Move nested conditions into match guards and simplify equivalent Boolean
expressions without changing validation or short-circuit behavior. Keep
the explicit permitted Testkit dev-dependency arm in the layer checker.
Co-Authored-By: OpenAI Codex <noreply@openai.com>
Translate the shared animated image rectangle into the Before pane for
whole-frame drawing, cached window tiles and input mapping. Translate the
final target separately so render requests stay stable during the motion.
Cover intermediate mesh bounds and final geometry in both layouts,
including clicks in either pane, return to Fit and cached region tiles.
Fixes#1821
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Task: LRC-LIB-CATALOG-IMPORT
Count offline records, Local promotions and virtual copies correctly. Keep destination, warnings and errors in a persistent import result, with Show in Library clearing filters. Preserve library-switch ownership, reimport identities, existing edits and one-step undo.
Avoid repeated missing-original probes and unnecessary staged-photo, catalog-operation and develop-patch copies. Preserve upstream phased progress and typed cancellation. Address Astra review findings with regression coverage.
Validated the combined main tree with all eight cargo xtask ci gates (3,462 passing test executions; zero failures), fresh 30k import/persistence/metadata checks, duplicate-free same-session and reopened reimports, and visually inspected headless screenshots. Historical performance comparisons apply to the f730805d production changes before main integration; native Windows import/GUI acceptance remains unqualified.
Generated with Codex.
Photos dragged from the grid onto a Folders row move there on disk and in the library
(`photo.moveToFolder`: files and XMP sidecars, virtual copies follow, never overwrites, one undo
step), as Move Selected Photos Here in the row's menu does.
- The row lights up, and the label following the pointer says where they go: "Move 2 photos to
“Porto”" (onto an album the same gesture only adds them, so the two read apart).
- A closed folder held under the pointer for 0.7 s opens, so a folder inside it can be reached.
- Refused, with the reason on release and a not-allowed pointer: a row that covers several disks,
and Local photos (not in the library). Photos already there say "Already in “Porto”".
- A drop asks first, as Lightroom Classic does: "Move 2 photos to “Porto”?" (their files and
XMP sidecars move on disk; Edit → Undo moves them back), with Don't ask again, kept in Settings
as "Confirm before moving photos to another folder on disk" (on by default). The menu's Move
Selected Photos Here, chosen by name, goes at once. A question left open while another library
opens (the menu bar and agents reach past a dialog) moves nothing: photo ids are per library.
- The toast says how many moved and that Edit → Undo moves them back; the menu item says the same.
- Photos that leave the view leave the selection, so a Delete or edit next can't reach them unseen.
For NAS and network shares: the copy across volumes now streams (it read the whole file into
memory, a problem for a large raw or video) into a free hidden name that takes the photo's name only
once complete (a share dropping mid-copy left a truncated file under the photo's name, which a
retry reported as the photo already there). New engine tests cover a share going away:
Undo with the share offline or read-only is refused with the reason and kept until it can run,
nothing is deleted and the library stays linked; a connection lost partway through a move puts
back what moved; a file the share can't give back rolls the whole Undo back; and a real
move across devices where the machine has a second file system (`/dev/shm`).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Translate the shared animated image rectangle into the Before pane for
whole-frame drawing, cached window tiles and input mapping. Translate the
final target separately so render requests stay stable during the motion.
Cover intermediate mesh bounds and final geometry in both layouts,
including clicks in either pane, return to Fit and cached region tiles.
Fixes#1821
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A Shift-click with the brush paints a straight line from where the last stroke ended, tapering size, feather and flow when they changed in between, as one undo step. mask.brushStroke takes connect: true for it.
At 100% the photo's corner could sit half a physical pixel off, for
example in a 1279.5 pt canvas on a 2x display. The GPU then resampled
the photo with its bilinear filter and blurred it. On a 5152x7728 JPEG
the 100% view kept 68% of the fine detail.
fit_rect now rounds the corner to a whole physical pixel when the photo
is 1:1 with the screen. Other zooms resample anyway, and a pinch must
not jump, so they stay as they were. With the fix the same view matches
the original (71.4 dB PSNR, 100% of the detail).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
library.browse (engine and the UI's background read) switched the source to the
folder without touching the selection, so the previous view's photo stayed
active: Photo Grid marked nothing, and Detail, the histogram and the info
showed a photo the folder doesn't hold. Browsing now follows library.source's
rule through a shared Session::keep_active_in_view: the active photo stays when
the folder shows it, else the folder's first photo becomes active (also as the
background read's batches arrive).
