Every demo-photo source load evaluated the procedural scene from scratch. A 24 MP demo
original costs ~200 CPU-seconds (AlpineLake, the photo the Detail tests open), and #323
added ~30 loupe tests that each load it at 1:1, Before and After separately. On the 4-vCPU
FreeBSD VM the ui-egui lib tests went from 1198 s to 3046 s, and the tests' 120 s/300 s
waits expired (loupe windows, fixture thumbnails starved of the shared rayon pool, the demo
detail snapshot).
Demo scene pixels are now generated once per (scene, size) and shared: concurrent loads
wait for one generation (Before/After at 1:1 no longer generate the original twice) and
the most recent results are kept up to half the memory budget. Locally, pinned to 4 CPUs,
the ui-egui lib suite now finishes in 155 s; without the change it had done 231 of 338
tests after 25 minutes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One config-dir helper (engine::config::config_dir; camera_profiles re-exports it) with both
the face-models and denoise-models defaults; hosts set up both model folders; Settings has
Faces and AI Denoise tabs; both command sets registered; both model dialogs; locale entries
from both; one [features] table per app manifest; Cargo.lock regenerated by cargo; the
DecodedSource cache cost counts both segmentation mattes and the denoised twin.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Keep current upstream UI behavior and the later detector contract checks. Validate candidate models before replacing working weights and records, with a synthetic regression that failed before the fix. Remove the unrelated GPU equivalence-test relaxation. All seven quality gates pass with the pinned hosted fonts.
Generated with Codex.
Photo > Detect Faces runs the YuNet detector (a small pure-Rust interpreter over the ONNX graph, no
runtime dependency) and adds the faces it finds as unnamed face regions in one undo step. The detector
and the recognition models are opt-in downloads: Settings > Faces shows each model's licence and
training-data note, Download stays disabled until the terms are accepted, and the file is fetched over
https with lightcraft-fetch from the model's own repository at a commit-pinned address, checked against
its recorded size and SHA-256 and then installed. Detect Faces offers the detector's download the first
time it is used. No model file is part of the repository (the real-model tests read LC_YUNET_MODEL; two
ignored tests download from the real hosts).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Recover the unpublished denoise work against current upstream. Preserve later opt-in model setup, shared Rust HTTPS downloads, pending-photo resume and an Amount-independent switch. Keep original RAWs unchanged and share the current color finish between ordinary and denoised sources.
Bound public tiling inputs, refuse corrupt-cache exports, and retain synthetic/real-model coverage. Include two Windows path spelling fixes in existing tests. No model weights, face code, media or Tract runtime are included.
Validation: cargo xtask ci (all seven gates); installed RawNIND CPU/GPU agreement 1.38e-6; native CPU preview/export on a public 24 MP Sony RAW, unchanged original SHA-256.
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Generated with Codex.
Keep upstream tests and the fixed SAM worker, refresh the workspace version and parity summary, and route the model picker through the case-insensitive native filter helper. Carry the shared translated font-symbol correction; validation includes the pinned hosted fonts and all seven CI gates.
Generated with Codex.
The model list marked the YuNet detector as bundled with LightCraft and installed, but no model ships
inside LightCraft (the weights are opt-in downloads or files the user adds). Drop the notion: a
model is installed when its folder is, it can be removed like any other, and the licence notice no
longer says it is bundled.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
An off-by-default "Recognise faces" toggle and the face model list in Settings,
with a one-drop install: pick a .onnx (Add a model file…) or drop one on the
window. The engine inspects it (recognised by hash, or described from its
shape), a dialog shows its licence, whether commercial use is allowed, what it
was trained on and what LightCraft had to assume, and Install stays disabled
until "I accept" is ticked. "Get…" opens a model's own page in the browser:
LightCraft never fetches a model itself. A file it cannot use says why.
The models folder is shared by the app, the CLI and the MCP server
(`engine::config`, `$LIGHTCRAFT_FACE_MODELS` overrides `<config>/models`), so a
model installed from one is there for the others. A session has none until a host
asks (`with_default_face_models`), which keeps tests hermetic.
