | 8 | 8 | A folder named with either slash is one recent (7641f00). |
-| Import, bin, drag and drop | `ui/workspace/media-pane.slint` | done | 8 | 8 | ★ A click previews a file without placing it (e43e106); "N selected" still untranslated. |
-| Timeline editing | `ui/timeline/`, `placement.rs` | done | 8 | 8 | ★ A clip never lands on top of another, in the model, 10 tests (faa6da9); a click sends the playhead (2cc3e02); ★ a detached sound moves alone (e11006e); a group dragged past 0 still collapses (finding 11). |
-| Volume line | `lanes.slint`, `studio.rs:6437` | done ↑ | 6 | 7 | The pending title commit is flushed first (5c97f20); untested. |
-| Undo and redo | `editor.rs` | done | 8 | 8 | Every snapshot still clones the bin. |
-| Transitions | `plan.rs`, `concat-export/src/lib.rs:768` | partial | 7 | 7 | The hold still plays keyed effects early. |
-| Effects and filters | `concat-effects/packages/` | done | 8 | 8 | `[[replaces]]` fenced with reasons and tests (ab1c482); shelves are data (f9584ab). |
-| Colour and HDR | `model.rs`, `gpu.rs`, `encode.rs` | done | 7 | 7 | AMF makes ten-bit HEVC as Main 10 (56e94bb). |
-| Scopes | – | removed | 7 | – | The pane and its GPU counting are gone (616d8fa). |
-| Colour grading | `concat.color-wheels`, `grading.slint` | done | 7 | 7 | Unchanged. |
-| Titles and text | `concat-text`, `titles.rs` | done ↑ | 7 | 8 | ★ Colour emoji in their own colours (a03e416); ★ base fonts every title can use, presets as data, a searchable picker that says when a font is missing (97b349b, aee2d61). |
-| Shapes / Stickers | `media-pane.slint`, `clips.rs` | partial | 6 | 6 | Still vanish under 0.2.5. |
-| Keyframes, animations | `model.rs` | done | 7 | 7 | Unchanged. |
-| Crop, transform, blend, masks | `plan.rs` | done | 7 | 7 | Unchanged; shape masks of 0.2.4 still absent (#239). |
-| Cutout, smart brush | `segment.rs` | done | 7 | 7 | Unchanged. |
-| Enhance | `enhance.rs` | thin | 6 | 6 | Copies still land in `cache/`. |
-| Auto-captions | `captions.slint`, `samples.rs` | done ↑ | 7 | 8 | Words sit at the right time on MTS files now (c6d7e17, with a fixture). |
-| Text-to-speech | `tts.rs` | done | 7 | 7 | No change; the provider is still not logged. |
-| ★ Voiceover | `concat-host/src/record.rs`, `panes/voiceover.rs` | partial | – | 6 | Records from the microphone onto the timeline (1370adb); a crashed take has a zero-length header and is lost; two quick takes can land in each other's place. |
-| ★ VU meters | `concat/src/meters.rs` | thin | – | 5 | Beside the timeline (f1f2f3e); the programme meter has no feed and sits at silence. |
-| Templates | `templates.rs` | done | 7 | 7 | Poster still copied under a name nothing reads. |
-| LUT import | `cube.rs` | done | 8 | 8 | Unchanged. |
-| Custom and community packages | `catalogue.rs` | done ↑ | 6 | 7 | Trials at three settings on the sibling compositor (ab1c482); still on the UI thread, still skipped with no sibling. |
-| Proxies | `proxy.rs` | partial | 5 | 5 | The encoder is dropped before the rename (543443a); names use `DefaultHasher`. |
-| Export | `export.slint`, `concat-export` | done | 7 | 7 | ★ Makes its folder (38e45a9); the mux can leave a broken file at the final path; the sheet overwrites without asking. |
-| ★ Export on a phone | `platform.rs:599`, `ConcatFiles.java:76` | partial | – | 6 | The file goes into Movies/Concat through the media store (63f2633), Browse is gone; unseen on a real phone; a 0-row update hides the file. |
-| Hardware encode | `encode.rs:325` | done | 6 | 6 | Unchanged. |
-| Playback and audio | `playback.rs`, `samples.rs` | done ↑ | 5 | 7 | Sound at the stream start (c6d7e17); the run-on clock is one word (02bb564); the full window decoded ahead (e5931a6); the mixer scan and the realtime frees remain. |
-| ★ Monitor stills | `pool.rs:774` | done | – | 7 | A still keeps its shape (89f5ddd, pixel test); still stretched when a clip has no dimensions. |
-| Hardware decode | `hardware.rs` | done | 7 | 7 | Unchanged. |
-| Remote API | `concat-api`, `concat-server` | partial ↓ | 6 | 5 | Nine window operations the API cannot name; `stop` blocks the window on a running job. |
-| Self-update | `updates.rs` | done | 5 | 5 | Unchanged. |
-| Logs | `logs.rs` | done | 7 | 7 | Unchanged. |
-| Keyboard menus, platform keys | `app.slint`, `platform.rs` | partial | 6 | 6 | Unchanged. |
-| Save Video Frame As, Save Audio As | `studio.rs:5768`, `6364` | done | 7 | 7 | Dialogs still open under the studio borrow (finding 7). |
-| Phone shell | `ui/phone/`, `concat-android` | partial ↑ | 6 | 7 | Exports work and are findable; stills right; the category pills' English list is still shown there. |
-| Relink missing media | `relink.slint`, `panes/relink.rs` | done ↓ | 8 | 6 | "Relink all" walks the folder on the UI thread and follows symlinks with no loop guard. |
-| Themes and accent colours | `ui/theme/` | done | 8 | 8 | Unchanged. |
-| Languages | `concat/locales/`, `fonts.rs` | partial ↑ | 6 | 7 | ★ A CJK fallback face covers every character the four CJK locales use, held by a test; ja translated for real; de, fr, ko, pt-BR still carry 137-204 English values. |
-| ★ Windows text | `src/patches/i-slint-renderer-skia` | partial | – | 5 | Pixel-snapped glyphs at 1x (3989e59); not yet seen on a Windows screen. |
-| Accessibility | the whole UI | thin | 2 | 2 | Unchanged. |
-| Windows data folders | `dirs.rs:95` | done | 6 | 6 | Unchanged. |
-
----
-
-## 5. The 4 October findings, re-verified
-
-Numbers are the 4 October audit's (its section 6), with the 28 September
-items it carried marked S#.
-
-| # | Finding | Today | Evidence |
-|---|---|---|---|
-| 1 | `main` red at fmt; nothing behind it ran | **Fixed, then red again** | fmt green since 3121107 and 7b5cc44; the Windows test job (9440a67) hangs in `cards::tests::every_card_draws_and_shows_its_package_at_work` and is killed at 90 min; last green push run 37212363814, 4 Oct 15:16. CI still on `stable` (`ci.yml:38`). |
-| 2 | `[[replaces]]` can hijack a live built-in; keys dropped; silent | **Fixed** | `Catalogue::add` refuses an installed, bare or stood-in-for id (`catalogue.rs:1021-1080`); `upgrade` leaves answered links alone (`:1127-1140`); `replacing` carries keys (`:589-614`); `Session::open` records rewrites (`session.rs:109-119`); tests at `catalogue.rs:1320, 1356`. Residue: a package may still claim any id this catalogue lacks (finding 22). |
-| 3 | Preview sound and transcript ignore the stream start | **Fixed** | c6d7e17, 18 minutes after the audit: `samples.rs:177-190` seeks at `start + origin`, the trim is offset, `fixtures/tone-at-1s.ts` added; `playback.rs:803-808` salts the decode key to retire the wrong spans. |
-| 4 | Locales hold English verbatim | **Mostly unchanged** | ja 137 → 7 (3aa6e5c); de 204 → 199, fr 207 → 204, ko 137, pt-BR 179 → 175. `locales.py` still compares key sets only (`scripts/locales.py:140-158`). |
-| 5 | Shapes vanish under 0.2.5; unknown kinds rewritten | Unchanged | `wire::clip_kind` falls back to `Video` (`model.rs:1172-1175`); `DOCUMENT_VERSION` is 1 (`doc.rs:41`). |
-| 6 | Install trial at defaults only; a timeout kills the monitor | **Partly fixed** | Trials at Default, Min and Max (`catalogue.rs:479-484`) on `monitor.sibling()` (`studio.rs:7236-7249`); the sibling shares the device (`gpu.rs:727-729`), so a real hang still takes the monitor; with no sibling the package installs untried (`studio.rs:7243-7245`); `preview.rs:222-227` still answers "lost" for good. |
-| 7 | The dissolve hold plays keyed effects early | Unchanged | `pre_roll` clears placement keys only (`concat-export/src/lib.rs:768-805`). |
-| 8 | Export rate is a fixed ladder | **Fixed** | a2ba094: the timeline's rate first, `rates_around` de-duplicates (`panes/export.rs:264-302`), test at `:663-677`. |
-| 9 | Volume drag drops typed title words | **Fixed** | 5c97f20: `studio.rs:6437-6443` flushes the pending commit first. No test. |
-| 10 | Art caches never released; ungated workers; export on close | **Partly fixed** | `forget_art` on open and close (`studio.rs:2093-2102`, 2433bab); `spawn_at` epoch-gated (`host.rs:312-322`); `save` still on the ungated `spawn` (`studio.rs:2437-2447`); `close_project` cancels cutouts, enhancers and reversers, not the exporter (`:6992-6995`). |
-| 11 | A full publish per monitor frame | **Partly fixed** | `spawn_frame` sets only the picture during playback (ee28f25, `host.rs:226-254`); every other full publish rebuilds both menus, `export.data()` and `settings.data()` (`studio.rs:9201-9207`). |
-| 12 | No keyboard reach; Escape closes nothing | Unchanged (slightly better) | `dialogs/confirm.slint:46` handles Escape; `Modal` and `Button` still have no `FocusScope`; one `accessible-label` in `ui/`. |
-| 13 | Strings that bypass i18n | **Partly fixed, partly worse** | The desktop dropped the category pills (79eafc7) but the phone still shows their English list (`media-pane.slint:1421-1430`, `phone/editor.slint:903, 965`); "{n} selected", "still", "Empty", "% of height", "Merge selected" unchanged; new Rust-side English in finding 15. |
-| 14 | Emoji paint as boxes on macOS | **Fixed** | a03e416: `own_colours` counts PNG, BGRA and simple `COLR` strikes (`concat-text/src/lib.rs:630-646`), `mark` draws them (`:997-1049`). The test returns early where no colour face exists, and CI installs none (finding 24). |
-| 15 | No Unreleased section in CHANGELOG | **Fixed, then recurred** | `## 0.2.6` written (cc19703); voiceover, meters, emoji, Android publishing, base fonts, the CJK fallback, the detach change and the Windows text patch have no section (`CHANGELOG.md:10`). |
-| 16 | Code grew without tests | Mixed | +37 tests on +1 503 lines: placement 10, catalogue 2, export 2, fonts 3, rates 1, elsewhere small; the window is at 71 tests for 22.5k lines; concat-android 0 on +155 lines; record.rs tests only its helpers. |
-| 17 | README sells an MCP transport | Unchanged | `README.md:62, 79, 114`; `concat-api/src/lib.rs:7`. |
-| 18 | The CLI installs no logger | Unchanged | `concat-cli/Cargo.toml` has no `log`; `logs::open_console` exists unused (`logs.rs:97`). |
-| 19 | The API is behind the window | **Worse** | Voiceover, VU meters, the font picker, presets as data and click-to-preview are window-only; the API gained one test line (3156ddb); `message.rs:24` still promises parity. |
-| 20 | Small correctness in concat-project | Partial | `has_non_finite` lists 8 of 11 `TextStyle` floats (`commands/mod.rs:931-940`) while `tidy` sanitises all 11; `"concat": "0.1.0"` still written (`doc.rs:320`). |
-| 21 | `matches_category` hard-codes nine ids | **Fixed** | f9584ab: categories and shelves from the manifest (`catalogue.rs:425-460`). |
-| 22 | concat-media lows | Unchanged | No fallback after a hardware encoder opens; Media Foundation labelled hardware; `sws_setColorspaceDetails` result ignored; a fresh reader seeks twice (`pool.rs:367, 981-984`); AMF handed `qp_b`. |
-| 23 | Host lows | Unchanged | Silent data-folder move (`dirs.rs:113-122`); any exe under Program Files is an MSI install (`updates.rs:289-299`); poster copied as `preview.jpg` (`templates.rs:356`); the dirs test runs against the real `%APPDATA%`. |
-| 24 | Window lows | Unchanged, docs worse | Extension swapped after the overwrite prompt (`studio.rs:5774-5783, 6370-6379`); JPEG on the UI thread in `freeze_at_playhead` (`6582-6595`); token from `RandomState` (`settings.rs:855-863`); docs stacked on the wrong items, now also on `forget_art` (`studio.rs:2082-2092`). |
-| 25 | UI lows | Unchanged | Menu look-ahead 4 rows; `in-chain` unrolled 8 deep; `IconButton.label` a tooltip, not `accessible-label`. |
-| 26 | gRPC has none of the JSON caps | Unchanged | `grpc.rs:70, 129-175`. |
-| 27 | Chatterbox on phones; DirectML silent; `DirectML.dll` not staged | Unchanged | `concat/Cargo.toml:130`; `chatterbox.rs:305-313`; no `DirectML` in `build-app.yml`; `tts.rs:7` still says whisper is a child process. |
-| 28 | Windows signing secrets undeclared; password on argv | Unchanged | `build-app.yml:37-53, 716`. Repo secrets are the two Android ones only: the 0.2.6 Mac builds are ad-hoc signed, the Windows installers unsigned. |
-| S1 | The update trusts a digest beside the package; quarantine stripped | Unchanged | `updates.rs:45, 418, 477, 504, 551, 597`. |
-| S2 | The Remote API is rooted at home; no root without HOME; open-then-save | Unchanged | `concat-server/src/lib.rs:80-86`; `concat-api/src/lib.rs:244-246, 454-456`; `settings.rs:573-577`. |
-| S3 | Every JSON connection leaks an fd; seats before auth; accept spins | Unchanged | `json.rs:100-122`; `lib.rs:284-288`. |
-| S4 | Clear cache deletes reversed and enhanced media | Unchanged | `reverse.rs:88`, `enhance.rs:71`, `projects.rs:341-350`. |
-| S5 | 0.2.4 reversed or animated clips open silently changed | **Fixed** | b00ffab: `Editor::drop_retired_fields` strips them and names the clips. |
-| S6 | The embedded API opens a second GPU device | Unchanged | `concat-api/src/lib.rs:179`; `concat-export/src/lib.rs:1247-1250`. |
-| S8 | HDR proxy name and contents disagree | Unchanged | `proxy.rs:59-87, 208`; `studio.rs:2802-2829`. |
-| S9 | Hub: no `catch_unwind`, heavy work inline, slow subscriber dropped | **Partly fixed** | `catch_unwind` (67e4f9d, `hub.rs:91-104`; no effect under `--profile app`, 848d3c5); `pixels` still composites on the hub thread; the silent drop remains. |
-| S10 | Downgrade is one click | Unchanged | `settings.rs:461-468`. |
-| S11 | No ceiling on frame rate or size | **Partly fixed** | `FrameRate::checked` (e9f09df, `time.rs:351-358`) for documents and the export; file rates unbounded (`probe.rs:201`); size only `> 0`; a vast clip start panics the export (finding 9). |
-| S12 | Every undo snapshot clones the bin | Unchanged | `editor.rs:193`. |
-| S13 | The export is serial with a readback per frame; the mixer scans every clip per sample | Unchanged | `concat-export/src/lib.rs:1320-1475`; `playback.rs:1140-1170`. |
-| S14 | Font by absolute path; a missing font silent | Partial | The picker says when a family is missing (aee2d61); `fonts.add_file` result still dropped (`titles.rs:361`). |
-| S16, S17 | Server and host hygiene | Unchanged | Socket permissions, `connections()` counting subscribers, no `project.create` checks; two TLS stacks. |
-| S18 | Dead `core::Project`; `ui/demo/` compiled in | Unchanged | `timeline.rs:499-514`; `app.slint:21`. |
-
-Of 42 items: 11 fixed, 7 partly, 1 fixed-then-recurred, 21 unchanged, 2
-worse. The 4 October audit's recommended order (its section 8) was taken
-up in part: step 1 (fmt, locale isolation, the 0.2.6 section) in full;
-step 2 (`[[replaces]]`, keys, the open notice, trials at Max on a
-sibling) in full; step 3 (the stream start) in full; step 5's rate ladder
-and step 6's volume flush and art caches; steps 4, 7 and 8 not at all.
-
----
-
-## 6. New findings, by severity
-
-Numbers continue from the re-verified table. Each names where to look.
-
-### High
-
-**1. A hand-started release builds `main`, not the tag it publishes.**
-`build-app.yml:152` and `mobile.yml:58, 173` check out with no `ref`;
-`release.yml:74-82` passes the build jobs no tag, while CI (`:67-72`) and
-`publish` (`:100-104`) do check the tag out. The v0.2.6 release was run
-37483649529, a `workflow_dispatch` at head `7b5cc44`; the tag `v0.2.6`
-resolves to `76c02c7`, 35 commits earlier. The shipped 0.2.6 binaries
-carry 35 commits the tag, the CHANGELOG section and the CI gate on the tag
-never saw. Fix: `ref: ${{ inputs.tag || github.ref }}` in both build
-workflows, passed from `release.yml`. (verified)
-
-**2. `main` has not been green for five days, and the gate is blocked by
-a hanging test.** The Windows job hangs in
-`concat-host cards::tests::every_card_draws_and_shows_its_package_at_work`
-and is killed at 90 minutes (runs 37521843321, 37624855588,
-37890980349); `release.yml:67-72` reuses `ci.yml`, so no tag can pass.
-The test draws every package on a GPU the runner does not have; it should
-skip where `WgpuCompositor::new()` is `None`, or run under a timeout, and
-the job should have one too. (verified)
-
-**3. The Android release key reaches pull-request builds and every
-dependency's build script.** `mobile.yml:36` runs on `pull_request` (a
-same-repo branch gets secrets) and on `workflow_dispatch` from any branch;
-`:113-130` decodes `ANDROID_KEYSTORE` into `$RUNNER_TEMP` and writes the
-password into `$GITHUB_ENV`, so `cargo apk build` (`:133-138`) and the
-build scripts of several hundred crates can read both; the signed APK of
-unmerged code is uploaded as an artifact (`:162-166`). Sign only on
-`release.yml` runs, and pass the password to the one step that signs.
-(verified)
-
-**4. An export with sound can leave a broken file at the chosen path and
-destroy the file that was there.** `audio::mux` opens the final `output`
-directly (`concat-media/src/audio.rs:720-721`; `format::output`
-truncates); on failure `render_on` removes `silent` and `mixed` only
-(`concat-export/src/lib.rs:992-996`). A mux failure (disk full, a read
-error) leaves a truncated mp4 where the user's previous export was. Only
-the no-audio path is atomic (`rename`, `:980`). The 4 October audit's
-section 7 claimed the export "cannot write a partial output"; that holds
-for the picture, not the sound. Mux into a third scratch name and
-rename. (verified)
-
-**5. The CJK fallback face shipped this morning covered half of
-Traditional Chinese and none of Japanese's own characters.** eb18dbd's
-subset was built from a Big5 sweep that walked only the `0x40-0x7E` trail
-bytes and no JIS set: 218 of the 972 characters in `zh-TW.json` and 121
-of the 673 in `ja.json` had no glyph (個 國 們 來; 両 働 円 図). The
-reviewer read the font's cmap; the test checked registration, not
-coverage. Fixed at the head: the subset holds GB 2312, Big5 level 1, JIS
-X 0208, the Hangul syllables and every character in the four locale files
-(2.7 MB), and `fonts::tests::the_cjk_face_covers_every_locale` reads the
-character map against the locale files. Kept here as a finding because it
-is the pattern of the week: a fix landed with a test that proved the
-wrong thing. (verified, fixed)
-
-### Medium
-
-**6. Native dialogs open while the `Studio` `RefCell` is mutably
-borrowed.** `handler!` holds `shell.studio.borrow_mut()` across the
-handler (`concat/src/lib.rs:446-456`); `ExportMsg::Browse`
-(`panes/export.rs:190-195`), `RelinkMsg::RelinkAll` (`relink.rs:55`) and
-`StartMsg::Browse` (`start.rs:142-144`) call `rfd` synchronously inside
-it, and `on_ui` closures (`host.rs:350-356`) open `save_file`
-(`studio.rs:5768, 6364`). `studio.rs:5755-5760` records that on macOS
-these dialogs pump Slint's timers, and the transport timer
-(`studio.rs:2645-2647`) and the meters timer (`meters.rs:79`) both
-`borrow_mut()`. A timer that fires during the dialog panics with
-`BorrowMutError`; Export → Browse while playing is the likely path.
-(reasoned, not reproduced)
-
-**7. The export sheet overwrites an existing file without asking and
-does not sanitise the name.** `panes/export.rs:427-431` builds
-`"{folder}/{name.trim()}.mp4"`; since 38e45a9 a name with `/` silently
-creates subfolders; an empty name writes a hidden `.mp4`; a same-name
-export replaces the previous file through `rename` or `mux`; no
-`exists()` check anywhere in the window; the path shown uses the
-untrimmed name (`:524-525`). (verified)
-
-**8. A frame wider or taller than 8192 panics wgpu, and nothing bounds
-it.** Devices are requested with `DeviceDescriptor::default()`
-(`gpu.rs:702`, limit 8192); no `on_uncaptured_error` handler is installed,
-so wgpu's default handler panics; the only cap is the API's `MAX_SIDE`
-(`concat-api/src/lib.rs:75, 885`); `render_on` and `preview_sources`
-never check the size (`concat-export/src/lib.rs:882-946`) and a document's
-`is_sane` is `> 0` (`model.rs:1953-1957`). A 10 000-px timeline from the
-window panics the export or preview thread. (verified)
-
-**9. A document with a vast clip start panics the export.** `Clip::tidy`
-floors `start` at 0 only (`model.rs:1474-1475`), so `"start": 1e300`
-loads; `quantise` saturates to `i64::MAX` (`resolve.rs:446-447`),
-`time_of_frame` multiplies by the rate, and `Rational::reduce` panics
-"rational overflowed i64" (`time.rs:100`). The same `reduce` is reached by
-file-derived rates, which `probe.rs:201` builds with no `MAX_TERM` check,
-against the crate's own rule at `time.rs:103-106`. (reasoned)
-
-**10. A SAFETY comment on `set_var` is not true as written.**
-`platform.rs:347-350` says the process is single-threaded there, but
-`Gpu::acquire()` (`:333`, `gpu.rs:58-66`) has already created a wgpu
-instance, adapter and device, and logging is open (`main.rs:13`). The
-branch runs only for software adapters (llvmpipe on Linux), which start
-threads that read the environment. Set the variable before `acquire`, or
-pass the setting through configuration. (verified)
-
-**11. Dragging a group past the timeline's start collapses the group or
-refuses the move.** The drag clamps each clip to 0 on its own
-(`studio.rs:4183`) and so does `resolve_moves` (`placement.rs:87`) before
-the group's shape is checked. Two adjacent clips (A at 1 s, B at 3 s, 2 s
-each) dragged left by 2 s become A [0,2) and B [1,3), overlap, and
-`resolve_moves` returns `None` (`:97-105`): the clips snap back. On
-different lanes they are reshaped instead of shifting by `−earliest`. No
-test. (reasoned)
-
-**12. A speed curve can grow a clip over its neighbour.**
-`SetClipSpeedCurve` lengthens the clip (`commands/properties.rs:285-306`)
-with none of the ripple faa6da9 gave `SetClipSpeed` (`:108-121`); the
-"no clip ever covers another" rule (`placement.rs:4-10`) is broken by one
-command. (verified)
-
-**13. Package reloads trial shaders on the window thread and leak a
-catalogue each time.** `reload_packages` runs every trial synchronously
-in the 2-second watch timer: three passes of up to 3 s per user package
-(`studio.rs:7226, 7238-7254, 7290-7320`); `install_with` rebuilds and
-`Box::leak`s a fresh catalogue on every reload (`catalogue.rs:962, 978`),
-each holding every built-in's stitched WGSL as `Arc<str>` (`shader.rs:922,
-936-945`), not the "few kilobytes" its doc claims (`catalogue.rs:946-947`).