In an album the Delete key ran photo.delete, which moved the photo to
Recently Deleted, so it also left its folder views. It now removes the
selection from the album only (album.removePhotos), as in Lightroom; in a
smart album it changes nothing. Photo > Remove from Album shows in normal
albums with the Delete key; Photo > Delete Photo still deletes from the
library. album.removePhotos defaults to the album on screen and returns
{removed, album}.
crs:Flipped="true" is how Lightroom Classic writes an ordinary radial mask
(effect inside); it was read as our invert, so every imported radial applied
outside the circle. Established from Lightroom Classic 15.6's exports of
synthetic DNGs:
- MaskGroupBasedCorrections, Version 2: Flipped is the effect-inside flag;
inverting is the component's MaskInverted.
- MaskGroupBasedCorrections without Version: inside whatever Flipped says
(Lightroom rewrites both values as Flipped=true, Version=2), and its Feather
is halved, as Lightroom rescales it on that upgrade.
- Legacy CircularGradientBasedCorrections Radial Filter: outside by default,
Flipped=true (Invert Mask) is inside.
One mapping serves XMP sidecars, DNG-embedded XMP, presets and .lrcat catalog
import. LightCraft writes no crs: masks, so there is no writer to match.
Found while testing crates/geom/src/crop.rs:59 (CropGeometry::constrained, via crates/engine/src/cmd/develop.rs:524 crop.set)
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Found while testing crates/engine/src/cmd/develop.rs:676 (crop.rotateAspect); also crates/geom/src/lib.rs:180 (Rect::fit_aspect)
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Buttons were egui's plain one, a fill set by hand where one had to stand out (a dialog's confirm,
About's Discord link), with no hover or press shade on a filled one and no shared sizes.
- `crate::button::Button`: what it is for (`Kind`: Primary filled with the accent, Secondary
with a neutral fill and a border, Danger filled red, Ghost with no fill until hovered), how
big (`Size`: Small 24, Medium 30, Large 36 points, each with its padding, text and icon size),
and what it shows (text, an icon before the text, or an icon alone, which takes a name for
screen readers and its tooltip).
- Hover and press darken it; disabled (or in a disabled part of the window) fades it and takes
no click, its tooltip or name still shown; a focused one shows an accent ring. A filled kind's
fill is darkened as little as needed for white to read at 4.5:1 (the light themes' accent and
the red would be too pale), in every state. Drawn within its own rectangle.
- Named for screen readers, registered for scripts and tests (`.widget(id)`). Its egui id comes
from the widget id (salted with `.id_salt`, e.g. the dialog's kind, so the keyboard's place
doesn't carry from one dialog to the next) or from its place in the ui, moved on so buttons
with the same text side by side stay apart.
- Dialogs' footers (Cancel, the confirm; at least 72 points wide) and About's Discord link use
it. The Export dialog's expected height in `tests_dialog_fit` grows by the 8 points of its
footer's 30-point buttons.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A design critique of the first run found the import review's "Import N Photos" a plain button at
the bottom left beside Cancel, though the welcome before it had taught that blue means go.
- Every dialog's footer is right-aligned: the confirm on the right, Cancel to its left, added in
that order so the keyboard goes through them as the eye does. The row spans the width the
window gives its content this frame, so it reaches the window's edge without widening it.
- The confirm is filled with the accent when it makes or brings in something (`confirm_style`);
Delete and Remove, and an informational dialog's Close, stay plain. Which dialogs only show
something is one list (`only_shows`).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CropGeometry::with_angle replaced a crop that no longer fit at the new
angle with the largest centred crop of its aspect. For a crop dragged in
from a corner (which still touches two edges) that happened at the first
fraction of a degree, so the crop jumped back to nearly the whole photo.
A crop that no longer fits now shrinks about its own centre, keeping its
aspect, to the largest size that fits there. The full photo, or a crop
an earlier turn shrank to the largest of its aspect, still follows the
largest crop and grows back towards 0 degrees (#1388). A crop whose
centre is off the turned photo is centred at its own size at most. No
crop the user made ever grows.