Nothing runs a model yet, and the tab says so. Headless tests cover the whole
flow (the dialog, refusing until accepted, installing, a file it cannot use).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
`faces.models.list / inspect / install / remove / select` and `faces.enable`, all
reachable from the UI, CLI, control channel and MCP. Models live in a folder the
host provides (`Session::face_models_dir`): one subfolder each with the model, its
manifest and a record of what the user accepted, plus settings.json.
- install needs `acknowledged: true` (the user has been shown the licence), copies
the file under a .part name, checks the copy's SHA-256 against the original, then
renames; a failure leaves nothing behind.
- ids are validated before they become folder names (and a manifest that disagrees
with its folder is ignored); remove only ever deletes a folder that has a model
manifest. A manifest id can no longer be `.`, `..` or start with a dot.
- inspect describes any .onnx without writing anything; unusable files are an
answer ("unsupported" with the reason), not a crash.
- the list says plainly that running a model is not built in yet.
`faces::synthetic` makes a minimal valid ONNX for tests (and the later self-check).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Raw-IFD flag 0xf006 = 1 marks RAF samples packed MSB-first into 32-bit little-endian words at any depth; 12-bit files with it (FinePix S1, X-A10, SL1000) were read as an LSB-first stream. Nikon uncompressed 12-bit strips of exactly 1.5 bytes per pixel in a file whose maker note has an NRW data block (COOLPIX B700, P950, P1000, P1100) use the same words in file byte order; without the block (E5700, E8400, E8800) they stay an MSB-first byte stream. Shared unpack_words32_msb and Packing::Words32Msb. RENDER_CACHE_VERSION 15 to 16.
A custom white balance on a raw with a colour matrix (DNGs, and other raws
without a file-local look) re-develops the pixels for the chosen white
through the file's own colour model: lightcraft_raw::color::rebalance
re-evaluates the camera transform at the new white and maps the as-shot
development onto it, so a neutral lit by that white renders neutral and
saturated colours move as the camera's matrices say (DNG spec ch. 6, as
Lightroom does). The decoder passes the colour tags and the white the
pixels were developed for in SourceInfo::camera_color; other photos, and
relative white balances, keep the Bradford adaptation. The eyedropper and
Auto invert the same model (color::neutral_white), so a picked grey card
renders neutral.
The Tint sign is main's (#349): unchanged here. Render cache version 15.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Read DNG 1.6 semantic masks (SemanticName, SemanticInstanceID,
MaskSubArea; any coding the tile/strip reader decodes, e.g. ProRAW's
lossy-JPEG matte) with lightcraft_raw::semantic_masks, map them onto the
developed image, and carry Apple's sky / portrait / skin / hair / teeth /
glasses mattes in SourceInfo::mattes (SourceInfo is no longer Copy).
Sky, Subject, Background and People components sample them through
orientation, warp (Warp::frame, as geometry.rs) and crop like the image,
so preview = export, instead of the heuristics; files without mattes
render as before. Malformed mask IFDs are skipped, and only the first
MAX_MASKS mask IFDs are decoded, so a file full of them costs bounded
work. The render cache version is bumped once, in the white-balance commit.
The corpus test reads corpus/raw/dng-apple-iphone12pro-proraw.dng, the
name `cargo xtask corpus` downloads it under.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Parse ProfileGainTableMap (DNG 1.6) and ProfileGainTableMap2 (DNG 1.7):
both versions, DNG 1.7 precedence (version 2 in IFD 0, then the raw IFD;
version 1 in the raw IFD, then IFD 0), u8/u16/f16/f32 gains, gamma.
Malformed tags (wrong size, zero or huge grids, non-finite or negative
gains, bad spacing or gamma) are ignored, never allocated for.
Byte order: the specification gives this tag no byte order of its own,
unlike opcode lists, so the payload is read in the file's byte order with
a big-endian fallback; a wrong-order read can't pass the size checks.
The module docs quote the DNG 1.6 "Byte Order" rule.
The map is kept in ColorData, written back by the DNG writer (version 1,
or version 2 when its gamma isn't 1) and dropped from merge outputs, whose
geometry no longer matches the capture's active area. It is not applied
to the render: Lightroom Classic renders Apple ProRAW without it.
Supersedes the parser of #273.