-Also: `breaks()` counts a `break` inside a `switch` as leaving the loop
-(`shader.rs:1128-1136`), so `loop { switch x { default { break; } } }`
-passes `bounded`. (verified)
-
-**14. "Relink all" walks the chosen folder on the UI thread and follows
-symlinks with no loop guard.** `panes/relink.rs:73-92` is an unbounded
-`read_dir` walk inside `update`, while the studio is borrowed;
-`path.is_dir()` follows symlinks, so a link to an ancestor never ends.
-The "first match wins, closest to the root" comment (`:85-86`) is false
-below the top level: a LIFO stack walks depth-first. The toast is English
-(`:130-136`). (verified)
-
-**15. User-visible strings are hard-coded in Rust, and the new panes add
-more.** Menu rows `studio.rs:9362` "Freeze frame", `9766-9768` "Sort by:
-Added/Name/Type"; toasts `3492`, `6564`, `6979-6980`; `panes/relink.rs:131`;
-`panes/export.rs:534-536` "{clips} clips · {titles} titles", `:301`
-"{name} fps", `:614` "10-bit". `locales.py` only sees `t()` keys, so none
-reaches a translator. With the 4 October's list (`media-pane.slint:2293,
-598`, `text-panel.slint:176`, `controls.slint:100`, `tray.slint:319`) and
-the phone's category pills, that is about fifteen strings in five
-languages' worth of boxes. (verified)
-
-**16. `Server::stop` blocks the window until every running API export
-has finished.** `shut` closes the Hub and joins the dispatcher
-(`concat-server/src/lib.rs:293-296`), the dispatcher runs `api.finish()`
-(`hub.rs:107`), which joins the export threads
-(`concat-api/src/lib.rs:285-287, 740-744`) and cancels none; the Remote
-page calls `server.stop()` on the UI thread (`settings.rs:557-558`).
-Turning Remote off during an API export freezes the window for the rest
-of it. Related: API export jobs run on a bare thread with no
-`catch_unwind` (`:760-788`), so a panic ends the job with no event and a
-socket client waits forever; audio decode workers (`playback.rs:675-718`)
-likewise die on a panic and leave their span stuck, and with
-`DECODE_WORKERS = 2` two such panics end audio for the session. (verified)
-
-**17. API reads are unconfined, and caller strings go straight to
-FFmpeg's opener.** `media.probe`, `media.import` and template fills pass
-caller strings to `concat_media::probe` and `Decoder`
-(`concat-api/src/lib.rs:342-344, 462-465, 486-490`), which call
-`ffmpeg::format::input(path)` (`probe.rs:93`, `decode.rs:606`) with no
-protocol or format whitelist. A token holder can name URLs or
-`concat:`/`subfile:` paths, or render any readable file FFmpeg can demux
-and read it back as PNG through `preview.frame` (`:626-640`). Whether the
-linked FFmpeg has network protocols on was not verified. Related: a
-caller that stops reading pins its seat until the server stops, since the
-outbox writer has no write timeout (`json.rs:305-311, 331-338`). (verified
-to the opener; protocols unverified)
-
-**18. Frames that fail in the export's pool path vanish silently, while
-the paced path aborts.** `if let Ok(frame) = sought.frame_at(...)`
-(`concat-export/src/lib.rs:1333-1358`) has no `else` and no log, so a
-reversed clip, a speed-curve clip or a dissolve hold that fails to decode
-exports as a missing layer; `Decoder::open(...)?` and `next_frame()?`
-(`:1400, 1408`) end the export. One policy is needed; either is
-defensible, both at once is not. Related: the still fix of 89f5ddd holds
-only when a clip carries dimensions; without them both callers fall back
-to the frame size (`:1329-1332, 1367-1370, 1817-1820`) and the stretch of
-#229 returns for API requests and older documents. (verified)
-
-**19. Prefetch "pinned" frames sit outside the pool's byte budget, and
-nothing bounds them.** Released only when the cursor passes them
-(`prefetch.rs:292-295`), not capped, not cleared on `clear()`; a backward
-seek during forward play leaves everything pinned ahead in place. Deep
-4K frames are about 66 MB each, so a few seeks pin gigabytes the
-256 MB-1 GB budget (`pool.rs:577-584`) never sees. (verified)
-
-**20. A crashed voiceover take is lost although its samples are on
-disk.** `record.rs:279, 283-298` writes the WAV header with data size 0
-and fixes it only in `finish`; a kill, a crash or a failed `write`
-(`:159`) leaves a file whose header says 0 samples, which the reader
-honours (`playback.rs:197-201`). Writing the `0xFFFFFFFF` placeholder that
-`playback.rs:180-183` already accepts would keep the take playable.
-Related: `landing` is pane state read when the asynchronous probe returns
-(`panes/voiceover.rs:89, 141`), so two quick takes can land in each
-other's place. (verified)
-
-**21. Unverified downloads go into shipped binaries, and mutable action
-tags run with write access.** FFmpeg comes from BtbN's moving `latest`
-release (`ci.yml:31`, `build-app.yml:57, 180`) and is copied into every
-desktop bundle; sherpa-onnx, ONNX Runtime and the mobile FFmpeg
-(`ci.yml:76, 96`, `sherpa-mobile.sh:38, 56`, `ffmpeg-mobile.sh:38`) are
-fetched with no digest; actions are pinned by tag, including
-`softprops/action-gh-release@v3` with `contents: write`
-(`release.yml:88-89, 139`); `release.yml:48` pastes `${{ inputs.tag }}`
-into a shell script. The default token being read-only is the one good
-line here. (verified)
-
-**22. The Android bridge has three ways to lose an import or a crash
-path.** `copy` catches `IOException` only (`ConcatFiles.java:277`) on a
-bare thread, so a provider's `SecurityException` or
-`IllegalArgumentException` kills the process and `filesPicked` never
-comes; the headless `Fragment` lives in an `InMemoryDexClassLoader`
-(`lib.rs:166-182`), so a process killed while the picker is open fails
-to restore it ("Unable to instantiate fragment"); `publishVideo` ignores
-`resolver.update`'s row count (`:115`) and then deletes the original
-(`:132`), so a 0-row update leaves the only copy pending, hidden and
-later purged. `files_picked` is `extern "system"` with no `catch_unwind`
-(`lib.rs:280-282`). (reasoned for the restore; the rest verified)
-
-### Low
-
-**23. Doc drift, again.** gpu.rs says in four places that a dead
-compositor falls back to the CPU (`gpu.rs:623-626, 1165-1166, 1502-1504,
-3161-3164`); the CPU compositor is gone and `render` returns
-`Frame::black` (`:3241-3251`). `studio.rs:2082-2092` stacks two docs on
-`forget_art` and leaves `row_track` and `ask_remove_track` undocumented.
-`FrameRequest::cover`'s doc (`pool.rs:64-68`) no longer describes stills.
-`catalogue.rs:946-947` claims a few kilobytes. `ARCHITECTURE.md:240, 260`
-still name a quarter-size proxy and a hardware-decode switch.
-`concat.iss:66-70` claimed the vendored translations were all 6.7.1
-files; the Traditional Chinese one is a 6.1 file with 276 of 296
-messages (corrected at the head).
-
-**24. Tests that prove the wrong thing or never run.** The emoji test
-returns early where no colour face exists and CI installs none
-(`ci.yml:53-56`); `a_still_keeps_its_shape_on_the_monitor` `expect`s a
-GPU, so a machine with FFmpeg and no adapter fails rather than skips;
-`platform.rs`'s tests use `edge_at` and `ResizeDirection`, which are
-`cfg(not(android))`, so `cargo test --target aarch64-linux-android` does
-not compile (`platform.rs:82, 93, 760-793`); device-loss recovery is
-tested by calling `recovered()` directly, never through
-`render_picture`; `concat-server` has no test for fd release or a slow
-reader; `concat-cli` none for `api` or `serve`; `concat-android` none.
-
-**25. The Rational edges.** `Neg` and `ceil` overflow for `num ==
-i64::MIN`, which `checked_reduce` lets through (`time.rs:116-121,
-155-157, 250-256`); `AddClipAtFirstFree` accepts a negative start
-(`commands/clips.rs:49-62`) while every other add clamps; `record.rs:287,
-315` overflow the RIFF size after about 12 hours of mono 48 kHz; proxy
-names use `DefaultHasher` (`proxy.rs:66`) against the reasoning at
-`playback.rs:783-785`; a proxy job that panics is never retried or swept
-(`proxy.rs:136-149`).
-
-**26. HDR side data sized by a Rust mirror of structs FFmpeg says are
-not ABI.** `size_of::<MasteringDisplay>()` and `size_of::<ContentLight>()`
-at `encode.rs:532-537, 1073-1079, 1281-1287` against structs at `:751,
-798`; `av_mastering_display_metadata_alloc_size` and
-`av_content_light_metadata_alloc` exist for this. The SAFETY comments are
-true only while the mirror matches the linked FFmpeg.
-
-**27. Realtime audio does work it should not.** The mix callback drops
-the previous `Vec<ActiveClip>` on `SetClips` (`playback.rs:1107`), which
-can `munmap` on the audio thread; samples are read from the mmap
-(`:169-174`) so a cold page faults inside the callback; the recorder
-allocates per input callback (`record.rs:222-227`); a seek while playing
-can stamp the old position for one buffer (`:521-527, 1195`). The mmap's
-SAFETY (`:159`) holds within one process only: a sync tool truncating a
-project folder's PCM cache is a SIGBUS.
-
-**28. Blocking work on the UI thread, beyond finding 14.**
-`clear_project_cache` deletes the cache folder inline (`studio.rs:6978`);
-`open_project` runs `proxy::sweep` inline (`:6849-6862`);
-`adopt_reversed`/`adopt_enhanced` probe with FFmpeg inline (`:6052,
-6077`); `freeze_at_playhead` decodes and encodes inline (`:6582-6595`);
-`Fonts::new` copies 12 MB of embedded faces to the heap
-(`concat-text/src/lib.rs:323-325`) and PNG emoji strikes are decoded on
-every render (`:1005-1011, 1058`).
-
-**29. Hygiene.** `wayland_drop::receive` sets CLOEXEC after `pipe()`
-rather than with `pipe2` (`wayland_drop.rs:341-350`) and grows `bytes`
-without bound (`:365-377`); `i18n::fill` substitutes placeholders in
-sequence, so an argument containing `{1}` is replaced again
-(`i18n.rs:337-343`); `english().by_text` maps duplicate values to an
-arbitrary key (`:94-101`); `std::mem::take` routing lets a re-entrant read
-see a blank pane (`studio.rs:7085-7138`); the crate uses
-`slint::private_unstable_api` for focus loss and drags
-(`platform.rs:156`, `lib.rs:23`) on a caret version, now beside a vendored
-renderer; `Config.socket` is unlinked before the bind whatever it names
-(`concat-server/src/lib.rs:195`); the token goes on argv and stdout
-(`concat-cli/src/main.rs:84-85, 228-232`); `serve` parks with no signal
-handling (`:247-249`); `project.create` accepts any size (`concat-api/src/
-lib.rs:377-393`); a user package may still claim any id this catalogue
-lacks (`catalogue.rs:1047-1072`); `JObject::from_raw` wraps a global
-reference where jni 0.22 asks for a local one (`concat-android/src/
-lib.rs:139`), harmless today and unargued.
-
----
-
-## 7. What is good, and should stay that way
-
-- **The placement rule is a model rule.** `concat-project/src/placement.rs`
- decides where a clip lands, for the window, the API and the CLI alike,
- with ten tests; the drag asks it as it goes so the echo shows where the
- drop will land (faa6da9).
-- **`Catalogue::add` refuses every take-over with a reason**, the built-in
- author is reserved, a LUT must be a plain `.cube` beside its manifest,
- and `replacing` carries keyframes through the knob expression and clamps
- them (`catalogue.rs:1021-1080, 552-622, 727-744`; `manifest.rs:533-537`).
-- **Model downloads are pinned to commit URLs with digests**, an empty
- digest is refused, archives are verified before `unpack_in` into a
- staging folder and renamed once (`chatterbox.rs:120-178`,
- `transcribe.rs:118`, `concat-vision/src/models.rs:83-101`,
- `concat-host/src/models.rs:295-301`, `tts.rs:522-580`).
-- **Rational time is done carefully**: i128 intermediates, `checked_new`
- for file values, `FrameRate::checked` judged in lowest terms with a test
- of 60000000/2002000 (`time.rs:100-122, 351-358, 445-456`).
-- **Hardware decode falls back at all three stages** (open, copy-back,
- send) and `resume_in_software` picks up after the last delivered pts
- (`decode.rs:640-647, 769-794, 830-843, 873-907`).
-- **The prefetcher cannot be wedged by a panic**: a `Finished` drop guard
- plus `catch_unwind`, and the queued-vs-pinned read documented under one
- lock order (`prefetch.rs:288-296, 415-450`).
-- **Constant-time token compare shared by both transports**, minted from
- `getrandom`, bounded line reads before auth, a 10 s auth timeout, with
- tests (`token.rs:16-36, 52-64`; `json.rs:35-40, 256-263, 385-414`).
-- **Atomic project save** (temp, `sync_all`, rename) and a session that
- refuses a newer or unreadable document (`projects.rs:171-195`;
- `session.rs:80-104`). The export's picture path is atomic the same way;
- the sound path should be (finding 4).
-- **The epoch gate covers art now**, `spawn_frame` publishes only the
- picture during playback, and `forget_art` drops every per-project cache
- on open and close (`host.rs:197-254, 312-322`; `studio.rs:2093-2102`).
-- **`platform.rs` keeps platform differences in one file with the reason
- and issue at each branch**, and the phone's publish hook is confined to
- it, off the UI thread, a no-op on a desk (`platform.rs:336-351, 439-461,
- 599-665`).
-- **`COLR` glyphs are probed, not trusted**: anything beyond solid layers
- falls back to the outline; a strike that lies about its size is
- rejected (`concat-text/src/lib.rs:630-713, 1056-1071`).
-- **Very few panics in the window**: three invariant `expect`s in 22.5k
- lines, and a startup failure reaches the user through a native dialog
- and the log (`lib.rs:126-130`).
-- **The fixes of the day each carry a test that fails without them**: the
- still's shape reads the stage either side of the picture; the export
- folder test asserts the folder whatever the GPU does; the CJK face is
- held to the locale files' characters.
-
----
-
-## 8. Where we stand
-
-The 4 October audit ended with eight recommended steps. Five days later:
-
-| Step (4 Oct) | Landed |
-|---|---|
-| 1. fmt, toolchain in CI, isolate the locale tests, write 0.2.6 | fmt and the locale lock, the 0.2.6 section; CI still on `stable`; 0.2.6 then built from `main` (finding 1) |
-| 2. `[[replaces]]` fence, keys through `replacing`, the open notice, trials at Max on a sibling | all four (ab1c482) |
-| 3. `AudioDecoder::open` takes the stream start, MTS fixture | done, 18 minutes after the audit (c6d7e17) |
-| 4. `DOCUMENT_VERSION` for shapes; `extra` for unknown kinds; `ClipPatch.shape`; round-trip all commands | nothing |
-| 5. locale check for English values; the six literals through `I18n`; export rate from the timeline; freeze keys at the hold | the rate (a2ba094); the rest open, and more literals arrived |
-| 6. Flush in `clip_volume`; gate art and save; publish-lanes-only frames | the flush and art (5c97f20, 2433bab, ee28f25); `save` still ungated |
-| 7. `FocusScope` on `Button`, Escape on `Modal`, `accessible-label` | Escape on the confirm sheet only |
-| 8. The 28 September list: sign the update, the Remote page's root and fd leak, processed media out of `cache/`, migration notices, the export pipeline, the API verbs | nothing; the API gap grew |
-
-Half the list was taken up, and taken up well: every item that landed
-has a test or a fixture, and two of the three headline findings of the
-4th are closed. What did not move is the same layer that has not moved
-since 28 September: the update chain, the server's seats and fds, the
-API's reach, the document's version. The week's new risk is in the
-release pipeline rather than the code: a release that is not what its
-tag says, a gate that cannot pass, and a signing key that every
-pull-request build can read. Those three are an afternoon.
-
-The fixes made during the audit followed the same shape as the week's:
-reported, reproduced where a machine allowed, fixed in one place with a
-test that fails without the fix, and explained to the reporter. One of
-them (the CJK face) was wrong on first landing and caught by this audit's
-own reviewer within hours, which is the argument for the audit cadence
-as much as for the reviewer.
-
-**Recommended order.** Each line is an hour to a few days, and each
-closes a finding above.
-
-1. `ref:` the tag in `build-app.yml` and `mobile.yml`; sign Android only
- on release runs, password to the signing step alone; skip or time out
- the cards test where there is no GPU, and give the Windows job a
- timeout; `rust-toolchain.toml` into CI. Then tag 0.2.7 from a green
- `main`. (1, 2, 3)
-2. Mux into a third scratch name and rename; one failure policy for the
- export's two decode paths; a size ceiling in `is_sane` and in
- `render_on`, with an `on_uncaptured_error` handler on every device;
- bound `Clip::tidy` and probed rates. (4, 8, 9, 18)
-3. The export sheet: refuse `/` and an empty name, ask before
- overwriting, show the trimmed path. Dialogs on the next loop turn
- everywhere the studio is borrowed. (6, 7)
-4. Trials off the UI thread with one catalogue per reload; the relink
- walk on a worker with a visited set; the cache clear and the sweep on
- workers. (13, 14, 28)
-5. `Server::stop` cancels jobs first and runs off the UI; `catch_unwind`
- on API jobs and decode workers; a protocol whitelist in front of
- FFmpeg's opener; a write timeout on the outbox. (16, 17)
-6. The pinned-frame budget; the WAV placeholder header; the group clamp
- by `−earliest`; ripple in `SetClipSpeedCurve`. (19, 20, 11, 12)
-7. The fifteen literals through `I18n`; a `locales.py` that flags English
- values; the Rust-side strings as keys. (15, and the 4th's #4, #13)
-8. Digests for FFmpeg, sherpa and ONNX downloads; actions pinned by
- SHA; `inputs.tag` through `env:`. (21)
-9. Look at Windows text at 100 % (3989e59) and a phone's export (63f2633)
- on real hardware; neither has been seen. Then the 28 September list, in
- its order.
-
-After those, phase 4 of the HDR plan (a true HDR preview), then phase 5,
-in the plan's order.
-
----
-
-## 9. Caveats
-
-Scores are judgements over code, calibrated against the 4 October scores
-but not measured against users or another editor. Every crate was read
-in full by one of four reviewers; the headline findings were re-read by
-hand. Findings marked (verified) were read at the line and, where a
-command could show it, run; (reasoned) means the path was followed
-through the code and not executed. The perf table ran on a loaded
-machine and is a floor. The app was not launched on a desk; the 0.2.6
-APK ran on an emulator only as far as its window's first second, which
-was enough to read the encoder probe and not enough to drive the editor.
-Two of the day's fixes, Windows text and the phone's Movies publishing,
-have not been seen on the hardware they are for. `cargo audit` was not
-run.
diff --git a/audit/AUDIT-2026-10-11-QUICK.md b/audit/AUDIT-2026-10-11-QUICK.md
deleted file mode 100644
index dab7bf60..00000000
--- a/audit/AUDIT-2026-10-11-QUICK.md
+++ /dev/null
@@ -1,132 +0,0 @@
-# Concat audit, 11 Oct 2026 — the quick version
-
-Same audit as `AUDIT-2026-10-11.md`, one screen. Each bar has 10 slots:
-🟢 filled = good, ⚪ empty = missing. Severity: 🔴 critical · 🟠 high · 🟡 medium · ⚪ low.
-Arrows are against the 9 Oct audit. Every crate was read in full by one of four reviewers.
-
-## 🎯 Verdict
-
-**Three of nine steps in two days, and the gate is still shut.** Eight commits after the 9 October audit, its three headline holes are closed and closed well: a release builds its tag and the Android key reaches the signing step alone (3f3b7de); an export can no longer leave a broken file, and no document can panic the GPU (a2dd5e3); the export sheet asks before replacing and refuses a name that is not a file's (9a2e5f2). Copy and paste works across timelines as one undo step. Each fix carries a test that fails without it. But `main` has now been red for a week: the Windows job still times out at 90 minutes **after** the fix meant to free it, nobody can read why, and no tag can ship. Three of the eight native dialogs still open under the studio borrow, a paste onto a busy lane lands in the wrong place with the wrong frames, and the toolchain action now runs from a moving branch. 🟠 ×1, 🟡 ×9 new, no 🔴.
-
-## 📏 Size
-
-| | Lines | Δ vs 9 Oct |
-|---|---:|---:|
-| Rust, all crates | 89 447 | +568 |
-| Slint UI | 29 790 | +73 |
-| WGSL, packages | 1 764 | 0 |
-| Tests (`#[test]`) | 719 | +4 like for like |
-| Largest file `concat/src/studio.rs` | 10 689 | +254 |
-
-Checks today: fmt clean · clippy 0 warnings · tests and perf see the full audit · locales 0 missing (the script crashes on a Windows console) · 23/23 models digested · CI on `main` red, 7 days.
-
-## 📊 The whole app
-
-| Aspect | Bar | 9 Oct | Today |
-|---|---|:-:|:-:|
-| 🧹 Code quality | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 🏛️ Architecture | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 |
-| 🧭 Design philosophy | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 🔧 Maintainability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 |
-| 😊 User likeability | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 |
-| ⚡ Performance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 📈 Scalability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 |
-| 🔐 Security | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 4 | 5 ↑ |
-| 🧪 Testing | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 📚 Docs | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 6 | 5 ↓ |
-| 🤖 CI and release | 🟢🟢🟢🟢⚪⚪⚪⚪⚪⚪ | 3 | 4 ↑ |
-
-Security up one: the key, the tag, the ceilings, the atomic export. Docs down one: the CHANGELOG is 45 commits behind and `ARCHITECTURE.md` had to be rewritten. CI up one for a release workflow that is now right, held at 4 by a gate that cannot pass.