The loupe's window rendering already evaluates every visible mask's alpha
for an `Overlay::AllMasks` (its union is composed from the same alphas the
selected-mask overlay uses), but `overlay_supports_window` refused it, so
at 1:1 and closer "Show All Masks" fell back to a whole-frame render at
the zoomed size: slower, and its detail stopped at the frame render's
resolution. Allowing it makes the zoomed view a canvas-sized frame plus a
native window, as a single mask's overlay already is.
Tests:
- tests_preview_limit::an_all_masks_overlay_at_one_to_one_is_a_window
(fails before: "native 6000: the whole-frame render is 6000")
- tests_window::an_all_masks_overlay_window_matches_the_whole_render
(a window of the union matches the whole render's pixels, for every view)
`header` accepts a lossless frame but the decoder routing only sent
CMYK/cropped/12-bit/non-interleaved files to jpeg-decoder: an 8-bit SOF3
file went to zune-jpeg, which refuses lossless, so a listed file could not
open at all. The files that did reach jpeg-decoder were wrong too:
jpeg-decoder emits its lossless samples native-endian (its `to_ne_bytes`),
so `from_be_bytes` swapped every sample (a 12-bit 2048 decoded as 8), the
raw `precision`-bit value was never left-justified into the 16-bit range
`Buf::U16` feeds onward, and a 3- or 4-component frame over 8 bits still
reports `RGB24`/`CMYK32` while holding two bytes per sample, so it was
wrapped as 8-bit data over double-length rows. `scale()` (a DCT size, not
a lossless resample) also cut a lossless frame to the top-left when a
thumbnail box was requested.
Every SOF3 file now routes to jpeg-decoder; lossless frames skip `scale`
(`finish` scales after); lossless samples are read native-endian and
left-justified; high-precision RGB/CMYK become `Buf::U16`.
Four hand-built SOF3 regression tests (12-bit gray, 12-bit RGB, 12-bit
CMYK, and 8-bit through the public `decode`); all four fail before the fix
(byte-swapped samples, U8-wrapped high-precision buffers, and the zune
refusal) and pass after. The crate's other tests stay green.
`version.*`, `pointColor.delete` and `redeye.catchlight`/`redeye.delete`
parsed `index` with `f64_req(..)? as usize`. Rust's float-to-int cast
saturates, so `index: -1` became 0 and the bounds checks never saw a
negative: a caller sending `index: -1` silently deleted, restored or
renamed version 0 (or the first point-colour sample / red-eye correction)
and got `ok` back. `history.restore` already guards this and `spot_index`
uses `as_u64`.
New `index_req` helper in `cmd/mod.rs` (the same finite, non-negative
guard) is used by the five sites. Behaviour for valid indices is
unchanged.
Regression test `negative_indices_are_refused` creates a version, then
asserts `version.delete/restore/rename` with `index: -1` error and version
0 is untouched (before the fix the first one deletes it).
A PEF or ORF maker-note crop with an origin of `u64::MAX` and a width of
1 made `x + width` wrap to 0, so the guard read it as inside the image and
`RawImage::check`'s identical comparison also wrapped: `validate()` passed
and `normalized()` then indexed past the sample buffer. Debug builds
panicked at the guard itself ("attempt to add with overflow").
The active-area check and both vendor crop guards (PEF AOC, ORF
ImageProcessing) now compare with `checked_add`, so a wrapping rectangle
is a `Corrupt` error at validate time. Exact fits still pass.
Regression test `an_active_area_that_wraps_past_the_image_is_corrupt`
builds such an image and asserts `validate()` and `normalized()` return
`Err` (fails before the fix with the overflow panic above).
A script line for `ui.screenshot` without a `path` had its generated path
assigned with `params["path"] = …`: `serde_json`'s mutable index panics on
an array, a string or a number, and the CLI installs no `catch_unwind`, so
a script handed to the documented headless workflow aborted the process.
Only an object (or null, like an absent `params`) takes the path now;
anything else is a script error naming the line, like the existing handling
of a malformed line.
The regression test runs the real binary on such a script and asserts a
non-zero exit, no "panicked" in stderr and the `params` must be a JSON
object message (it fails before the fix with `cannot access key "path" in
JSON array`).