The CC0 iPhone 12 Pro ProRAW from raw.pixls.us joins the corpus.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Place an album or folder before a sibling of its kind (also in another
folder), or last of its kind, in one undo step; album.sort goes back to
by name for one folder.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Album.order and Op::SetAlbumOrder; album_children lists folders first, then
by place, then by name. Moving an album drops its place and undo restores it.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A full-size load asks the decoder for the largest box (`SourceLevel::Full` is
`usize::MAX`, cast to `u32::MAX`). `jpeg::decode` then computed `2 * u32::MAX` to
decide on DCT scaling: an overflow panic in debug builds (caught, so every
full-size JPEG load failed with "decoder panicked"), and a wrap that happened to
give the right answer in release. The comparison saturates now, and the engine
converts its `usize` edge with `u32::try_from` instead of a truncating `as`.
Found decoding camera JPEGs (no JFIF/EXIF header) at full size.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Lightroom Classic reads `ToneCurvePV2012Red/Green/Blue` only together with
`ToneCurvePV2012`: a packet with a channel curve and no master curve leaves the
photo unchanged there (rendered on a chart: the channel curve alone has no
effect, with an identity master curve it applies). Imports now do the same.
The preset-packet test helper uses made-up as-shot values.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
XMP presets usually say LensProfileEnable=0, which turned optics.lens_profile
off and with it the DNG-embedded WarpRectilinear / FixVignetteRadial
corrections (iPhone ProRAW and other files that carry them). Lightroom
applies a file's built-in corrections whatever that switch says (our
inference), and LightCraft's profile corrections are only those embedded
ones, so crs import now carries over only an enabling switch. The Enable
Lens Corrections checkbox and the amount sliders keep working as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
XMP presets carry fields that describe the preset or the photo rather than
an adjustment (Cluster, SupportsAmount2, RequiresRGBTables, SortName,
Description, CropConstrainToWarp, OverrideLookVignette, AsShotTemperature /
AsShotTint). preset.import listed them as settings that couldn't be carried
over; so did an HDREditMode of 0 and PointColors slots that are all -1.
They are now treated as bookkeeping; an HDR edit, a used Point Color slot and
unknown adjustments are still reported.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CR2 and PEF decode without a camera colour matrix, like ARW, NEF, RW2,
RAF and CR3, but camera_preview::file_local_look (and the catalog's
relative white balance) left them out, so they rendered with the neutral
fallback matrix: the camera preview showed first, then a flat,
desaturated render replaced it. They now use the same guarded fit and
relative WB; a fit that fails the gates keeps the fallback, so a file can
only stay as it was or move closer to its camera JPEG. Not measured on
the corpus here (raw.pixls.us is blocked from this environment).
RENDER_CACHE_VERSION 15 re-renders cached CR2/PEF thumbnails.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01912WaK6Rmew3wxZuPyX7E8
At first open of Before/After the two window jobs were asked for together, each
decoded its own 24 MP copy of the original (a review saw two decodes and three
uploads for one photo) and the shared device copy was swapped between them. The
Before's window now waits until the After's job has handed the original to the
session (region::defer_before_window, Session media has_source); later windows
of both share the one copy.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Review findings: window growth for spots was capped silently and then rendered
wrongly; Auto Mask growth took the whole stroke's bounding box; hidden masks grew
windows; the wasm window memory cap was not enforced. window_for_reads_checked
returns None when what a window reads would need more than 8192 px (4096 on
wasm) along an axis, Session::region_job returns None then (the loupe keeps the
whole-frame render), Auto Mask growth is per dab near the request (once, not
chained along the stroke) and ignores hidden masks. Docs say what the limits are.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
An adversarial review found that sizing the loupe render from the cropped frame
(previous commit) left the source level chosen from the output edge alone: a
crop whose own long edge is under 2560 px was drawn from the 2560 px preview of
the whole photo, soft though sized right. The level now follows the pixels the
output needs along the source's long edge (output edge / crop fraction); grid
thumbnails keep their cheap level.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Session::region_job renders one PixelWindow of the loupe's frame from the source
level that frame size needs, with a key of its own and no cache writes.
ui-egui::region decides when a window is needed (the whole-frame render is
stretched) and which one: the visible pixels plus a margin, snapped to a 256 px
grid so small pans reuse the render, bounded in size, clamped into the frame.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The import probe of a non-raw image decoded it (at a 64 px fit, which for
JPEG still entropy-decodes every block and for PNG inflates everything)
only to learn its stored size and orientation: ~95 % of import CPU.
lightcraft-codecs gets `read_header(bytes) -> Header { format, width,
height, orientation }`: JPEG from its markers (EXIF orientation), PNG from
its chunks (IHDR, eXIf anywhere), TIFF from the first IFD (tiff crate,
no strips read), WebP from its RIFF chunks / bitstream header (image-webp),
PSD from its header and image resources. GIF, BMP and JPEG XL keep the
small decode. `files::probe_bytes` uses it; raw handling is unchanged.