-
-## 🧱 Per crate (average of its scales)
-
-| Crate | Lines | Δ | Bar | Avg | One thing |
-|---|---:|---:|---|:-:|---|
-| concat-vision | 2 760 | 0 | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7.5 | 🟢 untouched since 23 Sept · ⚪ runtime wrapper untested |
-| concat-core | 2 897 | +6 | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7.4 ↑ | 🟢 `MAX_SIDE` is the one ceiling everyone reads · ⚪ two dead items, `Neg` on `i64::MIN` |
-| concat-effects | 7 256 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.3 | 🟢 no churn · 🟡 trials on the UI thread, a catalogue leaked per reload |
-| concat-project | 10 285 | +69 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.1 ↓ | 🟢 every clip time held to 100 h, with a test · ⚪ all 114 tests in one 4 258-line module |
-| concat-media | 9 854 | +22 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.8 ↑ | 🟢 absurd rates refused at the probe; 38 `unsafe`, 38 SAFETY · ⚪ a VFR average past 1e6 now fails the import |
-| concat-render | 7 450 | +18 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.6 ↑ | 🟢 an uncaptured error logs, nothing dies · 🟡 but the frame it leaves is written unmarked |
-| concat-server | 1 393 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.6 | 🟢 Hub panics are one caller's error · 🟡 `stop` can hang for good on `0.0.0.0` on Windows; a refused thread kills the accept loop |
-| concat-text | 2 607 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.6 | 🟢 colour emoji probed, not trusted · ⚪ 12 MB to the heap, PNG per render |
-| concat-speech | 3 109 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.5 | 🟢 downloads pinned to commits · ⚪ the DirectML failure of #291 is a bare error |
-| concat-export | 3 871 | +27 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.1 ↑ | 🟢 one rename, one failure policy · ⚪ no test for the failure paths; scratch names collide |
-| concat-api | 2 352 | +1 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.1 ↓ | 🟢 `MAX_SIDE` is the engine's · 🟡 two more window operations with no verb; `export.run` overwrites silently |
-| concat-android | 398 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.1 ↑ | 🟢 the release key stays out of pull requests · ⚪ 0 tests, three crash paths |
-| concat (Slint) | 29 790 | +73 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.7 ↑ | 🟢 the tray's nested popup is gone · 🟡 2 `accessible-*`, no keyboard into the new menu |
-| concat-cli | 514 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 | ⚪ no logger, no signals, token on argv |
-| concat-host | 10 640 | +10 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.8 | 🟢 the cards test skips on WARP by reading the adapter · 🟡 the update chain and `stop` on the UI thread as they were |
-| concat (Rust) | 22 969 | +415 | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5.4 ↑ | 🟢 `after_dialog` is the right seam; the paste plan is pure and tested · 🟡 three dialogs missed; the paste lands wrong on a busy lane; `studio.rs` +254 |
-
-## 🎛️ Per feature
-
-| Feature | Bar | 9 Oct | Today | State |
-|---|---|:-:|:-:|---|
-| Launcher | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ Browse off the borrow; Open still on it |
-| Import and bin | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Timeline editing | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ ★ copy and paste across timelines, one undo step |
-| ★ Copy and paste | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ the plan is right; the landing is not on a busy lane; a title keeps a quarter |
-| Export | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7 | 8 ↑ | ✅ ★ atomic with sound; ★ asks before replacing; a GPU error is an unmarked frame |
-| Save Frame, Save Audio | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7 | 8 ↑ | ✅ dialogs on their own turn of the loop |
-| Titles and text | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Auto-captions | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Effects and filters | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Undo / redo | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| LUTs | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Themes | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ |
-| Playback and audio | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Transitions | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ the hold plays keyed effects early |
-| Colour and HDR | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Colour grading | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Keyframes | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Crop, blend, masks | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ shape masks still absent (#239) |
-| Cutout | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Text-to-speech | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ Chatterbox on DirectML fails bare (#291) |
-| Templates | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Custom packages | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ trials on the UI thread |
-| Monitor stills | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Hardware decode | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| Phone shell | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ⚠️ unseen on a phone; iOS cannot create a project (#299) |
-| Languages | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ the week's keys in all 14; seven locales half English by value |
-| Volume line | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ untested |
-| Logs | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 | ✅ |
-| VU meters | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5 | 6 ↑ | ⚠️ ★ a ruler, a peak readout, fed from playback; no test |
-| Voiceover | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ a crashed take is lost |
-| Export on a phone | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ unseen on a phone |
-| Relink | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ picker off the borrow; the walk still on the UI thread |
-| Hardware encode | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ✅ |
-| Enhance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ copies in `cache/` |
-| Shapes / Stickers | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ vanish under 0.2.5; a pasted shape loses its transform |
-| Keyboard menus | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ |
-| Windows data folders | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ✅ |
-| Remote API | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ seven operations with no verb; `stop` blocks, and may hang on Windows |
-| Proxies | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ |
-| Self-update | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ no signature |
-| Windows text | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ unseen on Windows |
-| Accessibility | 🟢🟢⚪⚪⚪⚪⚪⚪⚪⚪ | 2 | 2 | ❌ |
-
-## 🚨 Findings
-
-🟠 **High**
-1. `main` is still red after the fix: the Windows job times out at 90 min in `cargo test` on both runs since 3f3b7de, the log is admin-only, the cards test should skip, so something else hangs. No tag can ship (`ci.yml:272`, `release.yml:67-72`).
-
-🟡 **Medium** — three dialogs still open under the studio borrow: Open, Import LUT, the menu's Import (2) · a paste lands wrong on a busy lane (the whole media is placed before the trim) and a pasted title keeps style, length and height only (3) · `dtolnay/rust-toolchain@master` in six jobs (4) · CHANGELOG 45 commits behind (5) · `Server::stop` hangs for good on a wildcard bind on Windows (6) · a refused thread kills the accept loop silently (7) · an uncaptured GPU error is an unmarked export frame (8) · `studio.rs` +254, one `impl` of 215 methods on 93 fields (9) · `export.run` and `project.create` write where the window now asks (10).
-
-⚪ **Low** — mux scratch names collide across exporters (11) · past-the-end differs between decode paths (12) · `pick_rate` and VFR averages (13) · the ceilings' five leftovers (14) · gRPC stream open after a drop, a poisoned mutex, a non-UTF-8 CLI line (15) · the cards skip under `CONCAT_REQUIRE_GPU`, no timeout on the engine job (16) · apksigner's alias and password edges, undeclared Windows signing secrets (17) · `refresh_art` skipped, the clipboard outlives the project (18) · concat-project's 4 258-line test module (19) · `locales.py` crashes on cp1252 (20) · five CPU-fallback comments, "Slint 1.17", three crates missing from `src/README.md` (21) · dead `source_duration` and `Project` (22).
-
-## 🧭 What the 9 Oct audit asked, and what landed
-
-✅ `ref:` the tag, Android signed on release runs only, the cards skip, the toolchain pin · ✅ atomic mux, one failure policy, `MAX_SIDE` with handlers, `MAX_TIME` and checked rates · ✅ the export sheet; 5 of 8 dialogs · ❌ trials, relink, cache clear off the UI thread · ❌ `Server::stop`, `catch_unwind`, the whitelist · ❌ pinned-frame budget, WAV header, group clamp, speed-curve ripple · ❌ the literals, the locale check · ❌ digests and SHA pins (one action moved to a branch) · ❌ anything seen on real Windows or a real phone.
-
-## ▶️ Next, in order
-
-1. Make the Windows job name its hang (nextest per-test timeout, or `--test-threads=1 --nocapture`); skip the GPU suites on WARP; pin the toolchain action to a SHA; write the CHANGELOG; tag 0.2.7 from a green `main`.
-2. The three dialogs through `after_dialog`; relink, cache clear, sweep and trials off the UI thread.
-3. The paste: place at the trimmed length, carry the full patch for titles, clear with the project; move the placement half into `concat-project`.
-4. `Server::stop` on loopback and cancelling first; `Builder::spawn`; `catch_unwind` on jobs and workers; a protocol whitelist; a write timeout; `exists` and `is_sane` on the API's writing verbs.
-5. Count uncaptured errors and say so; random scratch suffixes; one past-the-end rule; a VFR fixture.
-6. Digests and SHA pins; declare the Windows signing secrets; a release against a staging tag with the real key.
-7. Pinned-frame budget; WAV header; group clamp; speed-curve ripple; `DOCUMENT_VERSION` for shapes.
-8. The literals through `I18n`; a locale check that flags English and prints anywhere.
-9. The 28 Sept list; then Windows text and the phone's export on real hardware.
diff --git a/audit/AUDIT-2026-10-11.md b/audit/AUDIT-2026-10-11.md
deleted file mode 100644
index d5dc15b0..00000000
--- a/audit/AUDIT-2026-10-11.md
+++ /dev/null
@@ -1,641 +0,0 @@
-# Concat audit, 11 October 2026
-
-Where the codebase stands two days after the 9 October audit, at
-`7223d64` (10 October), 71 commits past the 0.2.6 tag and 8 commits past
-the last audit. Same method, same scales, every finding from the 9th (and
-what was carried from the 4th and the 28th) re-verified against today's
-tree. Every finding names a file and a line. The sibling
-`AUDIT-2026-10-11-QUICK.md` is this audit as a one-screen scorecard.
-
-**How it was done.** Four reviewers each read one part of the tree in
-full: the window's Rust, Slint and locales; host, server, API and CLI;
-media, render, export and core; project, effects, text, vision, speech,
-Android, perf, CI and release. Each re-verified the 9 October findings in
-its part against today's lines, read the commits since the audit as
-commits (does each do what its message says, with what test, missing
-what), scored its part and reported new findings with locations. The
-headline items (the Windows job still timing out after the fix that was
-meant to free it; the three dialogs still opened under the studio borrow;
-the paste that does not keep its shape) were re-read by hand against the
-code and, for CI, against the workflow runs through GitHub's API. Line,
-test and `unsafe` tallies come from the tree by the same counting as the
-9th (Rust lines are `src/` only, without `build.rs` and `tests/`
-directories; the unwrap column's bracket is the count outside test
-modules). `cargo fmt --check`, `cargo clippy --workspace --all-targets`,
-`cargo test --workspace`, `scripts/locales.py --check`,
-`scripts/models.py --check` and `cargo run -p concat-perf --release --
---check` were run on this machine (Windows 10, x86_64, FFmpeg n8.1 shared
-build) today. The app itself was not launched; no phone build was made.
-
-**What the eight commits are.** All on 9 and 10 October, six by the
-maintainer and one by a contributor merged on the 10th: the 9 October
-audit itself (48e48a1); the release pipeline's three holes closed
-(3f3b7de: a release builds its tag, Android is signed on release runs
-alone, the cards test skips on a software adapter, the toolchain is read
-from the pin); the data-loss class closed (a2dd5e3: an atomic mux, one
-failure policy for the export's two decode paths, `MAX_SIDE` and
-`MAX_TIME` as the ceilings a document is held to, uncaptured-error
-handlers on both devices); the export sheet made careful (9a2e5f2: asks
-before replacing, refuses a name that is not a file's, dialogs through a
-new `host::after_dialog`); copy and paste of a selection across timelines
-as one undo step (c260c2d, #290); the meters pane as one instrument
-(948b6db); and the add-track menu's rows taking their clicks (441544e,
-#296). The first three took the 9 October audit's steps 1 to 3 the same
-day. The reviewers judged all seven; section 6 is largely about them.
-
-**Scale.** 0 to 10. 8 and up means "keep doing this"; 5 to 7 means
-"works, with known debt"; 4 and under means "will bite at the next order
-of magnitude, or already bites".
-
----
-
-## 1. The numbers
-
-### 1.1 Size, tests and hygiene per crate
-
-Deltas are against the 9 October table. Crates no commit touched carry
-the 9 October figures unchanged. The test count now includes
-concat-server's two `#[tokio::test]`s, which the earlier counts left out;
-like for like the suite grew by 4.
-
-| Crate | Rust lines | Δ | Slint / WGSL / Java | Tests (files) | `unsafe` | `unwrap`/`expect` (non-test) | `panic!` family (non-test) |
-|---|---:|---:|---:|---:|---:|---:|---:|
-| concat (window) | 22 969 | +415 | 29 790 slint (+73) | 73 (18) | 12 | 29 (3) | 0 |
-| concat-host | 10 640 | +10 | | 85 (19) | 4 | 243 (2) | 0 |
-| concat-project | 10 285 | +69 | | 114 (1) | 0 | 356 (11) | 6 |
-| concat-media | 9 854 | +22 | | 97 (11) | 38 | 204 (23) | 0 |
-| concat-render | 7 450 | +18 | shaders inline in `gpu.rs` | 86 (5) | 1 | 86 (68) | 3 |
-| concat-effects | 7 256 | 0 | 1 764 wgsl | 72 (9) | 0 | 134 (5) | 2 |
-| concat-export | 3 871 | +27 | | 31 (3) | 0 | 28 (0) | 0 |
-| concat-speech | 3 109 | 0 | | 19 (3) | 3 | 23 (2) | 0 |
-| concat-core | 2 897 | +6 | | 49 (7) | 0 | 26 (4) | 2 |
-| concat-vision | 2 760 | 0 | | 31 (8) | 0 | 22 (2) | 0 |
-| concat-text | 2 607 | 0 | | 24 (1) | 1 | 67 (1) | 0 |
-| concat-api | 2 352 | +1 | | 21 (2) | 0 | 23 (0) | 0 |
-| concat-server | 1 393 | 0 | | 12 (4) | 0 | 28 (0) | 0 |
-| concat-perf | 1 092 | 0 | | 2 (1) | 0 | 34 (34) | 0 |
-| concat-cli | 514 | 0 | | 3 (1) | 0 | 2 (2) | 0 |
-| concat-android | 398 | 0 | 291 java | 0 (0) | 10 | 1 (1) | 0 |
-| **Total** | **89 447** | **+568** | **~31 850** | **719 (93)** | **69** | **1 306 (158)** | **13** |
-
-The tree grew by 0.6 % in two days, almost all of it the window (+415
-Rust: the paste, +300 in `studio.rs`; the export sheet's name check and
-confirmation; `after_dialog`) and the data-loss fixes spread over six
-crates (+143). `studio.rs` is 10 689 lines (10 435 on the 9th; 47 % of the
-window's Rust): one `impl Studio` of 8 371 lines and 215 methods over a
-struct of 93 fields, and `run()` in `lib.rs` is 1 727 lines. No `unsafe`
-was added; the 69 all still carry a SAFETY line, and the one that is not
-true as written (9 October's finding 10) is as it was. Four tests landed
-on 568 lines: `paste_plan`, the export name check, the hundred-hour clamp,
-the absurd probed rate.
-
-Other facts: 123 effect packages, unchanged; 14 locale files at 1 016
-keys each (+5 this week: the overwrite question, the name error, the peak
-readout's tooltip), 0 missing and 0 stale, every new key translated in
-all fourteen, and de, fr, ko, pt-BR, it, es and tr still carrying 155 to
-204 English values (ja, ru, zh-Hans, zh-TW and fa fewer than ten); 882
-commits since 25 August, 8 since the last audit (40 files, +2 036 /
-−368); 30 tags, latest `v0.2.6`, 71 commits behind `main`; `Cargo.lock`
-unchanged at 817 packages plus the vendored Skia renderer patch. 89 open
-issues and 16 open pull requests on GitHub, 11 of the PRs from one
-contributor between 6 and 7 October, unreviewed.
-
-### 1.2 The checks, run today
-
-| Check | Result |
-|---|---|
-| `cargo fmt --all -- --check` | clean |
-| `cargo clippy --workspace --all-targets` | 0 warnings |
-| `cargo test --workspace --no-fail-fast` | see section 1.3 |
-| `scripts/locales.py --check` | 14 locales, 1 016 keys, 0 missing, 0 stale; **crashes on a Windows console** (finding 20) and passes under `python -X utf8` |
-| `scripts/models.py --check` | 23 models, 3.37 GB, mirrored on `models-v1` |
-| `cargo run -p concat-perf --release -- --check` | see section 1.3 |
-| `cargo audit` | not installed on this machine; not run |
-| CI on `main` | **red at head, and on every completed push since 4 October** |
-
-**CI is red for the same reason, after the fix.** 3f3b7de made the cards
-test skip on a software adapter (`concat-host/src/cards.rs:598-605`,
-`WgpuCompositor::on_software_adapter` at `concat-render/src/gpu.rs:
-737-739`; wgpu-hal 30.0.1 reports WARP as `DeviceType::Cpu`), and gave the
-Windows job a 120-minute limit (`ci.yml:201`). The two push runs since
-(37964005984 at 948b6db, 38022075165 at 7223d64) both failed in the same
-place: the step `Run cargo test --workspace -j 1` "has timed out after 90
-minutes" (`ci.yml:272`), so the job limit never fires. The engine, wasm and
-macOS jobs pass. The step's log needs admin rights to read (HTTP 403), so
-which test now hangs cannot be told from outside; the cards test should
-skip, which leaves the GPU parity suite, which the job's own comment says
-runs on WARP (`ci.yml:182-183`), as the next suspect. `release.yml:67-72`
-reuses `ci.yml` as the gate, so a tag today still fails before it builds
-anything. Every job now reads the toolchain from `src/rust-toolchain.toml`
-(`ci.yml:41-46`), through `dtolnay/rust-toolchain@master` (finding 4).
-
-### 1.3 Tests and the perf table
-
-**Not complete at the time of writing.** `cargo test --workspace
---no-fail-fast` and the perf table were started on this machine and were
-still compiling the window crate's test target when the audit was closed
-(the window reviewer's own `cargo test -p concat --lib` held the build
-lock for the first hour). fmt, clippy, the locale and model checks ran to
-completion and are in the table above. The 9 October run was 711 passed,
-0 failed, 1 ignored, and 22 of 22 perf scenarios in budget; the four
-tests added since each pass by inspection and in their authors' runs.
-Rerun both and paste the results here before this audit is cited for a
-release:
-
-```sh
-cd src
-PATH="../vendor/ffmpeg-n8.1-latest-win64-gpl-shared-8.1/bin:$PATH" cargo test --workspace --no-fail-fast
-cargo run -p concat-perf --release -- --check
-```
-
----
-
-## 2. Scorecard for the whole app
-
-| Aspect | 9 Oct | Today | One line |
-|---|:-:|:-:|---|
-| Code quality | 6 | 6 | fmt green, clippy clean, three `expect`s in 23k lines of window code; against that, a commit message that claims more than its code does (the paste), three dialogs the new seam missed, and the same English in Rust strings. |
-| Design architecture | 8 | 8 | Held, and `after_dialog` is the right seam: a dialog's answer is a message, like a worker's. `MAX_SIDE` is one constant every caller shares. |
-| Design philosophy | 6 | 6 | The paste reuses the model's placement rule; but a title's copy keeps a quarter of the clip, and the API fell two more operations behind the window. |
-| Maintainability | 5 | 5 | `studio.rs` 10 689 lines, +254 in two days, one `impl` of 215 methods on a 93-field struct; `run()` 1 727 lines; a vendored Slint crate; concat-project's whole test suite in one 4 258-line module. |
-| User likeability | 7 | 7 | The export asks before replacing and says what is wrong with a name; copy and paste across timelines; one instrument of meters. Against that, a paste onto a busy lane lands wrong, and the Windows text fix is still unseen on Windows. |
-| Performance | 6 | 6 | Unchanged: pinned prefetch frames outside any budget, trials and the relink walk on the UI thread, the serial export. |
-| Scalability | 5 | 5 | Unchanged: the fd per connection, the one Hub thread, the per-sample mixer scan, the per-edit bin clone. |
-| Security | 4 | 5 | A release builds its tag and the Android key reaches the signing step alone; a document cannot panic the GPU; an export cannot leave a broken file. Against that, a toolchain action on a moving branch, downloads still without digests, the update chain as it was. |
-| Testing | 6 | 6 | Four tests on 568 lines, each failing without its fix; but the mux's failure path, `render_on`'s refusal and `after_dialog` have none, and CI still cannot finish on Windows, so nothing is proven there. |
-| Documentation | 6 | 5 | `CHANGELOG.md` has no line for the 45 non-merge commits since 0.2.6; `ARCHITECTURE.md` was stale in two places and is rewritten today beside this audit; `src/README.md`'s crate table omits three crates; five comments still describe a CPU compositor fallback that is gone. |
-| CI and release | 3 | 4 | The release workflow is right now: the tag, the key, the pin. The gate still cannot pass, two runs after the fix, for a reason the public logs do not show. |
-
----
-
-## 3. Scores per crate
-
-Columns: code quality, maintainability, scalability, performance,
-architecture, design philosophy, security, testing. A dash means the
-scale does not apply. Arrows are against 9 October.
-
-| Crate | CQ | Maint | Scale | Perf | Arch | Phil | Sec | Test | Note |
-|---|:-:|:-:|:-:|:-:|:-:|:-:|:-:|:-:|---|
-| concat-core | 8 | 8 | 6 | 7 | 8 | 8 | 7 ↑ | 7 | `MAX_SIDE` and `MAX_TERM` are the one ceiling every caller shares; dead `Project` and `source_duration` remain; `Neg` and `ceil` still unchecked on `i64::MIN`. |
-| concat-project | 8 | 6 ↓ | 6 | 6 | 8 | 7 | 6 | 8 | `Clip::tidy` holds every time field to a hundred hours, with a test; `lib.rs` is 36 lines of re-exports and a 4 258-line test module; `has_non_finite` lists 8 of 11 floats in two places; a speed curve still grows over a neighbour. |
-| concat-media | 7 | 6 | 5 | 7 | 8 | 7 | 7 ↑ | 7 | `pick_rate` refuses a rate the arithmetic cannot hold, with a test; the mux writes a scratch sibling; every `unsafe` argued; a VFR average past 1e6 now falls to the base rate or fails (finding 13). |
-| concat-render | 7 | 5 | 6 | 7 | 8 | 7 | 6 ↑ | 7 | An uncaptured error logs instead of killing the thread; `on_software_adapter`; but a validation error is now a frame nothing marks bad; four docs still promise a CPU fallback; `gpu.rs` 3 392 lines. |
-| concat-effects | 8 | 7 | 6 | 5 | 8 | 8 | 6 | 8 | No commits; the catalogue leaked per reload and the UI-thread trials as they were. |
-| concat-export | 7 ↑ | 4 | 6 | 5 | 7 | 7 | 7 ↑ | 6 | One rename for both paths and one failure policy; `render_on` refuses a size past `MAX_SIDE`; no test for the failure paths; scratch names collide across exporters; `lib.rs` 3 042 lines. |
-| concat-host | 7 | 6 | 5 | 6 | 6 | 6 | 4 | 6 | A test guard and nothing else; the update chain, the cache deletion, the HDR proxy name, `Server::stop` on the UI thread as they were. |
-| concat (Rust) | 6 | 4 | 5 | 6 | 6 ↑ | 6 | 6 | 4 ↑ | `after_dialog` and the export sheet are well shaped, with a table test; the paste's plan is pure and tested; three dialogs missed, a title's copy loses its keys, `studio.rs` +254 lines. |
-| concat (Slint) | 7 | 7 ↑ | 6 | 7 | 7 | 6 | – | – | The tray lost a nested popup; the meters pane is one instrument with named constants; 2 `accessible-*` properties, 143 `TouchArea`s (+1), 15 `FocusScope`s, no keyboard path into the new menu. |
-| concat-api | 8 | 7 | 5 | 5 | 8 | 4 ↓ | 5 | 7 | `MAX_SIDE` is the engine's; the parity promise at `message.rs:23-26` lost two more operations (copy and paste, the overwrite question); `export.run` and `project.create` still overwrite without asking. |
-| concat-server | 7 | 8 | 5 | 6 | 8 | 8 | 5 | 6 | No commits; two things newly read: `stop` can hang for good on a wildcard bind on Windows, and a thread the OS refuses kills the accept loop silently. |
-| concat-cli | 7 | 8 | – | – | 7 | 6 | 6 | 2 | No commits; no logger; `serve` parks with no signal handling; a non-UTF-8 line ends `api`. |
-| concat-speech | 7 | 5 | 6 | 7 | 6 | 7 | 8 | 6 | No commits; the DirectML failure of #291 is a `?` with no fallback and no log of the provider. |
-| concat-vision | 8 | 8 | 5 | 6 | 9 | 8 | 8 | 7 | Zero churn since 23 September. |
-| concat-text | 7 | 6 | 5 | 6 | 8 | 8 | 7 | 6 | No commits; 12 MB copied to the heap at start, PNG strikes decoded per render. |
-| concat-android | 7 | 7 | – | 7 | 7 | 7 | 7 ↑ | 1 | The release key no longer reaches a pull-request build; the three crash paths of the bridge and 0 tests as they were. |
-
----
-
-## 4. Scores per feature
-
-Status: **done**, **partial** (works with a known gap), **thin** (present,
-weak). Rows that changed since the 9th carry a reason; ★ marks a row new
-this week.