The host name lookup blocked with no deadline before the cancel check, and each address then got the full connect timeout. The lookup now runs on a helper thread under the connect deadline, which all addresses share.
Each deferred cache write started its own thread holding its image, with no limit. One writer thread per cache now takes them in turn, and a write is skipped once 128 MB is waiting.
A compressed PEF too short for its size, an uncompressed tile claiming far more samples than it holds, and an ARW6 with a failed tile now get refused before the big buffer is allocated.
A wrongly typed id or ids fell back to the selection, or kept only the valid part of the list. Session::targets now returns an error for those, and callers pass it on.
A sidecar write that failed because the folder was offline or read-only was only logged. Keep those photos waiting, write them again once their files are reachable, and tell the user.
Found while testing crates/geocode/src/lib.rs:172-195
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Adds Georgian (ქართული, code ka, system tags ka / ka-GE / ka_GE) with
complete plain and formats catalogs, the Edit > Language command and the
Settings choice. Sentence-cased messages go through i18n::upper_first,
which leaves Georgian letters alone (upper-casing Mkhedruli would turn it
into Mtavruli). The Settings language choices wrap now that eleven
endonyms no longer fit on one line, and the face-model Remove button goes
through tr().
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YMTAbGNRxnMEzRvYnQGphF
A library with no photos opened on the Detail view, whose empty state said "Choose a photo in
the grid or filmstrip" while the grid was empty and hidden: the only screen that mentioned
import was off screen, and Import itself was nowhere but the File menu and its shortcut.
- An empty library (no photos, or only deleted or browsed ones) opens on the Photo Grid with the
sidebar (Square Grid stays as it is), in the desktop app and in `lightcraft-cli snapshot`. The
desktop app switches before it notes what its settings file holds, so the switch alone isn't
saved.
- The grid and the Detail view show "Your library is empty" with Import Photos… (primary) and
Import Lightroom Catalog… (not on the web) as buttons, and a line saying photos and folders
can be dropped into the window, with the key Import Photos… is bound to now (`⌘⇧I` on macOS,
`Ctrl+Shift+I` elsewhere, or a remapped one). Not while a folder is browsed under Local or
Recently Deleted is shown: their photos aren't in the library but are on screen.
- The right-hand panel says "No photos yet / Import photos to get started" there instead of
"Select a photo to edit".
- The top bar has an Import button beside Export in every library. The centred search field's
reserve grows by both sides' share (620, and a 190 minimum), so the filter icon stays clear of
the icons on the right at every window width, also with macOS's title bar in the window. The
top bar's icons are named for screen readers, and a refused click says why in a toast.
- `empty_state` is `empty_message` with buttons that do the next step.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Lightroom Classic's Loupe Overlay ▸ Layout Image: a PNG with transparency (a
cover's masthead and cover lines, a banner's safe areas) drawn over the photo
in the loupe, fitted and centred, following zoom and pan. Opacity and matte
(the photo darkened outside the graphic's box) in steps, ⌥⌘O shows / hides it,
the choice is saved with the UI state. The file is read off the UI thread; an
unreadable one is reported on the canvas. engine::export::read_graphic is the
watermark logo's reader, now shared.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Mj2w3LwYjFhMHALZuFp2rU
Retain the complete reviewed change, including prior integration resolutions,
on current upstream.
#736: deliver Thai UI fonts in desktop and web releases
Use the merged craft-fonts input containing Noto Sans Thai across release,
FreeBSD and Nix builds. Retain the small Thai face in the wasm subset and
reject required release inputs that lack it. Exercise the reported Thai
folder name through all UI families and include original font licence texts
in Windows installers as well as portable packages.
Refs #736; completes delivery of the fallback added in #750.
Validation: UI tests, UI-only strict Clippy, 28 wasm checks, HEIF checks,
assets, layers, parity and formatting pass. All 21 unrelated workspace
failures reproduce with upstream Rust sources and the old font input;
existing Rust 1.95 metadata/engine Clippy findings remain unchanged.
A pass drawn again after a discard still has the widgets of the pass
thrown away, so a control request answered there found a row where it no
longer is; it also added a 240 fps sample to perf.fps. Control requests
and the fps sample now run on a frame's first pass only. The Keyword
List's hold is kept for the pass right after it was asked for, so it
never moves a panel that was not drawn in between.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brandon Thomas