The TIFF colour-layout check is shared with the decoder (`layout`).
Measured (release, 4 cores, throwaway benchmark, per file, bytes in
memory): 200 generated 4000x3000 JPEGs (3.6 MB): 72.7 ms -> 2.1 ms
(read_header alone 1.05 ms; the rest is the content hash and metadata);
20 PNGs (30.8 MB): 236-265 ms -> 9.0-9.6 ms.
Corrupt files (behaviour note): the header readers also refuse what a
decode refuses before the pixels (unsupported layout, compression, frame
type, bad IHDR, ...) and truncated files: a JPEG whose markers don't reach
its EOI, a PNG cut before the chunk after its IDAT run, TIFF strips/tiles
that start past the end, a WebP image chunk or PSD merged data past the
end. So truncated files (interrupted copies, partial downloads) are still
marked `error` in the import review, as before; tests check that whenever
the old decode accepted a (cut or byte-mutated) file the header agrees.
Changed: damage *inside* compressed data that is all present (flipped or
garbled bytes) was refused at scan and is now imported, and shows as
unreadable ("!" in the grid) when rendered. Also: tiny JPEGs (< 128 px)
were decoded leniently by the old probe, so a truncated one imported; it
is now refused. A TIFF truncated inside its last strip is not detected
(StripByteCounts are often overstated, so only strip starts are checked).
Tests: header readers vs decode on generated JPEG/PNG/TIFF/WebP/PSD/BMP
with and without EXIF/TIFF orientation; truncations and random mutations
agree with decode; hostile headers (zero/absurd sizes, bad lengths,
truncation, garbage) are errors; read_header added to the fuzz tests;
probe_bytes oriented dims match the decode-based values.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01912WaK6Rmew3wxZuPyX7E8
Root cause (in 0.2.1 and still on main): files and folders named on the
command line were imported from the eframe app-creator closure with
`app.run("library.import", …)`, which runs the engine command synchronously:
`import::expand` walked the folder recursively (one `is_dir()` open per entry,
no depth or count bound, following links) and probed every file before the
first frame could be drawn. An argument that resolves to a drive root (`.` or
`..` with a launcher's working folder at `C:\`, `/` or `\` on Windows) or to
the home folder therefore walked the whole drive with the window stuck as a
black "not responding" box. The library-problem banner's pending import had
the same synchronous call. Every other recursive walk (Local's Include
Subfolders, Find Missing Photos, preset and LUT folders, MCP) was also
unbounded, which is the "Local folders with subfolders: not responding"
pattern of filmcraft#23, and a link loop recursed until the stack overflowed.
- engine `walk::files_in`: one bounded, iterative walker (32 folder levels,
1,000,000 entries; `Walked::truncated` says it stopped), folder or file
taken from the directory listing (no open per entry), links and junctions
to a folder followed once. `import::expand` (+ `expand_within`), preset and
LUT folder expansion and MCP `expand_paths` use it.
- engine `import::launch_path`: an empty launch argument is ignored; a
relative one is resolved against the working folder; a drive or file-system
root becomes its memory card's DCIM folder or is refused, and so is the home
folder (a notice says why and to use File > Import Photos).
- desktop: launch arguments go through it and are imported with
`import::start_paths` (the import worker), like dropped files; so does the
library-problem banner's pending import.
Not verified on Windows (no machine here). The report says PhotoCraft 0.5.0
and `--memory` hang the same way with no arguments; nothing in LightCraft
walks the disk at startup without arguments (strace of a headless demo run
with HOME=/ and cwd /: no recursive listing), so that part may be outside
this code.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01912WaK6Rmew3wxZuPyX7E8
Brandon Thomas