-
-| Feature | Where | Status | 9 Oct | Today | Why |
-|---|---|---|:-:|:-:|---|
-| Launcher, new project, recents | `ui/start.slint` | done | 8 | 8 | Browse goes through `after_dialog`; Open still under the borrow (finding 2). |
-| Import, bin, drag and drop | `ui/workspace/media-pane.slint` | done | 8 | 8 | Unchanged; "N selected" still untranslated. |
-| Timeline editing | `ui/timeline/`, `placement.rs`, `studio.rs` | done | 8 | 8 | ★ Copy and paste of a selection, between timelines, as one undo step, with a tested plan (c260c2d); a paste onto a busy lane does not keep its shape (finding 3); a group dragged past 0 still collapses. |
-| ★ Copy and paste | `studio.rs:1903`, `:6774-6964` | partial | – | 6 | The plan is right; `AddClip` lands the whole media before the trim, and a title's copy keeps style, length and height only. |
-| Volume line | `lanes.slint`, `studio.rs:6510` | done | 7 | 7 | Unchanged, untested. |
-| Undo and redo | `editor.rs` | done | 8 | 8 | Every snapshot still clones the bin. |
-| Transitions | `plan.rs`, `concat-export/src/lib.rs:738` | partial | 7 | 7 | The hold still plays keyed effects early. |
-| Effects and filters | `concat-effects/packages/` | done | 8 | 8 | Unchanged. |
-| Colour and HDR | `model.rs`, `gpu.rs`, `encode.rs` | done | 7 | 7 | Unchanged. |
-| Colour grading | `concat.color-wheels`, `grading.slint` | done | 7 | 7 | Unchanged. |
-| Titles and text | `concat-text`, `titles.rs` | done | 8 | 8 | Unchanged. |
-| Shapes / Stickers | `media-pane.slint`, `clips.rs` | partial | 6 | 6 | Still vanish under 0.2.5; a pasted shape loses its transform. |
-| Keyframes, animations | `model.rs` | done | 7 | 7 | Unchanged. |
-| Crop, transform, blend, masks | `plan.rs` | done | 7 | 7 | Unchanged; shape masks of 0.2.4 still absent (#239). |
-| Cutout, smart brush | `segment.rs` | done | 7 | 7 | Unchanged. |
-| Enhance | `enhance.rs` | thin | 6 | 6 | Copies still land in `cache/`. |
-| Auto-captions | `captions.slint`, `samples.rs` | done | 8 | 8 | Unchanged. |
-| Text-to-speech | `tts.rs`, `chatterbox.rs` | done | 7 | 7 | Unchanged; the DirectML failure (#291) is a bare error. |
-| Voiceover | `concat-host/src/record.rs`, `panes/voiceover.rs` | partial | 6 | 6 | A crashed take is still lost (the RIFF-overflow claim of the 9th was wrong: `finish` caps at `u32::MAX − 36`). |
-| VU meters | `concat/src/meters.rs`, `meters-pane.slint` | partial ↑ | 5 | 6 | ★ A ruler, a peak readout, the clip light over its bars, fed from playback's peaks (948b6db); no test. |
-| Templates | `templates.rs` | done | 7 | 7 | Unchanged. |
-| LUT import | `cube.rs` | done | 8 | 8 | Unchanged. |
-| Custom and community packages | `catalogue.rs` | done | 7 | 7 | Unchanged; trials on the UI thread, a catalogue leaked per reload. |
-| Proxies | `proxy.rs` | partial | 5 | 5 | Unchanged. |
-| Export | `export.slint`, `panes/export.rs`, `concat-export` | done ↑ | 7 | 8 | ★ The mux cannot leave a broken file and the output path is touched once (a2dd5e3); ★ the sheet asks before replacing and refuses a name that is not a file's, with a test (9a2e5f2); an uncaptured GPU error is now a wrong frame nobody is told about (finding 8). |
-| Export on a phone | `platform.rs`, `ConcatFiles.java` | partial | 6 | 6 | Unchanged; unseen on a phone. |
-| Hardware encode | `encode.rs` | done | 6 | 6 | Unchanged. |
-| Playback and audio | `playback.rs`, `samples.rs` | done | 7 | 7 | Unchanged. |
-| Monitor stills | `pool.rs`, `concat-export/src/lib.rs:1349` | done | 7 | 7 | Still stretched when a clip has no dimensions; the test still `expect`s a GPU. |
-| Hardware decode | `hardware.rs` | done | 7 | 7 | Unchanged; the fallback chain is complete at open, transfer and send. |
-| Remote API | `concat-api`, `concat-server` | partial | 5 | 5 | Seven window operations with no verb; `stop` blocks the window, and on a wildcard bind on Windows may never return (finding 6). |
-| Self-update | `updates.rs` | done | 5 | 5 | Unchanged. |
-| Logs | `logs.rs` | done | 7 | 7 | Unchanged. |
-| Keyboard menus, platform keys | `app.slint`, `platform.rs` | partial | 6 | 6 | Unchanged; the flattened add-track menu has no keyboard path. |
-| Save Video Frame As, Save Audio As | `studio.rs:5838`, `6438` | done ↑ | 7 | 8 | ★ Both dialogs open on their own turn of the loop now; the extension is still swapped after the prompt. |
-| Phone shell | `ui/phone/`, `concat-android` | partial | 7 | 7 | Unchanged; the category pills' English list still shown there. |
-| Relink missing media | `relink.slint`, `panes/relink.rs` | done | 6 | 6 | The picker is off the borrow; the walk still runs on the UI thread, inside `after_dialog`'s `then`, and follows symlinks with no loop guard. |
-| Themes and accent colours | `ui/theme/` | done | 8 | 8 | Unchanged. |
-| Languages | `concat/locales/`, `fonts.rs` | partial | 7 | 7 | The week's five keys translated everywhere; seven locales still half English by value. |
-| Windows text | `src/patches/i-slint-renderer-skia` | partial | 5 | 5 | Untouched since 3989e59; still unseen on a Windows screen. |
-| Accessibility | the whole UI | thin | 2 | 2 | Unchanged. |
-| Windows data folders | `dirs.rs` | done | 6 | 6 | Unchanged. |
-
----
-
-## 5. The 9 October findings, re-verified
-
-Numbers are the 9 October audit's (its section 6).
-
-| # | Finding | Today | Evidence |
-|---|---|---|---|
-| 1 | A hand-started release builds `main`, not its tag | **Fixed** | `build-app.yml:159-161` and `mobile.yml:72-74, 211-213` check out `inputs.ref`; `release.yml:80-81, 89-91` pass the tag. Every checkout in the three workflows has a `ref`. |
-| 2 | `main` red; the gate blocked by a hanging test | **Unchanged** | The cards test skips on WARP (`cards.rs:598-605`) and the job has a 120-minute limit (`ci.yml:201`), but runs 37964005984 and 38022075165 still die at the step's 90 minutes (`ci.yml:272`). Cause undiagnosed; see section 6, finding 1. |
-| 3 | The Android key reaches pull-request builds | **Fixed** | The build runs under a key made for the run (`mobile.yml:133-141`); the repository's key is decoded in the `sign` step alone (`:158-175`), password through `--ks-pass env:` (`:174`), keystore removed after; `release.yml:91` alone passes `sign: true`. Two unverified edges (finding 19). |
-| 4 | An export with sound can leave a broken file | **Fixed** | `audio::mux` writes a third scratch sibling (`concat-export/src/lib.rs:979, 1009`); one `rename` for both paths (`:1012-1013`); three scratch files removed on every outcome (`:1016-1018`). No test makes the mux fail. |
-| 5 | The CJK face shipped with half its coverage | **Fixed** (on the 9th) | `fonts::tests::the_cjk_face_covers_every_locale`. |
-| 6 | Native dialogs open under the studio borrow | **Partly fixed** | `host::after_dialog` (`concat/src/host.rs:352-368`) carries Export Browse (`panes/export.rs:199`), Start Browse (`start.rs:149`), Relink all (`relink.rs:51`), Save Audio (`studio.rs:5838`) and Save Frame (`:6438`). Three remain under `borrow_mut`: `StartMsg::Open` → `start.rs:220`, `import_lut` → `studio.rs:7557`, and the menu's Import at `lib.rs:1627`, whose `pick_files_async` blocks synchronously on a desk (`platform.rs:682`). The timers still borrow (`studio.rs:2718`, `meters.rs:80`). |
-| 7 | The export sheet overwrites without asking; no name check | **Fixed** | `file_name` (`panes/export.rs:703-710`) trims, refuses empty, dots, control characters and Windows' set, with a table test (`:719`); `exists()` raises `Confirm::ReplaceExport` (`:452-453`), answered as `ExportMsg::Overwrite` (`:217`); the summary shows the trimmed name (`:560-566`). |
-| 8 | A frame past 8192 panics wgpu | **Fixed** | `MAX_SIDE` (`concat-core/src/frame.rs:32`); `render_on` refuses (`concat-export/src/lib.rs:900-908`); `is_sane` and `or` (`model.rs:1962-1970, 1928-1940`); the API's constant is the engine's (`concat-api/src/lib.rs:76`); `on_uncaptured_error` on both devices (`gpu.rs:719-721`; `concat/src/gpu.rs:110-112`), the only two `request_device` sites. Residue: findings 8 and 15. |
-| 9 | A 1e300 clip start panics the export | **Fixed** | `Clip::tidy` clamps `start`, `duration`, `source_start` and both fades to `MAX_TIME` (`model.rs:1479-1484`); `pick_rate` takes only a rate `FrameRate::checked` accepts (`probe.rs:228-231`); tests at `concat-project/src/lib.rs:937-969` and `probe.rs:286-298`. `transition_in.duration` keeps a floor only (`:1510-1512`). |
-| 10 | A SAFETY comment on `set_var` is not true | Unchanged | `platform.rs:347-350`. |
-| 11 | A group dragged past 0 collapses | Unchanged | `studio.rs:4254`; `placement.rs:49`. No test. |
-| 12 | A speed curve grows over a neighbour | Unchanged | `commands/properties.rs:285-308`; and the paste re-applies it (`studio.rs:6922`). |
-| 13 | Trials on the window thread; a catalogue leaked per reload | Unchanged | `catalogue.rs:978`; `shader.rs:1139` still counts a `switch` break. |
-| 14 | "Relink all" walks on the UI thread | Unchanged | The picker moved to `after_dialog` (`relink.rs:51`); the walk (`:85-91`) runs in its `then`, on the UI thread, LIFO, following symlinks; toast English (`:140`). |
-| 15 | User-visible strings in Rust | Unchanged | `studio.rs:9588, 9992-9994, 3563, 6645, 7205-7206`; `panes/export.rs:314, 568, 574, 654`; `relink.rs:140`; the Slint set as it was. |
-| 16 | `Server::stop` blocks the window; jobs and workers die silently | Unchanged | `concat-server/src/lib.rs:293-296`; `hub.rs:107`; `concat-api/src/lib.rs:741-745, 761-786`; `playback.rs:675-720`; `settings.rs:558`. |
-| 17 | API reads unconfined; caller strings to FFmpeg's opener | Unchanged | `concat-api/src/lib.rs:326-328, 466, 486-490`; `json.rs:305-311, 331-338`. |
-| 18 | One export path drops failed frames silently | **Fixed**, with a residue | All three failure points are `?` (`concat-export/src/lib.rs:1372, 1427, 1435`). A time past the media's end still differs: the pool answers the last frame (`pool.rs:429-432, 1001-1008`), the paced path's `None` contributes nothing (`:1432-1435`). The still-size fallback is unchanged (`:1349-1352, 1844-1847`). No test for the new `?`. |
-| 19 | Pinned prefetch frames outside any budget | Unchanged | `prefetch.rs:181, 292-295, 321`. |
-| 20 | A crashed voiceover take is lost | Unchanged | `record.rs:279, 312-319`; a `write` error at `:159` returns and never finishes the header. |
-| 21 | Unverified downloads; mutable action tags | **Worse** | FFmpeg from BtbN `latest` (`ci.yml:31, 203`; `build-app.yml:62, 197, 320`), ONNX Runtime (`ci.yml:83-84`), sherpa (`ci.yml:103, 247`; `sherpa-mobile.sh:38, 56`), mobile FFmpeg (`ffmpeg-mobile.sh:38`), all without digests; every action by tag; and now `dtolnay/rust-toolchain@master` in six jobs (finding 4). `release.yml:48` still pastes `inputs.tag` into a script. |
-| 22 | Three Android bridge crash paths | Unchanged | `ConcatFiles.java:277, 200-214, 115, 132`; `concat-android/src/lib.rs:169, 256-282`. |
-| 23 | Doc drift | Mostly unchanged | gpu.rs's four CPU-fallback docs (`:623-625, 1183-1184, 1521-1522, 3179-3182`); `studio.rs:2126-2137` three docs on `forget_art`; `pool.rs:64-68`; `catalogue.rs:946-947`. `ARCHITECTURE.md` is rewritten today. |
-| 24 | Tests that prove the wrong thing or never run | Partly | New: the parity harness and the cards test skip on a software adapter unless `CONCAT_REQUIRE_GPU` (`gpu/tests.rs:21-30`, `cards.rs:598-605`). Still: the emoji test returns where no colour face exists (`concat-text/src/lib.rs:2595-2596`); `a_still_keeps_its_shape_on_the_monitor` `expect`s a GPU (`concat-export/src/lib.rs:2478`); device loss tested by calling `recovered()` (`:2574-2580`); `platform.rs:760-786` does not compile for Android. |
-| 25 | The Rational edges and other smalls | Unchanged, one correction | `time.rs:114-121, 159-161, 252-259`; `commands/clips.rs:49-61`; `proxy.rs:66, 135-149`. The RIFF-overflow claim was wrong: `finish` caps at `u32::MAX − 36` (`record.rs:315`). |
-| 26 | HDR side data sized by a Rust mirror | Unchanged | `encode.rs:536, 1077, 1285`; mirrors at `:752, 799`. |
-| 27 | Realtime audio does work it should not | Unchanged | `playback.rs:1107, 168-175, 521-527, 1195`; `record.rs:222-227, 242-259`. |
-| 28 | Blocking work on the UI thread | Unchanged | `studio.rs:7204, 7086, 6129, 6154, 6663, 6670`; `concat-text/src/lib.rs:323-325, 1058`. |
-| 29 | Hygiene | Unchanged | `wayland_drop.rs`; `i18n.rs:331-337, 90-99`; `studio.rs:7312-7322`; `lib.rs:23`, `platform.rs:156`; `concat-server/src/lib.rs:195`; `concat-cli/src/main.rs:82-85, 228-232, 247-249`; `concat-api/src/lib.rs:380-397`; `catalogue.rs:1047-1072`; `concat-android/src/lib.rs:139`. |
-
-Of the 29: 8 fixed, 2 partly, 18 unchanged, 1 worse. The items the 9th
-carried from the 4th and the 28th are as they were, with three
-exceptions: S11 (no ceiling on rate or size) is now **fixed** on every
-side (`FrameRate::checked`, `pick_rate`, `MAX_SIDE`, `MAX_TIME`); the 4th's
-#15 (the CHANGELOG) is **worse** (45 non-merge commits since 0.2.6 with no
-section, `CHANGELOG.md:10`); and the 4th's #19 (the API behind the
-window) is **worse** again (copy and paste and the overwrite question
-have no verb; `message.rs:51-237` lists 21). The rest: #4 locales by value
-(de 199, fr 204, ko 137, pt-BR 175, it 175, es 174, tr 155), #5 shapes,
-#7 the hold, #10 and #11 partly, #12, #13, #17 README's MCP, #18 the
-CLI's logger, #20, #22, #23, #24, #25, #26, #27, #28, S1 to S4, S6, S8,
-S9 partly, S10, S12, S13, S14 partly, S16, S17, S18: unchanged, at the
-lines the 9th gave.
-
----
-
-## 6. New findings, by severity
-
-Numbers are this audit's. Each names where to look.
-
-### High
-
-**1. `main` is still red after the fix meant to free it, and nobody can
-say why.** Both push runs since 3f3b7de time out in `cargo test
---workspace -j 1` at the step's 90 minutes (`ci.yml:272`; runs
-37964005984, 38022075198; annotation "The action … has timed out after 90
-minutes"). The cards test should now skip on WARP (`cards.rs:598-605`),
-so another test hangs; the step's log is readable by administrators only,
-and the next candidate from the job's own comments is the GPU parity suite
-on WARP (`ci.yml:182-183`). With the step timeout firing first, the job's
-new 120-minute limit (`:201`) does nothing. The release gate reuses the
-job (`release.yml:67-72`): no tag can ship. Make the next failure legible
-before guessing again: run the suite under a per-test timeout
-(`cargo nextest` with `slow-timeout = { period = "60s", terminate-after =
-5 }`, or `--test-threads=1 --nocapture` so the log names the last test
-started), and have the Windows job skip the GPU suites on a software
-adapter the way the cards test does. (verified against the runs; the
-cause is not)
-
-### Medium
-
-**2. Three dialogs still open while the studio is mutably borrowed.**
-9a2e5f2's message says every dialog goes through `after_dialog`; five of
-eight do. `StartMsg::Open` calls `pick_folder` inside its handler
-(`panes/start.rs:220`, reached from `on_window!` at `lib.rs:512-514` and
-the menu at `:1616-1625`); `import_lut` calls `pick_files` (`studio.rs:
-7557`, under `on_window!` at `lib.rs:748-750`); and the menu's Import
-(`lib.rs:1627`) calls `pick_files_async`, which on a desk blocks
-synchronously (`platform.rs:682`) while `state` is borrowed. The transport
-and meter timers still `borrow_mut` (`studio.rs:2718`, `meters.rs:80`), so
-File › Open or Import while playing is the panic path that finding 6 of
-the 9th described, on macOS. (verified)
-
-**3. A paste does not keep its shape on a busy lane, and a pasted title
-keeps a quarter of what it had.** `place_copy` adds the clip with
-`AddClip`, which lands the **whole media** (`concat-project/src/commands/
-clips.rs:34-38`, `clip_length` `:739-743`) at `nearest_free_start`
-(`placement.rs:42-66`), and only then trims it to the source's footage
-(`studio.rs:6885-6912`). A two-second clip from a sixty-second file needs
-a sixty-second gap on the target row; without one it lands elsewhere, and
-`head = start − placed.start` (`:6895`) trims from the wrong anchor, so
-the copy sits in the wrong place with the wrong frames. `paste_plan`
-never sees the existing clips, so its test cannot catch this. Duplicate
-had the same bug; a multi-clip paste makes it the normal case. And a text
-or shape clip returns early from `place_copy` (`:6852-6878`) with style,
-duration and `offset_y` only: the keys, transform, opacity, fades, blend,
-filters, effects and transition at `:6916-6964` are reached by footage
-alone, against the commit message. Also: the clipboard is never cleared
-(`:771`), so after switching projects a paste toasts `MediaGone` once per
-clip; a copied speed curve re-applies `SetClipSpeedCurve` (`:6922`,
-finding 12 of the 9th); `name`, `cutout`, `audio_stream`, `transition_in`,
-`enhanced` and `extra` are not carried for footage either. Place the clip
-with its trimmed length (an `AddClipAtFirstFree` that takes the footage,
-or trim before placing), and carry the full patch for titles. (verified)
-
-**4. The toolchain action now runs from a moving branch.** 3f3b7de
-replaced `dtolnay/rust-toolchain@stable` with `@master` in six jobs
-(`ci.yml:41-46, 157-162, 208-214`; `build-app.yml:163-170`; `mobile.yml:
-81-88, 219-226`) so that the channel could be read from the file. It is
-the one action in the repository pinned to a branch, and it runs in every
-build including the release's. The same action takes a SHA, and
-`@stable` with the file's channel passed as `toolchain:` worked before.
-(verified)
-
-**5. The CHANGELOG has nothing after 0.2.6.** `CHANGELOG.md:10` is
-`## 0.2.6 — 2026-10-04`; the 45 non-merge commits since the tag have no
-line: voiceover, VU meters and their ruler, colour emoji, Android
-publishing, base fonts and the font picker, click-to-preview, Detach,
-the CJK face, the Windows text patch, copy and paste across timelines,
-the export sheet's confirmation and name check, the atomic export, the
-release and signing fixes, the add-track menu. `release-notes.sh:43-50`
-would fall back to commit subjects for 0.2.7. This recurred once already
-(the 4th's #15). (verified)
-
-**6. `Server::stop` can hang the window for good on a wildcard bind on
-Windows.** `shut` wakes the blocked `accept` by connecting to
-`listener.local_addr()` (`concat-server/src/lib.rs:269-272`); for
-`0.0.0.0:port` or `[::]:port` Windows refuses that connect
-(`WSAEADDRNOTAVAIL`), so `thread.join()` (`:281-283`) never returns. The
-Remote page accepts such addresses (`settings.rs:983, 996` test them) and
-calls `stop()` on the UI thread (`:558`); the window's `Drop` reaches it
-on quit (`lib.rs:300-306`). Linux routes 0.0.0.0 to loopback, so it shows
-on Windows only. Wake the accept by connecting to loopback on the bound
-port, or close the listener handle. (reasoned, not reproduced)
-
-**7. A thread the OS refuses kills the accept loop with no word.**
-`std::thread::spawn` per connection (`json.rs:128, 183`) panics when a
-thread cannot be made; the accept loop dies, the server stops seating
-callers, nothing is logged and `stop` still works, so the symptom is a
-Remote page that says On and a port that never answers. `Builder::spawn`
-and refuse the caller. (verified)
-
-**8. An uncaptured GPU error is now a silently wrong export frame.** The
-handler logs (`gpu.rs:719-721`), `dead` is set on readback failure only
-(`:3263-3268`), and the export writes whatever came back (`concat-export/
-src/lib.rs:1490`). Before a2dd5e3 that was a panic, so this is a trade,
-but it is one with no signal to the person exporting. Count the errors in
-the handler and have `render_on` report a frame drawn under one. (verified)
-
-**9. `studio.rs` keeps growing.** 10 689 lines; `impl Studio` from
-`:1962` to `:10332` is 8 371 lines and 215 methods on a struct of 93
-fields (`:709-957`); `run()` in `lib.rs:133-1860` is 1 727 lines. +254
-lines in two days, the paste among them, when the paste is exactly the
-kind of pure placement that belongs in `concat-project` beside
-`placement.rs` and would be testable there against existing clips. The
-panes moved out one at a time (eleven so far); the gestures, the clipboard
-and the stage have not. (verified)
-
-**10. The API still writes where the window now asks.** `export.run`
-overwrites an existing file without a question (`concat-api/src/lib.rs:
-519-520`) and the export's rename makes the replacement total;
-`project.create` writes any size (`:380-397`) and the document is opened at
-the fallback size later (finding 16) rather than refused at the door. A
-script with the token destroys a previous export silently. One `exists`
-check and one `is_sane` at the verb. (verified)
-
-### Low
-
-**11. Scratch-name collision across exporters.** The mux's scratch names
-derive from the stem alone (`concat-export/src/lib.rs:976-979`): `a.mp4`
-and `a.mov` in one folder, or the window's and the embedded API's
-exporters (separate `SingleFlight` slots, `concat/src/host.rs:82`,
-`concat-api/src/lib.rs:181`), share scratch paths and would truncate or
-delete each other's. A random suffix.
-
-**12. The past-the-end policy still differs between the export's two
-decode paths.** `concat-export/src/lib.rs:1432-1435` against `pool.rs:
-1001-1008`: a clip trimmed past its file freezes on the last frame when
-reversed or retimed and goes black at normal speed.
-
-**13. `pick_rate` and variable-rate files.** `MAX_TERM`'s doc says terms
-past it come from hand-edited documents (`time.rs:341-344`), but FFmpeg
-reduces `avg_frame_rate` to terms up to `INT_MAX`; a VFR file whose
-average reduces past 1e6 now takes `r_frame_rate` (`probe.rs:212-215`),
-and with both out the import fails "no usable frame rate" (`:174-178`)
-where it opened before. No VFR fixture exists. (reasoned)
-
-**14. Five things the ceilings left.** `transition_in.duration` keeps a
-floor only (`model.rs:1510-1512`); `VideoSettings::or` falls back a side
-past `MAX_SIDE` with no notice to the user (`:1930-1939`) where retired
-fields get one (`editor.rs:107`); `frame.rs:29-30` and `concat/src/gpu.rs:
-87-92` call 8 192 "the downlevel limit" (it is wgpu's default; downlevel
-is 2 048), and the window's device asks for `adapter.limits()` (`:94-96`)
-while the compositor's own takes the default (`gpu.rs:703`), so the two
-devices can hold different maximums; `preview_sources` has no check of
-its own and relies on its callers.
-
-**15. The server's quieter edges.** gRPC's `events()` keeps a stream open
-after the Hub dropped it as slow (`grpc.rs:74-76`, `hub.rs:58-66`), so a
-client whose channel filled once never learns its export ended; a
-poisoned subscriber mutex silences events for good with no log (`hub.rs:
-54-57`); the CLI's `api` ends the process on a non-UTF-8 stdin line
-(`main.rs:172-173`); the `key_of` doc is attached to `impl Drop for Api`
-(`concat-api/src/lib.rs:868-879`).
-
-**16. The cards skip and the engine job.** The no-adapter branch of the
-cards test skips even under `CONCAT_REQUIRE_GPU` (`cards.rs:598-601`),
-against the CI comment (`ci.yml:127-128`) and unlike concat-render's
-harness, which asserts (`gpu/tests.rs:27-28`); and the `engine` job, which
-now alone runs the cards test for real, has no `timeout-minutes`
-(`ci.yml:27-29`).
-
-**17. The signing step's two unverified edges.** No `--ks-key-alias`
-(`mobile.yml:158-175`), so a keystore with two entries fails; a key
-password that differs from the store password makes apksigner prompt on
-stdin and the step hang. A release with the real key has not run since
-the change. And `WINDOWS_CERTIFICATE_PFX`/`_PASSWORD` are used at
-`build-app.yml:710-711` but not declared in its `secrets:` block
-(`:43-57`), so any caller but `release.yml`'s `secrets: inherit` builds
-unsigned without saying so (the 9th's #28, restated).
-
-**18. The window's smalls.** `after_dialog` skips the `refresh_art()`
-that `handler!` runs (`lib.rs:454` against `host.rs:365-368`); the
-clipboard outlives the project (`studio.rs:771`, not cleared in
-`close_project` `:7213`); `dock.rs:19-20`'s doc arithmetic gives 74 for a
-constant of 76; the meters pane's "dB", "L" and "R" are literals
-(`meters-pane.slint:198, 206, 214`); the new tray menu has no keyboard
-path.
-
-**19. concat-project's test suite is one file.** `lib.rs` is 36 lines of
-re-exports and a 4 258-line `mod tests` holding all 114 of the crate's
-tests (`lib.rs:36-37`); `commands/`, `editor.rs`, `doc.rs` and `model.rs`
-carry none of their own.
-
-**20. `scripts/locales.py` crashes on a Windows console.** `:150` prints
-each locale's name in its own script; under cp1252 that is a
-`UnicodeEncodeError` at "Español" and the check reports nothing. CI's
-Ubuntu is UTF-8, so only a Windows contributor sees it; `python -X utf8`
-works, and `sys.stdout.reconfigure(encoding="utf-8")` at the top would
-make it work everywhere.
-
-**21. Doc drift, continued.** Five comments describe a CPU compositor
-fallback that no longer exists (`concat-export/src/lib.rs:19`,
-`concat-host/Cargo.toml:37`, `concat/Cargo.toml:124`, `gpu.rs:1522, 3181`),
-beside the four in gpu.rs the 9th listed; `src/Cargo.toml:32` says wgpu is
-pinned to what "Slint 1.17" shares (the lock has 1.18.1); `src/README.md`'s
-crate table omits concat-text, concat-vision and concat-perf; `tts.rs:7`
-still says whisper is a child process; `concat-api/src/lib.rs:7` still
-names an MCP server. `ARCHITECTURE.md` is rewritten today against the
-tree and should be read with this audit.
-
-**22. Dead code in core.** `Clip::source_duration` (`timeline.rs:
-314-321`) has no caller outside its own tests, beside the dead `Project`
-(`:499-514`), both re-exported at `core/lib.rs:31`.
-
----
-
-## 7. What is good, and should stay that way
-
-- **The 9 October audit's first three steps were taken the same day,
- and each closed what it named.** A release builds its tag
- (`build-app.yml:159-161`); the Android key is decoded in the step that
- signs and removed after (`mobile.yml:158-175`); the mux writes a
- sibling and the output path is touched once by a rename, whichever way
- the file was made (`concat-export/src/lib.rs:979-1018`).
-- **One ceiling, shared.** `concat_core::frame::MAX_SIDE` is what the
- document's `is_sane`, the export's `render_on` and the API's
- `checked_size` all read (`frame.rs:32`; `model.rs:1962`; `concat-export/
- src/lib.rs:900`; `concat-api/src/lib.rs:76`), and `MAX_TIME` holds every
- clip time field (`model.rs:1479-1484`); a hand-edited document opens at
- a size the GPU can hold instead of killing its thread.
-- **`host::after_dialog` is the right seam.** A native dialog runs on its
- own turn of the loop with nothing borrowed, and its answer is a message
- to the pane like a worker's result (`concat/src/host.rs:352-368`;
- `panes/export.rs:199`, `start.rs:149`, `relink.rs:51`). The three sites
- it missed are three calls.
-- **The export sheet's name check is a table test** (`panes/export.rs:
- 703-719`), and the question before replacing goes through the confirm
- sheet that already existed, answered as `ExportMsg::Overwrite`.
-- **`paste_plan` is pure and tested** (`studio.rs:1903-1928, 10387`):
- rows matched by number, a locked row passed over for the nearest open
- one, a whole paste one undo step through the editor's gesture folding
- (`concat-project/src/editor.rs:202-209`). What is wrong with the paste
- (finding 3) is around the plan, not in it.
-- **The cards test's skip reads the adapter, not an environment
- variable** (`gpu.rs:737-739`), and the Linux job still runs it on
- lavapipe under `CONCAT_REQUIRE_GPU` (`ci.yml:133`).
-- **The toolchain is one fact in one place.** Every job reads
- `src/rust-toolchain.toml` (`ci.yml:41-46`); only the action that reads
- it is pinned wrong (finding 4).
-- **The 69 `unsafe` sites are unchanged and every one carries a SAFETY
- line**; concat-media's 38 (decode 3, encode 10, ffi 8, hardware 15,
- probe 1, treat 1) each argue their contract within four lines.
-- **The hardware-decode fallback is complete** at open, transfer and
- send (`decode.rs:644, 835, 848, 878`), and a failed device is remembered
- per kind (`hardware.rs:343-372`).
-- **The tray's menu is flat now** (`tray.slint:97-150`, −37 lines): the
- one nested popup in `ui/` is gone and no other nests one.
-- **The week's five locale keys are translated in all fourteen files**,
- and `scripts/locales.py --check` holds the key sets (once it can print).
-
----
-
-## 8. Where we stand
-
-The 9 October audit ended with nine recommended steps. Two days later:
-
-| Step (9 Oct) | Landed |
-|---|---|
-| 1. `ref:` the tag; sign Android on release runs; skip or time out the cards test; toolchain into CI; tag 0.2.7 from a green `main` | the first four (3f3b7de); `main` still red, no tag |
-| 2. Mux into scratch and rename; one failure policy; a size ceiling with an uncaptured-error handler; bound clip times and probed rates | all four (a2dd5e3) |
-| 3. The export sheet: refuse `/` and empty names, ask before overwriting, show the trimmed path; dialogs on the next loop turn | the sheet in full; five of eight dialogs (9a2e5f2) |
-| 4. Trials, relink, cache clear and sweep off the UI thread | nothing |
-| 5. `Server::stop` cancels first; `catch_unwind` on jobs and workers; a protocol whitelist; a write timeout | nothing |
-| 6. Pinned-frame budget; WAV placeholder header; group clamp; speed-curve ripple | nothing |
-| 7. The literals through `I18n`; a locale check that flags English | nothing (the week's new keys were done right) |
-| 8. Digests for downloads; actions by SHA; `inputs.tag` through `env:` | nothing, and one action moved to a branch |
-| 9. Windows text and the phone's export on real hardware | nothing seen |
-
-Three of nine steps in two days, taken up well: every fix landed with a
-test that fails without it, in one place, with the reason in the message.
-What the two days also show is the week's pattern again: a fix that
-claims a little more than it does (every dialog; a copy that carries
-everything; the Windows suite under an hour) and is caught by the next
-reading. The gate is the thing. `main` has not been green since 4
-October, the fix for the hang did not end the hang, and until the Windows
-job finishes nothing can be tagged and nothing the CI says about Windows
-is known. Everything else on this list waits behind an afternoon with a
-per-test timeout and a readable log.
-
-**Recommended order.** Each line is an hour to a few days, and each
-closes a finding above.
-
-1. Make the Windows job name its hang: `cargo nextest` with a per-test
- timeout, or `--test-threads=1 --nocapture`; skip the GPU parity suite
- on a software adapter as the cards test does; give the `engine` job a
- timeout too. Pin `dtolnay/rust-toolchain` to a SHA. Then tag 0.2.7
- from a green `main`, with a CHANGELOG section written first. (1, 4, 5,
- 16)
-2. The three remaining dialogs through `after_dialog`; the relink walk,
- the cache clear and the sweep on workers; trials off the UI thread
- with one catalogue per reload. (2, and the 9th's 13, 14, 28)
-3. The paste: place the clip at its trimmed length, carry the full patch
- for titles, clear the clipboard with the project; and move the
- placement half into `concat-project` where existing clips can be
- tested against. (3, 9)
-4. `Server::stop` wakes the accept on loopback and cancels jobs first;
- `Builder::spawn` for connections; `catch_unwind` on API jobs and
- decode workers; a protocol whitelist in front of FFmpeg's opener; a
- write timeout on the outbox; an `exists` check and `is_sane` on the API's
- two writing verbs. (6, 7, 10, and the 9th's 16, 17)
-5. Count uncaptured errors and report a frame drawn under one; a random
- suffix on the mux's scratch names; one past-the-end rule for both
- decode paths; a VFR fixture. (8, 11, 12, 13)
-6. Digests for FFmpeg, sherpa and ONNX downloads; actions pinned by SHA;
- `inputs.tag` through `env:`; declare the Windows signing secrets; run
- a release with the real Android key against a staging tag. (17, and
- the 9th's 21)
-7. The pinned-frame budget; the WAV placeholder header; the group clamp;
- ripple in `SetClipSpeedCurve`; `DOCUMENT_VERSION` for shapes. (the
- 9th's 19, 20, 11, 12, and the 4th's #4)
-8. The literals through `I18n`; a `locales.py` that flags English values
- and prints on any console. (20, and the 9th's 15)
-9. The 28 September list in its order: sign the update, the Remote
- page's root and fds, processed media out of `cache/`. Then Windows
- text and the phone's export on real hardware.
-
-After those, phase 4 of the HDR plan (a true HDR preview), then phase 5,
-in the plan's order.
-
----
-
-## 9. Caveats
-
-Scores are judgements over code, calibrated against the 9 October scores
-but not measured against users or another editor. Every crate was read in
-full by one of four reviewers; the headline findings were re-read by hand.
-Findings marked (verified) were read at the line and, where a command
-could show it, run; (reasoned) means the path was followed through the
-code and not executed. The checks ran on a Windows machine with a
-discrete GPU, which is not the runner's WARP, so the CI hang could not be
-reproduced here and was read from the workflow runs' annotations only.
-The app was not launched; no phone build was made; Windows text at 100 %
-and the phone's export remain unseen on the hardware they are for.
-`cargo audit` was not run.
diff --git a/audit/AUDIT-2026-10-12-QUICK.md b/audit/AUDIT-2026-10-12-QUICK.md
deleted file mode 100644
index c83f2adf..00000000
--- a/audit/AUDIT-2026-10-12-QUICK.md
+++ /dev/null
@@ -1,208 +0,0 @@
-# Concat audit, 12 Oct 2026 — the quick version
-
-Same audit as `AUDIT-2026-10-12.md`, one screen. Each bar has 10 slots:
-🟢 filled = good, ⚪ empty = missing. Severity: 🔴 critical · 🟠 high · 🟡 medium · ⚪ low.
-Arrows are against the 11 Oct audit. Every crate was read by one of four reviewers; the headline findings were re-read by hand.
-
-## 🎯 Verdict
-
-**`main` is green; 0.2.7 is not ready to tag.** In the 39 commits since the 11 Oct audit, its top step landed in full:
-- **CI names its own hangs.** Per-test timeouts showed which suites hung on Windows, and those suites now skip on the software GPU.
-- **Downloads and actions are pinned.** Every desktop download is pinned by digest and every action by SHA.
-- **Release bundles must start.** Every desktop bundle is launched where it was built before the release can use it.
-- **The window is fixed:** the last three dialogs are off the studio borrow, and ten tracker bugs are fixed, each with a test that fails without its fix.
-- **25 languages:** each has a font that covers its script.
-
-**But no release run has passed:**
-- **rc.1:** all six desktop builds failed.
-- **rc.2:** Linux arm64 crashed on exit. That is fixed in `c06560e`, but no release run has tried the fix yet.
-- **The phones' new launch checks have never run.**
-- **The final 0.2.7 notes would be three lines long.**
-
-Windows is green because the tests that hit its slow shader compile no longer run there. On an iPhone, Import does nothing.
-
-Count: 🟠 ×4, 🟡 ×17 new, no 🔴.
-
-## 📏 Size
-
-| | Lines | Δ vs 11 Oct |
-|---|---:|---:|
-| Rust, all crates | 90 362 | +915 (+2 007 like for like; concat-perf removed) |
-| Slint UI | 29 901 | +111 |
-| WGSL, packages | 1 764 | 0 |
-| Tests (`#[test]`) | 748 | +31 like for like |
-| Largest file `concat/src/studio.rs` | 10 866 | +177 |
-
-Checks today:
-- fmt clean, clippy 0 warnings.
-- **745 tests passed, 0 failed, 1 ignored**, on Metal.
-- 25 locales, 0 missing.
-- 23 of 23 models have their digests.
-- CI on `main` green, 6 pushes running.
-
-## 📊 The whole app
-
-| Aspect | Bar | 11 Oct | Today |
-|---|---|:-:|:-:|
-| 🧹 Code quality | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 🏛️ Architecture | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 |
-| 🧭 Design philosophy | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 🔧 Maintainability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 |
-| 😊 User likeability | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7 | 7 |
-| ⚡ Performance | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 📈 Scalability | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 |
-| 🔐 Security | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5 | 6 ↑ |
-| 🧪 Testing | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 |
-| 📚 Docs | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 |
-| 🤖 CI and release | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 4 | 6 ↑ |
-
-- **Security up one:** downloads pinned by digest, actions by SHA, the release tag passed through `env:`, Windows signing by thumbprint.
-- **CI up two:** a green gate, hangs that name themselves, and bundles that must start. It is held at 6 because:
- - two release candidates have been spent;
- - phones are never built on a push to main;
- - the bundle workflow can only be run by pushing a tag.
-
-## 🧱 Per crate (average of its scales)
-
-| Crate | Lines | Δ | Bar | Avg | One thing |
-|---|---:|---:|---|:-:|---|
-| ★ locales + `locales.py` | – | – | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7.7 | 🟢 25 languages, a glyph test per locale · 🟡 a missing key does not fail CI |
-| concat-core | 2 897 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.4 | 🟢 untouched · ⚪ `MAX_SIDE`'s doc now wrong; two dead items |
-| concat-vision | 2 760 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 7.3 ↓ | 🟢 untouched · 🟡 CPU-only on shipped Windows; the Intel Mac's runtime may refuse it |
-| concat-effects | 7 256 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.9 ↓ | 🟢 123 packages pass on Metal · 🟠 none compiled on D3D12 in CI any more |
-| concat-project | 10 285 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.9 | ⚪ no commits; `AddClip` places the whole media; tests in one file |
-| concat-media | 10 066 | +212 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.8 | 🟢 nothing vendor-specific opened at launch · 🟡 an HEVC export still opens NVENC first (#309 moved) |
-| concat-render | 7 723 | +273 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.8 ↑ | 🟢 bounded waits, error scopes, phone limits · 🟡 "30 s" is 2.15 s on GLES; `combine` unscoped |
-| concat-server | 1 393 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.6 | 🟡 `stop` on `0.0.0.0` on Windows, as before |
-| concat-text | 2 607 | 0 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.6 | ⚪ no commits |
-| concat-speech | 3 307 | +198 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.5 | 🟢 a typed error and one CPU retry, tested · 🟡 Windows ships no DirectML, the crate still asks for it |
-| concat (Slint) | 29 901 | +111 | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 6.5 ↓ | 🟢 a dropdown scrolls under a finger · 🟡 three primitives still break their bindings; the captions sheet re-broke |
-| concat-export | 4 039 | +168 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.3 ↑ | 🟢 device loss tested end to end · 🟡 a validation error reads as a lost GPU |
-| concat-api | 2 352 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.1 | 🟡 captions from a clip has no verb either; the `gpu` capability can never appear |
-| concat-android | 398 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.1 | ⚪ 0 tests; the emulator check is uncommitted |
-| concat-cli | 514 | 0 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6.0 | ⚪ no logger, no signals |
-| concat-host | 10 842 | +202 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.8 | 🟢 titles painted once; iPhone directories · 🟡 24 held stills undo it mid-drag; saves share a scratch file |
-| concat (Rust) | 23 923 | +954 | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 5.6 ↑ | 🟢 every dialog through one seam; 14 tests on pure plans · 🟠 iPhone Import does nothing |
-
-## 🎛️ Per feature (rows that moved, and the weakest)
-
-| Feature | Bar | 11 Oct | Today | State |
-|---|---|:-:|:-:|---|
-| Languages | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 7 | 8 ↑ | ✅ 25, with faces and READMEs; no right-to-left layout |
-| Export | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ⚠️ a stuck GPU ends the export now; a Windows blur or glow can still sit for minutes, uncancellable |
-| Timeline editing | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ✅ a click no longer moves a clip; Detach says why |
-| Titles and text | 🟢🟢🟢🟢🟢🟢🟢🟢⚪⚪ | 8 | 8 | ⚠️ painted once; past 24, they drop out of the monitor during a drag |
-| Auto-captions | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | 8 | 7 ↓ | ⚠️ ★ on the clip's menu; the sheet can transcribe a clip it is not showing |
-| ★ Command-line launch | 🟢🟢🟢🟢🟢🟢🟢⚪⚪⚪ | – | 7 | ✅ `--project`, `--version`, `--smoke`; any other argument refuses to start |
-| Phone shell | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 7 | 6 ↓ | ❌ iPhone Import does nothing; a failed start vanishes |
-| Cutout | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 7 | 6 ↓ | ⚠️ CPU-only on Windows; at risk on Intel Mac |
-| Export on a phone | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ phone limits right; #286 fix uncommitted; unseen on a phone |
-| Hardware encode | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | 6 | 6 | ⚠️ #309 moved from launch to the first HEVC export |
-| ★ Bundle launch checks | 🟢🟢🟢🟢🟢🟢⚪⚪⚪⚪ | – | 6 | ⚠️ well built; no target through every check yet |
-| Enhance | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 6 | 5 ↓ | ⚠️ CPU on Windows; copies in `cache/` |
-| Remote API | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ |
-| Self-update | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ no signature check |
-| Windows text | 🟢🟢🟢🟢🟢⚪⚪⚪⚪⚪ | 5 | 5 | ⚠️ unseen on Windows |
-| Accessibility | 🟢🟢⚪⚪⚪⚪⚪⚪⚪⚪ | 2 | 2 | ❌ |
-
-Every other feature is unchanged; the full table is in the audit, section 4.
-
-## 🚨 Findings
-
-🟠 **High**
-1. **The final 0.2.7 notes would be three lines long.** `release.yml:170` takes `v0.2.7-rc.2` as the previous tag, and the CHANGELOG is gone. Adding `--exclude 'v*-*'` gives `v0.2.6`.
-2. **On an iPhone, Import does nothing and says nothing.** Only Android installs a file picker (`platform.rs:689-691`).
-3. **Phones are never built on a push to `main`.** The new phone checks have never run, and once committed they gate `publish`.
-4. **A Windows blur or glow export can sit for minutes compiling shaders (#297).**
- - Nothing bounds or cancels the compile.
- - The shaders are compiled again on every export.
- - CI no longer runs the suites that showed it; they skip and report PASS.
-
-🟡 **Medium**
-- **Titles and saves:**
- - Past 24 titles, the early ones drop out of the monitor during a drag, and every title is re-treated each frame (5).
- - Two saves at once share one scratch file (6).
-- **Phones:**
- - On GLES the "30 s" device wait is about 2.15 s, so a heavy phone frame fails the export (7).
- - A phone that cannot start just vanishes; an iPhone's Documents folder is hidden from the Files app (12).
-- **The captions sheet** can transcribe a clip other than the one it shows (8).
-- **GPU errors and encoders:**
- - An HEVC or AV1 export still opens NVENC first, so #309 moved from launch to export (9).
- - Any validation error reads as "lost the GPU 3 times" (10).
- - Transitions' `combine` is unscoped and a refused effect is silently missing from the export (11).
-- **Locales:** `locales.py --check` passes with keys missing (13).
-- **Release notes:** they carry CI-internal lines and the DXC pair, one of which reverts the other (14).
-- **ONNX Runtime:**
- - Windows ships the CPU ONNX Runtime while speech asks for, logs and retries DirectML (15).
- - On an Intel Mac, ONNX Runtime 1.23.2 is likely below ort's API (16).
-- **Launch checks:**
- - `--smoke` stops before the window exists (17).
- - The instruction-set limits are still provisional, and the emulated old CPU runs `--version` only (18).
-- **Release tooling:**
- - The bundle workflow can only be run by pushing a tag (19).
- - `CODE_SIGNING.md` contradicts the workflow (20).
- - Dependabot's cargo group is red by design over sherpa (21).
-
-⚪ **Low**
-- **Phone and launcher:**
- - The 4 px drag threshold is below a finger's slop (22).
- - #151 comes back after Browse (23).
-- **Timeline and launch:**
- - Detach unmutes a video the user muted on purpose (24).
- - Any unknown argument refuses to start, so Linux "Open with" through `%U` shows an error box (25).
-- **Process and CI:**
- - Ten "Fixes #N" commits; the push closed #299, which was reopened by hand (26).
- - The concurrency group cancels pending runs on `main`, and 7 pushes have no CI (28).
- - `ubuntu-latest` becomes Ubuntu 26 on 19 Oct (29).
-- **Rendering:**
- - Right-to-left languages are laid out left to right (27).
- - The live-title ceiling is fixed at 512 MiB rather than sized to the machine (30).
- - The 8-bit fallback clips HDR (31).
-- **API, server and CLI:** the `gpu` capability can never appear; a gRPC stream can hold `stop`; CLI `render` writes in place (32).
-- **Docs:** three misattached doc comments, the DirectML docs, the README badge and MCP, `ARCHITECTURE.md`'s known issues (33).
-- **PR hygiene:** #304 is still open; #314 needs a re-run; 11 contributor PRs are unreviewed (34).
-
-## 🧭 What the 11 Oct audit asked, and what landed
-
-✅ **Landed:**
-- Windows hangs named, GPU suites skipped on WARP, engine job timed out, toolchain action pinned (no tag yet).
-- The three dialogs.
-- Digests, SHA pins, `env:` for the tag, signing secrets declared, the real Android key used twice.
-- The locale check prints anywhere and counts English.
-
-🟡 **Partly:** uncaptured errors are counted and scoped, with half of it left.
-
-❌ **Not landed:**
-- Relink, cache clear and trials off the UI thread.
-- The paste.
-- `Server::stop` and `catch_unwind`.
-- The pinned-frame budget and its kin.
-- The literals through `I18n`.
-- Anything seen on real Windows or a real phone.
-
-## ▶️ Next, in order
-
-1. **Release:**
- - Fix the previous-tag lookup.
- - Finish and commit the uncommitted phone work, on its own:
- - the doc line;
- - a patient final drain;
- - `smoke-ios.sh` checks the exit status;
- - a `trap` in `smoke-android.sh`;
- - "Refs #286".
- - Run `mobile.yml` by hand.
- - Tag rc.3 and let every target finish.
- - Tag 0.2.7, with the draft's notes edited by hand.
-2. An iPhone document picker (or a notice); a failed start shown on a phone; the two plist keys.
-3. Make `Select`, `SegmentedControl` and `Switch` controlled.
-4. Held stills stamped per pass; one scratch name per save.
-5. Windows shaders: shared pipelines, cancel between builds, a visible skip, a scheduled real-GPU job, then DXC again on real hardware.
-6. GPU errors: refused vs lost, `combine` scoped, patient GLES waits; a sentinel and a setting for vendor encoders.
-7. CI:
- - `push` on `mobile.yml`, `workflow_dispatch` on `build-app.yml`;
- - `--smoke` through real frames, instruction-set limits from rc.2's counts;
- - pin `ubuntu-24.04` before 19 Oct;
- - Dependabot's sherpa exclusion;
- - `locales.py` fails on missing keys.
-8. Honest ONNX providers on Windows and Intel Mac; `CODE_SIGNING.md` in line with the workflow.
-9. Then the 11 Oct steps still open: the paste, the server, the pinned-frame budget, and the 28 Sept list.
diff --git a/audit/AUDIT-2026-10-12.md b/audit/AUDIT-2026-10-12.md
deleted file mode 100644
index cdef6f8a..00000000
--- a/audit/AUDIT-2026-10-12.md
+++ /dev/null
@@ -1,863 +0,0 @@
-# Concat audit, 12 October 2026
-
-Where the codebase stands one day after the 11 October audit, at
-`c06560e` (12 October), 39 commits past the last audit's `7223d64` and
-110 past the 0.2.6 tag, with two release candidates (`v0.2.7-rc.1`,
-`v0.2.7-rc.2`) tagged and failed in between. Same method, same scales,
-every finding from the 11th (and what it carried from the 9th, 4th and
-28th) re-verified against today's tree. Every finding names a file and a
-line. The sibling `AUDIT-2026-10-12-QUICK.md` is this audit as a
-one-screen scorecard.
-
-**How it was done.** Four reviewers each read one part of the tree: the
-window's Rust, Slint and locales; host, server, API and CLI; media,
-render, export and core; project, effects, text, vision, speech, Android,
-CI, release and docs. Each re-verified the 11 October findings in its part
-against today's lines, read the commits since the audit as commits (does
-each do what its message says, with what test, missing what), scored its
-part and reported new findings with locations. The headline items (the
-release notes' previous tag, the iPhone's Import, the pool's 24 held
-stills, the shared save scratch, the hard-coded hardware encode, the
-locale check's missing keys, the auto-closed issues) were re-read by hand,
-and where a command could show it, run. Line, test and `unsafe` tallies
-use the 11th's counting (Rust lines are `src/` only, without `build.rs`
-and `tests/` directories; the bracket is the count outside test modules),
-which the reviewers reproduced exactly on the 11th's tree before counting
-today's. `cargo fmt --check`, `cargo clippy --workspace --all-targets`,
-`cargo test --workspace --no-fail-fast`, `scripts/locales.py --check` and
-`scripts/models.py --check` were run on this machine (an Apple silicon
-Mac, macOS 26, Metal) today. The app itself was not launched; no phone or
-release bundle was built here.
-
-**The working tree is not clean.** Uncommitted on this machine and
-intended for 0.2.7: a send-and-receive retry for encoders that answer
-"try again" (`concat-media/src/encode.rs`, aimed at Android's MediaCodec,
-#286), Android-emulator and iOS-simulator launch checks
-(`scripts/ci/smoke-android.sh`, `smoke-ios.sh`, +197 lines of
-`mobile.yml`), and an error exit when the event loop fails
-(`concat/src/lib.rs`). This audit scores `HEAD`; section 6 reviews the
-uncommitted work separately, because it gates the release once
-committed. The test run below includes it.
-
-**What the 39 commits are.** All on 11 and 12 October; 33 by the
-maintainer, 6 by three contributors (Bengali, Vietnamese, `--project` on
-launch). In groups:
-
-- *The release pipeline* (d118484, b161c1d, 2d3a43e, c06560e): every
- bundle launched before upload (`--version`, `--smoke`, an emulated Ivy
- Bridge, an instruction scan), every download pinned by digest, every
- action by SHA, nextest with per-test timeouts, draft releases, Windows
- signing by thumbprint; then the fixes the two release candidates found.
-- *Windows* (be1b882, 3178b78, 0a61b65): DXC in, then out the same day;
- the GPU suites that took minutes on WARP skip there; no Microsoft Store;
- the installer in the app's languages. `main` is green because of these.
-- *Bugs from the tracker* (132f68d, 34fe37a, 590e483, f1ccdfe, 049823f,
- e5d8197, e1b64a4, b4adc57, 50c18ee): iPhone projects, settings and logs
- (#299); a second Untitled project (#151); a click that moved a clip
- (#247); Detach that said nothing (#310); no sound band on a silent file
- (#283); the last three dialogs off the studio borrow; the inspector's
- section strip (#250) and a title's words field (#139); titles painted
- once (#261); Chatterbox read again on the CPU (#291); an export the GPU
- stops answering ends, and a phone's GPU draws within its limits (#297,
- #312); no hardware encoder opened at launch (#309); a dropdown scrolls
- under a finger.
-- *Features* (2d3e651, 5982dc7, 45c061f): Generate captions on a clip's
- menu and a "Use voice from timeline" switch (#302); `--project` on launch.
-- *Languages* (30a1742, ad59b61, c65295b, 8a628c6, c0c8827, fa51932,
- 5c60bfd, 9737836): 14 languages to 25, a README in each, five new Noto
- fallback faces with a glyph test per locale, a locale check that prints
- on Windows and counts English.
-- *Housekeeping* (1d22c12, 7c5cb3d, 0714fb8, 4affa32, 2d27766, af62333,
- 91a6df0, the 11 October audit itself): the CHANGELOG, concat-perf and the
- web build removed; AGPL-only licence text; a code-signing policy; the
- version bumped to 0.2.7.
-
-**Scale.** 0 to 10. 8 and up means "keep doing this"; 5 to 7 means
-"works, with known debt"; 4 and under means "will bite at the next order
-of magnitude, or already bites".
-
----
-
-## 1. The numbers
-
-### 1.1 Size, tests and hygiene per crate
-
-Deltas are against the 11 October table. concat-perf (1 092 lines, 2
-tests) was removed in 7c5cb3d; like for like the tree grew by 2 007 lines.
-The 11th's concat-render row counted `src/gpu/tests.rs`, a `#[cfg(test)]`
-module (`gpu.rs:3560-3561`), as non-test code: its "(68)" unwraps and "3"
-panics were 59 and 3 in tests. Corrected below; the totals' brackets
-follow.
-
-| Crate | Rust lines | Δ | Slint / WGSL / Java | Tests (files) | `unsafe` | `unwrap`/`expect` (non-test) | `panic!` family (non-test) |
-|---|---:|---:|---:|---:|---:|---:|---:|
-| concat (window) | 23 923 | +954 | 29 901 slint (+111) | 87 (20) | 13 | 37 (4) | 0 |
-| concat-host | 10 842 | +202 | | 89 (19) | 4 | 249 (2) | 0 |
-| concat-project | 10 285 | 0 | | 114 (1) | 0 | 356 (11) | 6 |
-| concat-media | 10 066 | +212 | | 100 (11) | 38 | 204 (23) | 0 |
-| concat-render | 7 723 | +273 | shaders inline in `gpu.rs` | 91 (5) | 1 | 86 (9) | 0 |
-| concat-effects | 7 256 | 0 | 1 764 wgsl | 72 (9) | 0 | 134 (5) | 2 |
-| concat-export | 4 039 | +168 | | 34 (3) | 0 | 31 (0) | 0 |
-| concat-speech | 3 307 | +198 | | 21 (3) | 3 | 23 (2) | 0 |
-| concat-core | 2 897 | 0 | | 49 (7) | 0 | 26 (4) | 2 |
-| concat-vision | 2 760 | 0 | | 31 (8) | 0 | 22 (2) | 0 |
-| concat-text | 2 607 | 0 | | 24 (1) | 1 | 67 (1) | 0 |
-| concat-api | 2 352 | 0 | | 21 (2) | 0 | 23 (0) | 0 |
-| concat-server | 1 393 | 0 | | 12 (4) | 0 | 28 (0) | 0 |
-| concat-cli | 514 | 0 | | 3 (1) | 0 | 2 (2) | 0 |
-| concat-android | 398 | 0 | 291 java | 0 (0) | 10 | 1 (1) | 0 |
-| ~~concat-perf~~ | – | −1 092 | | – | – | – | – |
-| **Total** | **90 362** | **+915** | **~31 960** | **748 (96)** | **70** | **1 289 (66)** | **10** |
-
-The window took most of it again (+954: `launch.rs`, the untitled naming,
-the detach plan, the captions sources, the fonts table, `leave()`), then
-host's title cache (+202), render's bounded waits and phone limits (+273),
-media's encoder seams (+212) and speech's CPU retry (+198). `studio.rs` is
-10 866 lines (+177; 45 % of the window's Rust): `impl Studio` from `:1962`
-to `:10419` is 8 458 lines and about 220 methods over the same 93 fields,
-and `run()`'s body moved into `fn start` (`lib.rs:161-1932`), 1 772 lines.
-One `unsafe` was added (`leave`'s `libc::_exit`, with a SAFETY line); all
-70 carry one, and the one that is not true as written (`platform.rs:
-347-350`) is as it was. 31 tests landed on 2 007 lines, every one on a
-pure helper that fails without its fix.
-
-Other facts: 123 effect packages, unchanged; 25 locale files (en and 24)
-at 1 020 keys each, 0 missing and 0 stale, with every `{n}` placeholder
-matching English (checked by hand; the script does not); 993 values still
-the English word for word, as the script now counts it: fr 150, de 142,
-it 125, pt-BR 125, es 123, tr 109, ko 99, nl 44, hr 31, id 28, pl 10,
-vi 4, fa 2, bn 1, and ten locales at 0. 39 commits since the last audit
-(161 files, +21 465 / −3 445, most of it locales and READMEs); 32 tags,
-latest `v0.2.7-rc.2`; the last release is `v0.2.6`, 110 commits behind
-`main`. 82 open issues (89 on the 11th) and 19 open pull requests, 11 of
-them from one contributor, unreviewed since 6 October.
-
-### 1.2 The checks, run today
-
-| Check | Result |
-|---|---|
-| `cargo fmt --all -- --check` | clean |
-| `cargo clippy --workspace --all-targets` | 0 warnings |
-| `cargo test --workspace --no-fail-fast` | **745 passed, 0 failed, 1 ignored**, on Metal: the GPU suites ran for real |
-| `scripts/locales.py --check` | 25 locales, 1 020 keys, 0 missing, 0 stale; warns 993 English values; **would pass with keys missing** (finding 13) |
-| `scripts/models.py --check` | 23 models, 3.37 GB, mirrored on `models-v1` |
-| `cargo audit` | not installed on this machine; not run |
-| CI on `main` | **green** at `2d3a43e` and on every completed push since `3178b78` (six in a row, the first since 4 October); `c06560e` in progress at the time of writing |
-| Release, `v0.2.7-rc.1` | CI passed; **all six desktop builds failed** (Windows smoke silent for 15 minutes; Linux and macOS smoke scripts without the exec bit; Intel Mac build past 60 minutes; Windows on ARM's LLVM installer); Android and iOS passed |
-| Release, `v0.2.7-rc.2` | Linux arm64's `--smoke` logged `smoke: ok` and **segfaulted on its way out** (139); the rest cancelled by hand at 20:07; Android passed and was signed with the real key |
-
-**How `main` went green.** d118484 moved CI to nextest
-(`src/.config/nextest.toml:12-40`: slow at 60 s, killed at 5 minutes,
-JUnit kept on failure), which is what the 11th asked for, and the next
-Windows run named its hang at once: `every_package_probe_holds`,
-`every_shader_package_renders_at_its_defaults`,
-`the_gaussian_blur_is_a_direct_gaussian` and `every_edit_still_exports`,
-each past its limit on WARP (run 38209190014). That is the FXC compile of
-the blur and glow loops, the same cost #297 is about. be1b882 switched
-D3D12 to DXC to cure it; two GPU tests failed outright under DXC on WARP,
-and 3178b78 reverted it 38 minutes later and made those suites skip on a
-software adapter (`concat-render/src/gpu/tests.rs:41-48`,
-`concat-host/tests/export.rs:1550-1569`). Windows has run 740 tests green
-in 15 to 17 minutes since. The cost is that nothing in CI compiles an
-effect package through D3D12 any more (finding 4).
-
-**What has passed in a release run.** No desktop target has run every
-check to the end. Linux x86_64 came closest (host smoke, the emulated Ivy
-Bridge and the instruction scan, then cancelled in the clean-distribution
-installs). The Windows smoke has never got past its first line; the macOS
-smoke has never run; the Intel Mac has never finished a build. 2d3a43e
-(Windows smoke and emulator split, 100 minutes for the Intel Mac) and
-c06560e (`_exit` on Linux) are unproven until rc.3.
-
-### 1.3 Tests
-
-`cargo test --workspace --no-fail-fast` ran to completion here: 745
-passed, 0 failed, 1 ignored (the one ignored is concat-speech's model
-test). It ran on the working tree, so it includes the uncommitted
-encoder retry and its one test. The attribute count (748) is higher than
-the run because a few tests are compiled for other platforms only. On
-this Mac the GPU suites that skip on Windows' WARP ran on Metal and
-passed, including all 123 packages at their defaults, which is the only
-place today they were run on a real GPU. There is no performance table:
-concat-perf is gone (7c5cb3d), and nothing replaces it.
-
----
-
-## 2. Scorecard for the whole app
-
-| Aspect | 11 Oct | Today | One line |
-|---|:-:|:-:|---|
-| Code quality | 6 | 6 | fmt green, clippy clean, four `expect`s in 24k lines of window code, every new helper pure and tested; against that, three doc comments stacked or misattached, ten commits that say "Fixes" against the repository's rule, and d118484 claiming a Windows suite that could finish when it could not. |
-| Design architecture | 8 | 8 | Held. Every native dialog now goes through one seam; platform rules (`dirs::convention`, `windows_kind`, `location_on`, `encodable_among`) are pure functions of their inputs, so each platform's rule is tested on every runner. |
-| Design philosophy | 6 | 6 | The week's fixes go to the cause where they can (bounded waits, an honest revert, a launch check per bundle); but two binding-break fixes treat the symptom while the three primitives that cause it stay, and the captions sheet brought the bug back a day later. |
-| Maintainability | 5 | 5 | `studio.rs` 10 866 lines (+177), `start()` 1 772, `gpu.rs` 3 561, `concat-export/src/lib.rs` 3 210, `encode.rs` 2 173; concat-project's 114 tests in one file. |
-| User likeability | 7 | 7 | 25 languages with a face for every script; captions from the clip's menu; a click no longer moves a clip; Detach says why; the iPhone can make a project. Against that: on an iPhone Import does nothing; the captions sheet can transcribe a clip it is not showing; titles drop out of the monitor during a drag; a Windows blur or glow export can sit for minutes. |
-| Performance | 6 | 6 | #261's title cache is right, but the pool's 24 held stills undo it mid-gesture and re-treat every title each frame; the FXC compile is back on Windows and repeats on every export; concat-perf is gone. |
-| Scalability | 5 | 5 | Unchanged: the fd per connection, the one Hub thread, the per-sample mixer scan, the per-edit bin clone. |
-| Security | 5 | 6 | Every desktop download pinned by digest (`scripts/ci/deps.lock`, `fetch.sh`), every action by SHA, the tag validated and passed through `env:`, Windows signing by thumbprint, the Android key proven twice. Against that: mobile downloads still undigested, the signing policy and the workflow disagree, the update chain as it was. |
-| Testing | 6 | 6 | 31 tests, each failing without its fix; per-test timeouts; every desktop bundle launched before upload. Against that: the skipped GPU suites report PASS, `--smoke` stops before the window exists, the emulated old CPU runs `--version` only, concat-android has no tests, and there is no performance gate. |
-| Documentation | 5 | 5 | The CHANGELOG's staleness ended by deleting it, and the notes that replace it would give 0.2.7 three lines; `CODE_SIGNING.md` contradicts the workflow; `ARCHITECTURE.md`'s known issues were stale within a day; the README still sells an MCP server. |
-| CI and release | 4 | 6 | `main` green six pushes running; a hang names itself in minutes; draft releases that need every bundle launched. Against that: two release candidates spent and no desktop target through every check, phones never built on a push, the bundle workflow runnable only by a tag, and the final notes' previous tag wrong. |
-
----
-
-## 3. Scores per crate
-
-Columns: code quality, maintainability, scalability, performance,
-architecture, design philosophy, security, testing. A dash means the
-scale does not apply. Arrows are against 11 October.
-
-| Crate | CQ | Maint | Scale | Perf | Arch | Phil | Sec | Test | Note |
-|---|:-:|:-:|:-:|:-:|:-:|:-:|:-:|:-:|---|
-| concat-core | 8 | 8 | 6 | 7 | 8 | 8 | 7 | 7 | No commits. `MAX_SIDE`'s doc (`frame.rs:28-31`, "the downlevel limit every GPU meets") is now contradicted by render's own `max_side`; dead `Project` and `source_duration`; `Neg` and `ceil` on `i64::MIN`. |
-| concat-project | 8 | 6 | 6 | 6 | 8 | 7 | 6 | 8 | No commits. `AddClip` places the whole media (`commands/clips.rs:34-38`), the speed curve grows over a neighbour (`properties.rs:285-308`), `DOCUMENT_VERSION` is 1 (`doc.rs:41`), all tests in one 4 294-line file. |
-| concat-media | 7 | 6 | 5 | 7 | 8 | 7 | 7 | 7 | b4adc57's seams are pure with injected `opens` and environment, and its tests panic if anything is opened; but an HEVC or AV1 export still opens NVENC first (finding 9). The uncommitted retry is right by FFmpeg's contract. |
-| concat-render | 7 | 5 | 6 | 7 | 8 | 7 | 7 ↑ | 7 | Bounded waits, per-frame error scopes, the adapter's own limits, an 8-bit fallback tested against the half-float picture. But on GLES "30 seconds" is about 2.15 (finding 7); `combine` and refused passes are unscoped (finding 11); four docs still promise a CPU fallback. |
-| concat-effects | 8 | 7 | 6 | 5 | 8 | 8 | 6 | 7 ↓ | No commits. The package suites no longer run on Windows CI: the 123 shaders are proven on lavapipe, Metal and nothing on D3D12. |
-| concat-export | 7 | 4 | 6 | 5 | 7 | 7 | 7 | 7 ↑ | The device-loss recovery is tested end to end through `render_on`; progress at least once a second; a phone-sized ceiling with a reason. A deterministic validation error now reads as a lost device (finding 10); `lib.rs` 3 210 lines. |
-| concat-host | 7 | 6 | 5 | 6 | 6 | 6 | 4 | 6 | The #261 LRU and the iPhone directories are small, pure where they can be, and tested; the pool's FIFO undoes the first mid-gesture (finding 5) and overlapping saves share a scratch file (finding 6); the update chain as it was. |
-| concat (Rust) | 6 | 4 | 5 | 6 | 7 ↑ | 6 | 6 | 5 ↑ | Every dialog through one seam; 14 tests on pure plans (`detach_plan`, `moved_any`, `location_on`, `launch::parse`); `leave()`. Against: the iPhone's Import (finding 2), three misattached docs, the paste unchanged, `studio.rs` +177. |
-| concat (Slint) | 7 | 7 | 6 | 7 | 6 ↓ | 6 | – | – | Two binding-break fixes and a finger-scrollable Select; but `select.slint:127`, `segmented-control.slint:43` and `controls.slint:559` still assign their own value, about 33 one-way bindings remain, and the captions sheet re-introduced the bug (finding 8). |
-| locales and `locales.py` ★ | 8 | 8 | 7 | – | 8 | 8 | – | 7 | 25 languages with READMEs in parity, a face for every script with a charmap test per locale, English counted per locale. Missing keys do not fail; no placeholder check; no right-to-left layout. |
-| concat-api | 8 | 7 | 5 | 5 | 8 | 4 | 5 | 7 | No commits. The parity promise (`message.rs:23-26`) is behind again: captions' two new entry points are window-only. `export.run` overwrites; the `gpu` capability can never appear. |
-| concat-server | 7 | 8 | 5 | 6 | 8 | 8 | 5 | 6 | No commits. The wildcard-stop hang, the fd per connection, seats before auth; a gRPC `Events` stream can hold `stop`. |
-| concat-cli | 7 | 8 | – | – | 7 | 6 | 6 | 2 | No commits. No logger; `serve` parks with no signal handling; `render` writes in place. |
-| concat-speech | 7 | 5 | 6 | 6 ↓ | 7 ↑ | 7 | 8 | 6 | A typed `ReadError` and a once-only CPU retry, with two pure tests; but Windows ships the CPU ONNX Runtime while the crate asks for, logs and retries DirectML (finding 15). |
-| concat-vision | 8 | 8 | 5 | 5 ↓ | 9 | 8 | 8 | 7 | Code unchanged; cutouts, the brush and enhance are CPU-only on shipped Windows, and the Intel Mac's runtime may not serve ort's API at all (finding 16). |
-| concat-text | 7 | 6 | 5 | 6 | 8 | 8 | 7 | 6 | No commits. |
-| concat-android | 7 | 7 | – | 7 | 7 | 7 | 7 | 1 | No code commits; three bridge crash paths; 0 tests. The uncommitted emulator check would be its first. |
-
----
-
-## 4. Scores per feature
-
-Status: **done**, **partial** (works with a known gap), **thin** (present,
-weak). Rows that changed since the 11th carry a reason; ★ marks a row new
-this audit.
-
-| Feature | Where | Status | 11 Oct | Today | Why |
-|---|---|---|:-:|:-:|---|
-| Launcher, new project, recents | `start.slint`, `panes/start.rs` | done | 8 | 8 | Open off the borrow; "Untitled project N" (#151) and the iPhone's location (#299), with four tests. A Browse brings #151 back (finding 23); the Files app cannot see the iPhone's folder (finding 12). |
-| Import, bin, drag and drop | `media-pane.slint`, `host.rs:377-400` | done | 8 | 8 | Menu and bin Import through `pick_files_then`. On an iPhone it does nothing (finding 2). |
-| Timeline editing | `lanes.slint`, `studio.rs` | done | 8 | 8 | A click no longer moves a clip (#247, tested), Detach says why (#310, tested), no band on a silent file (#283). The 4 px threshold is under a finger's slop (finding 22); Detach unmutes a video muted on purpose (finding 24). |
-| Copy and paste | `studio.rs:6855-6966` | partial | 6 | 6 | Unchanged. |
-| Volume line | `lanes.slint` | done | 7 | 7 | Absent on a file with no sound, consistently. |
-| Undo and redo | `editor.rs` | done | 8 | 8 | Unchanged. |
-| Transitions | `plan.rs`, `gpu.rs:3450-3518` | partial | 7 | 7 | `combine` runs outside the new error scopes. |
-| Effects and filters | `concat-effects/packages/` | done | 8 | 8 | Code unchanged; on Windows the blur and glow compile cost of #297 ships again (finding 4). |
-| Colour and HDR | `model.rs`, `gpu.rs`, `encode.rs` | done | 7 | 7 | The 8-bit fallback on a phone clips HDR with a log line only. |
-| Colour grading | `concat.color-wheels` | done | 7 | 7 | Unchanged. |
-| Titles and text | `concat-text`, `titles.rs` | done | 8 | 8 | #261: a timeline past the budget is painted once (049823f, tested); more than 24 titles drop out of the monitor during a drag (finding 5). |
-| Shapes / Stickers | `media-pane.slint`, `clips.rs` | partial | 6 | 6 | Unchanged. |
-| Keyframes, animations | `model.rs` | done | 7 | 7 | Unchanged. |
-| Crop, transform, blend, masks | `plan.rs` | done | 7 | 7 | Unchanged. |
-| Cutout, smart brush | `segment.rs` | done | 7 | 6 ↓ | CPU-only on shipped Windows; the Intel Mac's ONNX Runtime may refuse its first session (finding 16). |
-| Enhance | `enhance.rs` | thin | 6 | 5 ↓ | The same runtime as cutout; copies still in `cache/`. |
-| Auto-captions | `captions.slint`, `panes/captions.rs` | done | 8 | 7 ↓ | ★ On a clip's menu, with a source list and a switch (#302); but the sheet can show one clip and transcribe another, and show one placement and use another (finding 8). |
-| Text-to-speech | `tts.rs`, `chatterbox.rs` | done | 7 | 7 | The #291 retry helps CoreML; on Windows DirectML is not shipped at all, so the retry doubles a real CPU failure (finding 15). |
-| Voiceover | `record.rs` | partial | 6 | 6 | A crashed take is still lost. |
-| VU meters | `meters.rs` | partial | 6 | 6 | No commits. |
-| Templates | `templates.rs` | done | 7 | 7 | Unchanged. |
-| LUT import | `cube.rs` | done | 8 | 8 | Off the borrow now. |
-| Custom and community packages | `catalogue.rs` | done | 7 | 7 | Unchanged. |
-| Proxies | `proxy.rs` | partial | 5 | 5 | Unchanged. |
-| Export | `panes/export.rs`, `concat-export` | done | 8 | 8 | A GPU that stops answering ends or recovers the export instead of freezing it, and a slow frame moves the bar. Against that: an FXC compile on Windows that nothing bounds or cancels, a refused effect silently absent, a validation error reported as a lost device. |
-| Export on a phone | `platform.rs`, `ConcatFiles.java` | partial | 6 | 6 | GLES devices open within their own limits and draw in 8 bits where half floats fail; the uncommitted retry targets #286; a heavy frame fails after about 2.15 s (finding 7). Still not seen on a phone. |
-| Hardware encode | `encode.rs` | done | 6 | 6 | Nothing vendor-specific opened at launch (#309); the crash moves to the first HEVC or AV1 export (finding 9); "· hardware" is blank until an export has run. |
-| Playback and audio | `playback.rs` | done | 7 | 7 | Unchanged. |
-| Monitor stills | `pool.rs` | done | 7 | 7 | A dead headless compositor is replaced. |
-| Hardware decode | `hardware.rs` | done | 7 | 7 | Unchanged. |
-| Remote API | `concat-api`, `concat-server` | partial | 5 | 5 | Unchanged; one more window operation (captions from a clip) without a verb. |
-| Self-update | `updates.rs` | done | 5 | 5 | The Store branch gone; the chain as it was, and no Authenticode check although every PE is now signed. |
-| Logs | `logs.rs` | done | 7 | 7 | On an iPhone the log is now written at all. |
-| Keyboard menus, platform keys | `app.slint`, `platform.rs` | partial | 6 | 6 | Unchanged. |
-| Save Video Frame As, Save Audio As | `studio.rs` | done | 8 | 8 | Unchanged. |
-| Phone shell | `ui/phone/`, `concat-android` | partial | 7 | 6 ↓ | Dropdowns scroll under a finger; captions on the clip menu. On iOS, Import is dead and a failed start is silent (findings 2, 12); on Android a failed start vanishes too. |
-| Relink missing media | `relink.rs` | done | 6 | 6 | Unchanged. |
-| Themes and accent colours | `ui/theme/` | done | 8 | 8 | Unchanged. |
-| Languages | `locales/`, `fonts.rs` | done ↑ | 7 | 8 | ★ 25 languages and 25 READMEs; five subset Noto faces with digests and a charmap test per locale; English counted per locale. No right-to-left layout for ar, he, ur, fa; seven older locales 99 to 150 English; a missing key does not fail CI. |
-| Windows text | `src/patches/i-slint-renderer-skia` | partial | 5 | 5 | Untouched; unseen on a Windows screen. |
-| Accessibility | the whole UI | thin | 2 | 2 | 2 `accessible-*` properties. |
-| Windows data folders | `dirs.rs` | done | 6 | 6 | Unchanged. |
-| ★ Command-line launch | `launch.rs`, `main.rs` | done | – | 7 | `--project`, `--version`, `--smoke`, parsed and tested; any other argument now refuses to start (finding 25); `--version` prints nothing in a Windows console. |
-| ★ Bundle launch checks | `scripts/ci/smoke-*`, `isa-scan.sh` | partial | – | 6 | Well built: an import audit on Windows, clean-distribution installs, the CUDA strace, an instruction scan. No target through every check yet; `--smoke` stops before the window; the emulated old CPU runs `--version` only; the instruction limits are provisional. |
-
----
-
-## 5. The 11 October findings, re-verified
-
-Numbers are the 11 October audit's (its section 6).
-
-| # | Finding | Today | Evidence |
-|---|---|---|---|
-| 1 | `main` red; the Windows hang undiagnosed | **Fixed, by skipping** | nextest named four suites past their limits on WARP (run 38209190014); `heavy_gpu()` (`gpu/tests.rs:41-48`) and the glow skip (`concat-host/tests/export.rs:1550-1569`) return early on a software adapter; six green pushes since `3178b78`. What the skip hides is finding 4. |
-| 2 | Three dialogs opened under the studio borrow | **Fixed** (132f68d) | Start › Open `panes/start.rs:283-292` (answered as `StartMsg::OpenChosen`), Import LUT `studio.rs:7566-7572`, menu Import `lib.rs:1672`, bin Import and Import Font `lib.rs:710, 733`, all through `host::after_dialog` or `pick_files_then` (`host.rs:352-400`). No dialog call remains outside one. |
-| 3 | A paste does not keep its shape; a title's copy keeps a quarter | Unchanged | `place_copy` `studio.rs:6855-6966`: `AddClip` lands the whole media (`:6888-6896`, `commands/clips.rs:34-38`) then trims (`:6899-6918`); titles return early (`:6862-6887`); the clipboard is not cleared in `close_project` (`:7218`). |
-| 4 | The toolchain action on a moving branch | **Fixed** | Pinned to a SHA in every job (`ci.yml:56, 252`; `build-app.yml:201`; `mobile.yml:111, 333`). Dependabot cannot move a branch pin, so it stays frozen until moved by hand. |
-| 5 | The CHANGELOG has nothing after 0.2.6 | **Obsoleted** (1d22c12 deleted it) | A release's notes are now its commits alone, and the previous tag is found wrong (finding 1). |
-| 6 | `Server::stop` can hang on a wildcard bind on Windows | Unchanged (reasoned) | `concat-server/src/lib.rs:269-272, 281-283`; called on the UI thread (`settings.rs:557-558`). |
-| 7 | A refused thread kills the accept loop silently | Unchanged | `json.rs:128, 183`. |
-| 8 | An uncaptured GPU error is a silently wrong export frame | **Partly fixed** (e1b64a4) | `render()` now draws under OutOfMemory and Validation scopes and compares an uncaught-error counter (`gpu.rs:3393-3441`); a failed frame is redrawn on a fresh device (`concat-export/src/lib.rs:1525-1541`). The rest is finding 11. |
-| 9 | `studio.rs` keeps growing | **Worse**, slowly | 10 866 lines (+177); `detach_audio`, `moved_any` and `detach_plan` were added to it (`:10266`, `:10439`, `:10455`). The week's other new code went into panes and pure functions. |
-| 10 | The API writes where the window asks | Unchanged | `concat-api/src/lib.rs:519-520` (no `exists`), `:380-397` (no `is_sane`); a `rate_den` of 0 leaves an empty project folder (`projects.rs:234-236`). |
-| 11 | Mux scratch names collide across exporters | Unchanged, **one correction** | `concat-export/src/lib.rs:976-978`. The window's embedded API shares the window's export slot (`settings.rs:582-583`), so the collision is between processes and same-stem outputs, not between the window and its API. |
-| 12 | The export's two decode paths disagree past the end | Unchanged | `lib.rs:1484-1487` against `pool.rs:427-432, 999-1008`. |
-| 13 | `pick_rate` and variable-rate files | Unchanged | `probe.rs:174-178, 212-232`; no VFR fixture. |
-| 14 | Five things the ceilings left | **Partly fixed** | Both devices now ask for the adapter's limits (`gpu.rs:548-552`; `concat/src/gpu.rs:94-98`), a device under 8 192 is refused at export start (`gpu.rs:849-851`, `lib.rs:1321-1328`), and the window's comment is right. `frame.rs:28-31` is still wrong; `transition_in.duration` has a floor only (`model.rs:1510-1512`); `VideoSettings::or` falls back silently. |
-| 15 | The server's quieter edges | Unchanged | `grpc.rs:74-76`, `hub.rs:54-66, 149-151`, `concat-cli/src/main.rs:172-173`, `concat-api/src/lib.rs:868-871`. |
-| 16 | The cards skip and the engine job | **Partly fixed** | The engine job has `timeout-minutes: 130` (`ci.yml:41`); the no-adapter branch still skips under `CONCAT_REQUIRE_GPU` (`cards.rs:598-601`). |
-| 17 | The signing step's unverified edges | **Fixed** | The alias from `keytool -list`, `--key-pass env:`, stdin from `/dev/null` (`mobile.yml:205-220`); the Windows secrets declared (`build-app.yml:68-73`); the real key used in both release candidates. |
-| 18 | The window's smalls | Unchanged | `host.rs:357-367` (no `refresh_art`), `dock.rs:19-20`, `meters-pane.slint:198, 206, 214`, the tray's keyboard path. |
-| 19 | concat-project's tests in one file | Unchanged | `lib.rs`, 4 294 lines, all 114 tests. |
-| 20 | `locales.py` crashes on a Windows console | **Fixed** (9737836) | `scripts/locales.py:43-47`. |
-| 21 | Doc drift | **Partly fixed** | `src/README.md` dropped perf and wasm but still omits concat-text and concat-vision (`:9-21`); the CPU-fallback docs remain (`concat-export/src/lib.rs:18-20`; `gpu.rs:676-678, 1295-1296, 1644-1648`; `concat-host/Cargo.toml:37-38`; `concat/Cargo.toml:124`; `platform.rs:335, 520-523`); `tts.rs:7`; MCP at `concat-api/src/lib.rs:7` and `README.md:58, 75, 106`. |
-| 22 | Dead code in core | Unchanged | `timeline.rs:314-321, 497-514`. |
-
-Of the 22: 5 fixed, 1 obsoleted, 4 partly, 11 unchanged, 1 worse. Of what
-the 11th carried from the 9th: #6 (dialogs) is **fixed**; #21 (downloads
-and action tags) is **mostly fixed**: every desktop and CI download goes
-through `scripts/ci/fetch.sh` against `deps.lock` (24 files, FFmpeg a
-dated build rather than `latest`), every action is pinned by SHA, and
-`inputs.tag` is validated and passed through `env:` (`release.yml:50-57,
-65-83`), leaving only the phone downloads (`ffmpeg-mobile.sh:38-39`,
-`sherpa-mobile.sh:38-39, 56-57`); #24 (tests that prove the wrong thing)
-is **partly** better (device loss now driven end to end, the hardware
-decode test honest) and partly worse (two of e1b64a4's tests restate the
-code: `gpu/tests.rs:1773-1781, 1850-1856`); and the 11th's correction to
-#20 needs one of its own: `finish` caps the RIFF size only past
-`u32::MAX`, so a take within 36 bytes below it still overflows at
-`record.rs:287` (a 12-hour take; trivial). The rest (#10, #11, #12, #13,
-#14 the walk, #15, #16, #17, #19, #20, #22, #23, #25 to #29, and the 4th's
-#4, #5, #7, #17 to #19, #23, #26, S1 to S4, S6, S8, S9 partly, S14, S16,
-S17) are unchanged, at the lines the 11th gave or a few lines on.
-
----
-
-## 6. New findings, by severity
-
-Numbers are this audit's. Each names where to look.
-
-### High
-
-**1. The final 0.2.7 release would ship three lines of notes.** The
-release finds the previous tag with `git describe --tags --abbrev=0
---match 'v*' "$tag^"` (`.github/workflows/release.yml:170`), and the two
-release candidates match `v*`. Run here, it answers `v0.2.7-rc.2`, and
-`release-notes.sh v0.2.7-rc.2 HEAD 0.2.7` prints three lines (the `_exit`
-fix, the Windows smoke split, the captions switch); from `v0.2.6` it
-prints 82. With `CHANGELOG.md` gone (1d22c12) nothing else carries
-0.2.6 to 0.2.7, and the draft (`:179`) catches it only if someone reads
-it. Add `--exclude 'v*-*'` when the tag has no hyphen (run here, that
-answers `v0.2.6`). (verified by running)
-
-**2. On an iPhone, Import does nothing and says nothing.** The phone's
-Import (`phone/editor.slint:850-854`) and Import Font (`:1103`) reach
-`platform::pick_files_async`, which on a phone asks `FILE_PICKER`
-(`platform.rs:686-692`); only `concat-android/src/lib.rs:223` ever
-installs one, so on iOS the answer is `log::warn!("no file picker on this
-platform yet")` and nothing on screen. 34fe37a and 132f68d made an iPhone
-able to save settings and create a project; without import it cannot
-edit anything. Export is the same shape: no publisher on iOS
-(`platform.rs:629-635`), so a finished file stays in the sandbox (and
-finding 12). Install a `UIDocumentPickerViewController` picker; until
-then, `notify` instead of logging. (verified)
-
-**3. Phones are never built on a push to `main`, and the new phone checks
-have never run.** `mobile.yml`'s triggers are `workflow_call`,
-`pull_request` and `workflow_dispatch` (`HEAD:.github/workflows/
-mobile.yml:4, 26-27`); all work lands on `main` directly, so an Android
-or iOS break shows only on a contributor's PR, a hand-started run or a
-tag. The uncommitted emulator and simulator checks inherit this, and once
-committed they gate `publish` (`release.yml`'s `publish` needs `phones`),
-so their first run would be rc.3. Run `mobile.yml` by hand on `main`
-before tagging; add `push: branches: [main]`. (verified)
-
-**4. On Windows, a blur or glow export can sit for minutes compiling
-shaders, and no test watches it.** With DXC reverted (3178b78), D3D12
-compiles through FXC, which be1b882's own message says takes minutes over
-the blur and glow loops (#297, open). Pipeline creation is CPU work
-(`gpu.rs:2497-2558`), so `DEVICE_WAIT` does not bound it; cancel is read
-at the top of each frame (`concat-export/src/lib.rs:1381`) and progress
-after one (`:1551-1554`), so Cancel does nothing until the compile ends.
-It repeats on every export: the window lends `monitor.sibling()`
-(`panes/export.rs:516`), whose `drawing_in` starts an empty `shaders` map
-(`gpu.rs:836-846, 1291`), so every effect the monitor compiled is compiled
-again. And CI no longer sees it: the suites that took past five minutes
-on WARP now return early and count as passed (740 run, 740 passed, run
-38302292964), their "skipping" lines swallowed by `store-success-output =
-false` (`nextest.toml:31`). Windows is the platform with the most
-downloads. Share compiled pipelines between siblings; check cancel between
-pipeline builds and show a "preparing effects" stage; make a skip visible
-in the step summary; run the full suites on a schedule with
-`CONCAT_REQUIRE_GPU=1`; say in the 0.2.7 notes that the compile cost
-remains. (verified against the runs and the code; the minutes are from
-#297 and be1b882's message)
-
-### Medium
-
-**5. More than 24 titles drop out of the monitor during a drag, and every
-title is re-treated each frame.** While a gesture is live
-(`begin_echo`, `studio.rs:2474`) the monitor paints every title live and
-hands each to `monitor.hold_still` (`studio.rs:2600-2618` →
-`preview.rs:152-154` → `pool.rs:628-652`); the pool keeps held stills
-FIFO, capped at `STILLS_HELD = 24` (`pool.rs:541, 634-638`). Each pass
-re-inserts in clip order, so only the last 24 survive; a title not held is
-looked for as a file at `memory://titles/…`, fails, and is dropped without
-a word (`concat-export/src/lib.rs:1913-1945`). Auto-captions add clips in
-time order, so on a captioned timeline the early captions vanish while
-anything is dragged, including a caption whose own size is being dragged.
-And `hold_still` calls `forget` for every title every pass (`pool.rs:640,
-675-685`), even for the same `Arc`, so each is treated again: the
-gesture half of #261 is undone. Hold only the titles the plan at the
-playhead covers, or stamp held stills per pass as `LivePaintings` does;
-skip `forget` when the frame is the same. (verified by reading)
-
-**6. Two saves of one project write the same scratch file.**
-`projects::save` always writes `concat.json.saving` (`concat-host/src/
-projects.rs:174`) and renames it. The window saves on a pool of three
-workers (`concat/src/host.rs:146-190`; `studio.rs:2500-2518`) and
-`close_project` saves on the UI thread (`studio.rs:7230-7235`): Ctrl+S
-twice, Ctrl+S during an autosave, or a close during a save puts two
-writers on one path. Interleaved writes can rename a manifest that does
-not parse ("not a Concat project", `projects.rs:224-226`); the loser's
-rename fails and toasts "could not save", and a close aborts. Take a
-per-path lock, give each save its own scratch name, and skip a save older
-than the one on disk. (mechanism verified; the corrupting interleaving
-reasoned)
-
-**7. On OpenGL ES, the 30-second device wait is about 2.15 seconds.**
-`wait()` passes `Some(DEVICE_WAIT)` (`gpu.rs:1311-1318`); wgpu-hal's GLES
-backend clamps it to `u32` and then `i32::MAX` nanoseconds
-(`wgpu-hal-30.0.1/src/gles/device.rs:1757-1764`, `gles/fence.rs:
-159-163`). On a GLES phone, the #312 device, a frame slower than that
-fails `read_back`, kills the compositor, is retried on a fresh device,
-asks for a software adapter a phone does not have, and ends the export as
-a lost GPU (`concat-export/src/lib.rs:1245-1247`). Not a regression (the
-old indefinite wait was capped the same way), but the doc
-(`gpu.rs:183-191`) is wrong on the platform it was written for. Loop
-short waits until `DEVICE_WAIT` has passed. (reasoned from wgpu's source)
-
-**8. The captions sheet can transcribe a clip other than the one it
-shows.** `Select`, `SegmentedControl` and `Switch` assign their own value
-(`select.slint:127`, `segmented-control.slint:43`,
-`inspector/controls.slint:559`), which in Slint drops the parent's
-binding. The sheet binds Switch `checked: root.from-voice`
-(`captions.slint:109`), Audio (`:123`), Model (`:134`), Placement (`:207`)
-and Size (`:216`) that way, is only hidden when closed
-(`modal.slint:34`), and Rust resets the fields on every open
-(`captions.rs:201-221`). So: Generate captions on clip A, pick B in
-Audio, close, Generate captions on clip C: the list shows B, Rust
-transcribes C. Placement and size have shown stale values since before
-this week. 590e483 and f1ccdfe fixed this class at two sites a day
-earlier; about 33 such bindings remain in `ui/`. Make the three
-primitives controlled (emit, never assign), which fixes every site at
-once; the sheet's own script box shows the pattern (`captions.slint:
-72-86`). (verified in code; the behaviour reasoned)
-
-**9. An HEVC or AV1 export still opens NVENC first, so #309 moved from
-launch to export.** b4adc57 stops start-up from opening a vendor encoder
-(`encode.rs:171-180, 442-479`, with tests), but the export always asks for
-hardware (`hardware: true`, `concat-export/src/lib.rs:1354`;
-`ExportRequest` has no field for it), and `encoders(true)` leads with
-`hevc_nvenc` or `av1_nvenc` (`encode.rs:117-130`). On the Optimus laptops
-#309 describes, `cuCtxCreate` segfaults at the first such export; the way
-out is `CONCAT_HARDWARE_ENCODE=0`, which nothing in the app mentions.
-Write a sentinel before a vendor-driver open and treat it as refused at
-the next launch if it is still there; expose a "Hardware encoding"
-setting. (verified by reading)
-
-**10. Any validation error now reads as a lost GPU.** A Validation scope
-error marks the compositor dead (`gpu.rs:3410-3437`); the export opens a
-second device, then WARP, draws the same frame on each, and ends "the GPU
-device was lost 3 times part way through the export - reset, out of
-memory, or not answering for 30 seconds" (`concat-export/src/lib.rs:
-1231-1243`). For a deterministic refusal that sends the person to their
-driver. On Windows it also opens the second device that `render_on`'s own
-doc (`:874-880`, #202) records as dying in NVIDIA's driver on laptops.
-Separate refused from lost: a validation error fails the export with
-wgpu's text; only loss, out-of-memory and the timeout recover. (verified
-by reading)
-
-**11. The 11th's finding 8, the half not closed.** `combine`, which
-packaged transitions use, has no error scope and the counter is not read
-around it (`gpu.rs:3450-3518`); a pass whose pipeline the driver refuses
-is filed in `refused` and skipped for good (`:2552-2558, 3056-3059`), so
-the exported file lacks the effect and only the log says so; and the
-window's lent export compositor carries a counter nothing increments
-(`drawing_in` makes a fresh one, `gpu.rs:1291`; the window's handler only
-logs, `concat/src/gpu.rs:113-115`). Scope `combine`; fail an export whose
-effect could not be drawn; share the counter. (verified by reading)
-
-**12. A phone that cannot start just vanishes, and an iPhone's work cannot
-leave the app.** `report_startup_failure` is a no-op on Android and iOS
-(`platform.rs:459-460`); at `HEAD` the end of `start()` exits 0 whatever
-`app.run()` returned (`lib.rs:1928-1932`), and `android_main` logs and
-returns (`concat-android/src/lib.rs:395-397`). The uncommitted `leave(1)`
-gives CI an exit code, not the person a message. Separately,
-`panes/start.rs:94-95, 116-118` makes Documents/Concat "so that the Files
-app shows it", but `src/scripts/ios-app.sh:54-97` sets neither
-`UIFileSharingEnabled` nor `LSSupportsOpeningDocumentsInPlace`, so Files
-does not. Show a Toast or dialog through the bridge on a failed start;
-add the two plist keys. (verified)
-
-**13. `locales.py --check` passes with keys missing.** A stale key sets
-`failed` (`scripts/locales.py:242-244`); a missing one is printed only
-(`:245-246`), against the docstring's "exit 1 on anything out of step"
-(`:5-7`). With 25 locales, three from outside contributors, a key added
-to `en.json` alone ships in English in 24 languages with CI green
-(`ci.yml:141`), and `i18n.rs:243-247` falls back silently. Fail on
-missing; add a placeholder check. (verified)
-
-**14. The release notes carry CI internals and a claim the next commit
-reverts.** `release-notes.sh:43` drops `ci:`, `docs:` and `chore:`, not
-`fix(release):`, so "The Windows smoke test and its emulated Ivy Bridge
-are two steps…", "The Linux and macOS smoke scripts can run…" and "A
-closing app ends with _exit, so v0.2.7-rc.2's arm64 bundle…" become 🐛
-lines; be1b882's "a blur or a glow no longer stalls an export" sits beside
-3178b78's "no DXC in 0.2.7"; and `i18n:` subjects keep their prefix
-because the type pattern `[a-z]+` (`:33, 53`) has no digits. Filter the
-`release` and `ci` scopes and `i18n`, and edit the 0.2.7 draft by hand.
-(verified by running)
-
-**15. Windows ships the CPU ONNX Runtime; concat-speech still asks for,
-logs and retries DirectML.** `build-app.yml:381-395` stages Microsoft's
-CPU `onnxruntime.dll` (a deliberate trade for old CPUs);
-`concat-speech/src/lib.rs:27-41` keeps `accelerated()` true on Windows,
-`chatterbox.rs:303-314` registers DirectML (silently dropped) and logs
-"asked to run on DirectML". So e5d8197's retry treats a real CPU failure
-on Windows as an accelerator failure, loads the model again and reads the
-whole text twice before reporting it, and every Windows diagnostic names
-a provider that is not there. `ARCHITECTURE.md:458` and
-`concat-vision/src/runtime.rs:7` say the same. Make `accelerated()`
-false on Windows until a DirectML runtime ships, and say "CPU" in the
-log. (verified by reading)
-
-**16. Cutouts and Chatterbox on an Intel Mac probably fail at their first
-session.** The workflow says so itself (`build-app.yml:91-96`): ort
-2.0.0-rc.13 asks for API 27, the bundled `onnxruntime 1.23.2` serves 23,
-"the first ONNX session … should fail". `--smoke` opens no session, and
-`macos-x86_64` has never finished a release build. An older `api-*`
-feature for that target, or the ONNX features hidden there, and a smoke
-flag that opens one small session. (reasoned from the workflow's note)
-
-**17. `--smoke` stops before the window exists, and the scripts say it
-draws.** It exits after `publish`, before `app.run()`
-(`concat/src/lib.rs:1900-1907`), as its own doc says (`:145-157`);
-`smoke-linux.sh:10-11`, `smoke-macos.sh:10-11`, `smoke-windows.ps1:17-18`
-and `build-app.yml:32-34` say "through the first frame of the window".
-Window and surface creation, the Skia renderer, the swapchain and the
-event-loop errors of #312 are never exercised by any launch check. Run
-the loop in smoke mode and quit from a one-shot timer after a few frames;
-the runners have Xvfb and WARP for it. (verified)
-
-**18. The instruction-set guard cannot catch a repeat of #193.**
-`isa-allow.txt:62-63` lets `concat` and `concat.exe` carry any number of
-AVX2, FMA, mulx/rorx and AVX-512 instructions, BMI only warns, and the
-"PROVISIONAL… set the limit once it has run" step (`:56-61`) was not
-taken after rc.2 printed real counts (avx2 18 127, fma 3 405, bmi 21).
-The emulated Ivy Bridge runs `--version` only (`smoke-windows.ps1:158`,
-`smoke-linux.sh:106`), which never enters ggml. So whisper's CPU backend
-going back to AVX2 would pass both. Set the limits from the measured
-counts with a margin, and run a short transcription under the emulator.
-(verified from the rc.2 log and the files)
-
-**19. The bundle workflow can only be run by pushing a tag.**
-`build-app.yml:45-74` has `workflow_call` only, and `release.yml` is its
-one caller: rc.1 found three script bugs, rc.2 one crash, and two fixes
-since are unproven. Add `workflow_dispatch` with a `ref`, or a nightly
-without `publish`. (verified)
-
-**20. `CODE_SIGNING.md` describes a different process from the
-workflow.** It says FFmpeg's and sherpa-onnx's DLLs are not signed with
-the certificate (`CODE_SIGNING.md:12-16`) and that every signing request
-is approved by hand (`:27`); the workflow signs every `.exe` and `.dll`
-not already validly signed, deliberately, because Smart App Control
-refuses unsigned DLLs (`build-app.yml:913-962`; #306, #311), from a PFX
-secret, automatically. One of them must change before SignPath reads the
-policy; the Smart App Control reason argues for the workflow. (verified)
-
-**21. Dependabot's cargo group is red by design.** `dependabot.yml:27-36`
-groups every crate; #315 moves `sherpa-onnx` 1.13.7 to 1.13.8, and the
-guard that `deps.lock` matches `Cargo.lock` (`ci.yml:125-128, 326-329`;
-`build-app.yml:433-436, 473`) fails every job, holding the other 16
-updates with it. Ignore `sherpa-onnx*` in the group and bump it by hand
-with `deps.lock`. (verified against run 38295637177)
-
-### Low
-
-**22. The 4 px drag threshold is under a finger's slop.** `lanes.slint:
-591-592` treats a press as a drag past 4 logical pixels; Android's own
-tap tolerance is 8 dp. A tap that wobbles 5 to 7 px on a phone still
-moves a clip: #247 survives on the platform with the most downloads. 8
-to 10 px when `platform::phone()`. (reasoned)
-
-**23. #151 comes back after Browse.** `free_untitled` runs on `Compose`
-and on an empty name at `Create` (`panes/start.rs:160-166, 243-247`), not
-on `FolderChosen` or `LocationEdited` (`:215, 175`); browsing to a folder
-that already holds "Untitled project" and pressing Create says it exists.
-(verified)
-
-**24. Detach on a video muted on purpose unmutes it.** `detach_plan`
-answers `Again` for any muted video with no detached sound
-(`studio.rs:10446-10447`), and `detach_audio` batches `muted:
-Some(false)` with the detach (`:10293-10305`); a mute the person chose
-cannot be told from one a deleted detach left. Mark the detach's mute, or
-mute the new sound clip in the `Again` case. (verified)
-
-**25. Any argument but the three flags refuses to start.**
-`launch::parse` returns the usage for anything else (`launch.rs:43-47`),
-which `main.rs:31-33` shows as "Concat could not start" in a dialog.
-Before 45c061f the arguments were ignored. The Linux packages' desktop
-entry is `Exec=/opt/concat/concat %U` (`build-app.yml:803`), and Windows'
-"Open with" passes a path: either now opens an error box instead of the
-app. Log and ignore unknown positionals, or drop `%U`. (verified)
-
-**26. Ten commits say "Fixes #N", against the repository's rule, and the
-push closed the issues.** 132f68d (five issues), 590e483, f1ccdfe,
-e5d8197 and e1b64a4 (two) use "Fixes"; GitHub closed #299 on the push at
-04:52 and it was reopened by hand at 04:54 (the issue's events). All are
-open now. "Refs #N" in the body, as the rule says.
-
-**27. Right-to-left languages are laid out left to right.** Arabic,
-Hebrew, Urdu and Persian ship with faces that shape them (`fonts.rs:53-58,
-104-112`), but nothing in `ui/` mirrors a row or flips an alignment. A
-Slint limit; say so in the README until it is handled.
-
-**28. The CI concurrency group cancels pending runs on `main`.**
-`cancel-in-progress: false` still cancels a pending run when a newer one
-queues; seven pushes have no CI record (d118484, 049823f, 9737836,
-0a61b65, 2d3e651, 0714fb8, 91a6df0), against the comment at
-`ci.yml:26-32`. Key the group on `github.sha` for pushes.
-
-**29. `ubuntu-latest` becomes Ubuntu 26 from 19 October.** Every run
-carries the annotation. Jobs on it: `ci.yml:37` (with apt names that may
-move), `mobile.yml:92` and the new emulator job, `release.yml:42, 118`,
-`nix.yml:35`, `models.yml:41`. Pin `ubuntu-24.04` before then.
-
-**30. The live-title ceiling is a constant.** `LIVE_CEILING` is 512 MiB
-(`titles.rs:98`), not sized from `memory::total()` (`memory.rs:10-14`)
-as the frame pool is (`pool.rs:577`); on a 3 to 4 GB Android phone a
-captioned timeline mid-gesture can hold that much on top of the pool.
-
-**31. The 8-bit fallback clips HDR with a log line.** On a phone that
-falls back to `Rgba8UnormSrgb` (`gpu.rs:174-181, 784-790`), an HDR
-timeline exports clipped and tagged HDR. Refuse HDR there, or say so in
-the sheet. (reasoned)
-
-**32. The API, the server and the CLI.** The `gpu` capability can never
-appear (`concat-api/src/lib.rs:180, 370-372`; the test pins its absence,
-`:1066`); a gRPC `Events` stream and `Server::stop` can wait on each
-other (`grpc.rs:167-170`, `lib.rs:281-293`; CLI only); `concat-cli
-render` writes in place with no `MAX_SIDE` check (`main.rs:417-426`).
-
-**33. Doc drift, continued.** Three doc comments stacked or misattached
-(`studio.rs:2127-2138`, `:7556-7565`, `:10258-10265`, the last leaving
-`clip_action` at `:10326` with none); `DEVICE_WAIT`'s doc (finding 7);
-two contradictory paragraphs on `read_back` (`gpu.rs:3312-3321`);
-`Gpu::acquire` drops a `request_device` error unlogged, so "no GPU
-adapter" is printed for a device failure too (`concat/src/gpu.rs:
-96-100`); `dirs.rs:32` ("downloaded models", now logs, titles, cards and
-updates too); `README.md:19` (a 0.2.7 badge whose alt text says 0.2.5,
-before 0.2.7 is out), `:52` (no iOS in About); `models.yml:19-20` against
-`:87`; `ARCHITECTURE.md`'s known issues (`:652-665`) stale since the day
-after they were written.
-
-**34. Pull-request hygiene.** #304 is still open although it landed as
-c0c8827; Dependabot's #314 failed on the old Windows hang and needs a
-re-run, and its `actions/cache` bump will conflict with the uncommitted
-`mobile.yml`; eleven contributor PRs (#262 to #275) are six days
-unreviewed.
-
----
-
-## 7. The uncommitted work, before it is committed
-
-Reviewed on the working tree, because once committed it is in the
-release.
-
-**The encoder retry (`concat-media/src/encode.rs`).** Right by FFmpeg's
-contract: on "try again", read the packets waiting and send the same
-frame again (`patiently`, `send`), for the frame and for the end of the
-stream; software encoders never answer it, so the desktop is unchanged;
-`written` still counts only sent frames; a stall leaves no file, because
-the scratch is removed and the output never touched. A good pure test
-with a fake. Before committing:
-- The old line `/// Writes every packet the encoder has ready.` now heads
- `send`'s doc, and `drain` has none. Move it back.
-- After the end of the stream, `drain` still stops at the first "try
- again" (`encode.rs:1341` in the tree). The contract says it cannot
- happen there; this change exists because MediaCodec's wrapper breaks the
- contract on the sending side, and if it does on the receiving side the
- tail of every phone export is lost with no error. Make the final drain
- patient too, in the same commit.
-- "Refs #286", not "Fixes": #286 says only "missing codecs", and no
- export has been seen working on a phone.
-- The Android launch check never exports. A one-second 720p H.264 export
- in it would exercise `h264_mediacodec` on the emulator's
- `c2.android.avc.encoder`.
-
-**The phone launch checks (`mobile.yml`, `smoke-android.sh`,
-`smoke-ios.sh`).** Well designed: Android's fatal lines are matched only
-for the app's own process or package, the foreground activity must be the
-app's, the native first line must appear, and logs and a screenshot are
-kept on every outcome; iOS checks the binary's platform and uses a fresh
-simulator, deleted in a trap. Before committing:
-- Run it once by `workflow_dispatch` on `main`; it gates `publish`
- (finding 3).
-- `smoke-ios.sh:184-188` passes on "smoke: ok" and ignores the exit
- status, so a crash after the line (rc.2's Linux failure exactly) passes
- on iOS.
-- `smoke-android.sh` has no `trap`; a `set -e` exit at `getprop` or
- `unzip` (`:128-136`) leaves `adb logcat` running and keeps no logs.
-- The simulator's environment is written inline in `mobile.yml`;
- `src/scripts/mobile-env.sh`, "the one description of the build
- environment", knows only the two device targets.
-- Check `git ls-files -s scripts/ci/smoke-*.sh` shows `100755` after
- `git add`; b161c1d exists because it did not once.
-- An emulator is not a phone: SwiftShader under ARM translation can pass
- what fails on a Mali Android 11 (#312). Record its first result as the
- baseline.
-
-**The failed-loop exit (`concat/src/lib.rs`).** Logging the error and
-exiting 1 through `leave` is right and fits the phone checks' patterns;
-it is for CI, not the person (finding 12).
-
----
-
-## 8. What is good, and should stay that way
-
-- **The 11th's first recommendation worked the first time.** nextest
- with a per-test timeout (`src/.config/nextest.toml`) turned a
- ninety-minute silent hang into four named tests on its first Windows
- run, and the JUnit report is kept on failure.
-- **The supply chain is one file.** `scripts/ci/deps.lock` pins 24
- downloads by digest, `fetch.sh` refuses and deletes a mismatch and
- prints both digests (`:66-78`), FFmpeg is a dated build, and a guard
- holds sherpa's pinned file to `Cargo.lock`'s version. Every action is
- pinned by SHA with its tag in a comment.
-- **A release cannot publish what did not start.** Every desktop bundle
- is launched where it was built before upload; Windows' launch audits
- the bundle's imports against a must-ship list and cuts `PATH` to
- System32 (`smoke-windows.ps1:96-150`); Linux installs the `.deb` and
- runs the AppImage in clean Ubuntu and Debian containers and checks that
- start-up opens no CUDA library (`smoke-linux.sh:85-150`, #309); the
- release is a draft that needs every bundle and both phones.
-- **Platform rules are pure functions of their inputs.**
- `dirs::convention` and `apple` (`dirs.rs:140-161`),
- `updates::windows_kind` (`:286-294`), `start::location_on`,
- `encodable_among` and `leads_with_hardware` with injected `opens` and
- environment (`encode.rs:442-479`): each platform's rule is tested on
- every runner, and b4adc57's tests panic if anything is opened.
-- **Every native dialog has one seam.** `after_dialog` and
- `pick_files_then` (`host.rs:352-400`): a dialog runs with nothing
- borrowed, and the phone and the desk share the call.
-- **Fonts are done properly.** Subset Noto faces with source and SHA-256
- per face (`fonts.rs:44-80`), each fallback once, and a test that every
- character of every shipped locale has a glyph.
-- **The phone GPU path is coherent.** `limits_for` asks for the adapter's
- own limits where it is below WebGPU's (`gpu.rs:548-552`), the 8-bit
- working format follows into every pool, target and byte count, and the
- fallback is tested against the half-float picture to ±2 levels
- (`gpu/tests.rs:1825-1847`).
-- **Device loss is finally tested end to end** through `render_on` with a
- dead lent compositor, redrawing the frame rather than dropping it
- (`concat-export/src/lib.rs:1517-1541, 2640-2706`).
-- **`LivePaintings` is the right shape** (`titles.rs:116-190`): one
- policy, stamped per pass, a pure test and a test that fails under the
- old FIFO. The pool's held stills should copy it.
-- **Honest reverts and honest messages.** 3178b78 says what failed and
- ships the known state; b161c1d, 2d3a43e and c06560e name the release
- candidate that found each bug, and `leave()`'s doc says what is guessed.
-- **The 70 `unsafe` sites each carry a SAFETY line**, and concat-core is
- still free of `unsafe`.
-
----
-
-## 9. Where we stand
-
-**Can 0.2.7 ship?** Not from today's tree. `main` is green, but no
-release run has passed every desktop bundle's launch check, the phones'
-new checks have never run, and the final release's notes would be three
-lines. What stands between the tree and a tag, in order:
-
-1. Fix the previous-tag lookup in `release.yml:170` (finding 1).
-2. Finish the uncommitted work: the doc line, a patient final drain,
- the two smoke-script fixes, "Refs #286"; commit it on its own (section 7).
-3. Run `mobile.yml` by hand on `main` and look at its result (finding 3).
-4. Tag `v0.2.7-rc.3` and let every target run to the end; c06560e and
- 2d3a43e are proven there or not at all.
-5. If green, tag `v0.2.7`, and edit the draft's notes by hand before
- publishing: drop the CI-internal lines and the DXC pair, and say that a
- blur or glow export on Windows can still take minutes to start
- (findings 4, 14).
-
-The 11 October audit ended with nine recommended steps. One day later:
-
-| Step (11 Oct) | Landed |
-|---|---|
-| 1. Make the Windows job name its hang; skip GPU suites on a software adapter; engine timeout; pin the toolchain action; tag 0.2.7 from a green `main` | all of it but the tag: named, skipped, timed out, pinned; two candidates tagged and failed |
-| 2. The three remaining dialogs; relink, cache clear and sweep on workers; trials off the UI thread | the dialogs (132f68d); nothing else |
-| 3. The paste: trimmed placement, the full patch for titles, clear the clipboard; move placement into concat-project | nothing |
-| 4. `Server::stop`, `Builder::spawn`, `catch_unwind`, a protocol whitelist, a write timeout, `exists` and `is_sane` on the API | nothing |
-| 5. Count uncaptured errors; scratch-name suffix; one past-the-end rule; a VFR fixture | the counter and scopes (e1b64a4), with half left (finding 11); nothing else |
-| 6. Digests, actions by SHA, `inputs.tag` through `env:`, Windows signing secrets, a release with the real Android key | all of it for the desktop (d118484 and after); the phones' downloads remain |
-| 7. Pinned-frame budget, WAV header, group clamp, speed-curve ripple, `DOCUMENT_VERSION` | nothing |
-| 8. The literals through `I18n`; a locale check that flags English and prints anywhere | the check prints and counts English (9737836); the literals as they were, and two new ones |
-| 9. The 28 September list; Windows text and the phone's export on real hardware | nothing seen |
-
-Three and a half of nine in one day, and the two that mattered most for a
-release (the gate and the supply chain) in full. The day's other work was
-the tracker, and it was taken well: every bug fix landed with a pure test
-that fails without it. The pattern the 11th named is still there, a
-little smaller: a fix that claims a little more than it does (the
-Windows suite could finish; a blur no longer stalls; #261's titles stay
-painted; "Fixes") and is caught by the next reading or the next tag. The
-new risk is the opposite of last week's: CI is green because the tests
-that see Windows' problem no longer run there, and the release pipeline
-can be exercised only by spending a tag.
-
-**Recommended order.** Each line is an hour to a few days, and each
-closes a finding above.
-
-1. The five release steps above. (1, 3, 4's notes, 14; section 7)
-2. iPhone import through a document picker, or a notice until it exists;
- a failed start shown on a phone; the two plist keys. (2, 12)
-3. The captions sheet: make `Select`, `SegmentedControl` and `Switch`
- controlled; then the ~33 bindings are right everywhere. (8)
-4. The pool's held stills stamped per pass and `forget` skipped for the
- same frame; one scratch name per save with a per-path lock. (5, 6)
-5. Windows shaders: share compiled pipelines between the monitor and the
- export; cancel between pipeline builds; a visible skip and a scheduled
- `CONCAT_REQUIRE_GPU` job; then try DXC again on real hardware. (4)
-6. The GPU's two outcomes, refused and lost; `combine` scoped; a refused
- effect fails the export; patient short waits on GLES; a sentinel and a
- setting for vendor encoders. (7, 9, 10, 11)
-7. CI: `push` on `mobile.yml`; `workflow_dispatch` on `build-app.yml`;
- `--smoke` through a few real frames; instruction limits from rc.2's
- counts and a transcription under the emulator; pin `ubuntu-24.04`
- before 19 October; the concurrency key; Dependabot's sherpa exclusion;
- `locales.py` failing on missing keys. (13, 17, 18, 19, 21, 28, 29)
-8. Windows' ONNX Runtime honest in speech and vision; the Intel Mac's
- runtime matched to ort or the features hidden; `CODE_SIGNING.md`
- brought in line with the workflow. (15, 16, 20)
-9. Then the 11th's steps 3, 4 and 7 as they stand (the paste, the
- server, the pinned-frame budget and the rest), and the 28 September
- list.
-
----
-
-## 10. Caveats
-
-Scores are judgements over code, calibrated against the 11 October scores
-but not measured against users or another editor. Every crate was read by
-one of four reviewers; the headline findings were re-read by hand.
-Findings marked (verified) were read at the line and, where a command
-could show it, run; (reasoned) means the path was followed through the
-code, or through wgpu's source, and not executed. The checks ran on an
-Apple silicon Mac with Metal, which is neither Windows' WARP nor a phone:
-the Windows stall, the GLES wait and the iPhone's import are read from the
-code, the workflow runs and the issues, not reproduced. The test run
-included the uncommitted encoder change. The app was not launched; no
-release bundle or phone build was made here; `cargo audit` was not run.
-The Linux arm64 exit fix (c06560e) compiles for that target and is
-explained in its commit, but it has not yet run on the runner that failed.