The E2E showed the previous fix missed the common case: enlarging a locked
crop until it stops leaves a frame that fills the photo's height, with its
top and bottom both at the edge of is_within_image's 1e-6 tolerance. No
translation keeps both there with room to spare, so the move stayed stuck.
When the frame can't move as it is, solve again with it shrunk about its
centre by the least that puts it strictly inside (at most 1e-5 of its size,
aspect unchanged). This also covers a frame at the edge on one side, so the
half-tolerance first pass is dropped and a click moves nothing again.
A handle drag stops where is_within_image's 1e-6 tolerance ends, so a
corner can sit right at it. The move solve then let that corner go 1e-9
further, the result failed the check, and the across-then-down fallback
left the frame stuck. Solve first with every corner kept within half the
tolerance, and only then where it is.
A handle drag can leave the frame up to is_within_image's tolerance past a
tilted edge. The move's bounds were tighter than that, so such a frame was
stuck again, and a click inside it moved it. The bounds are now never tighter
than where the frame is. A test pins how a smaller frame pushed past a tilted
edge slides along it, and the parity checklist's revision history notes the fix.
On a straightened photo, a crop that fills the photo's height could not be
moved sideways: the move went across and then down, and no straight sideways
step stays inside the turned photo. The frame now goes to the place nearest
the pointer's that keeps it inside, so it slides along the tilted edges to the
photo's side. Unstraightened, that place is the old across-then-down clamp.
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>
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.
develop.set / develop.adjust / develop.quickAdjust / develop.resetControl on crop.angle
(MCP set_develop, the control channel, lightcraft-cli render --set) and crop.set {angle}
now turn the crop as crop.straighten does, through the new CropGeometry::with_angle:
kept if it still fits, else the largest crop of its aspect inside the rotated photo.
Before, they left the full rect, and the export took in smeared edge pixels in the corners.
Dragging a crop handle past the image edge used to hand the raw rectangle to
CropGeometry::constrained, which slid the whole frame toward the centre and
moved the anchored corner. drag_crop owns the drag instead: the opposite
edges stay put, a locked ratio holds on all eight handles (edges grow about
their midline, sliding along a blocked side), the frame stops at the image
edge and slides along it, and it keeps a minimum size. Hostile numbers return
the start crop. Covered by Given/When/Then scenarios and property tests.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Homography::apply, Affine::apply, Warp::to_corrected,
Warp::corrected_to_source and embedded_warp are now generic over
lightcraft_geom::Real (private *_real variants in pipeline, public
apply_real in geom); the f64 API is a thin wrapper, so callers are
unchanged. Real is implemented for f64 (plain IEEE operators) and for
Interval, which rounds every bound one ulp outward and turns division by
a range containing zero, the homography's 1e-300 denominator clamp and
non-finite bounds into "undecided" (NaN). Because the interval runs the
same operations as the f64 path, it bounds the rounded f64 results, not
only the exact ones.
The f64 path keeps the same operations in the same order (only commuted
operands, which IEEE + and x leave bit-identical; no fma). Proof:
warp_f64_results_are_bit_identical_to_the_pre_generic_formulas hashes the
bits of 2000 random warps x 40 points through every formula (plus the
clamp branch, including +-1e-300 exactly) and matches the digest computed
by the same test on thumb-split (419261c) before the change; reassociating
one expression, or `<=` for `<` in the clamp, changes it.
The framing chain (output px -> transformed -> lens-corrected -> green
source, then inside the edges) is written once: Warp::framing, generic.
Warp::frame (f64, returns the corrected and green source points) serves
both the CPU resample (sample_warped) and Warp::covers; block_coverage
evaluates the same chain with intervals over a whole block of output
pixels and compares the bounds against the same edges() inside uses, with
no margin: the bounds enclose every pixel's f64 result, so the decision
equals the per-pixel one; else None.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Regions are stored on the upright (EXIF-oriented) photo, but the loupe's normalized frame and the
close-up crop include the user's orientation. The loupe overlay now maps each box into that frame
(and a resize drag back out of it), and face_job maps the face and swaps the photo's size for a
quarter turn. Flips need nothing: they are applied after the normalized frame. Adds
Orientation::inverse and Orientation::map_norm_rect.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- color: RgbSpace::to_xyz/from_xyz and bradford() no longer expect a
matrix inverse. Degenerate (collinear) primaries, which readers reject
before building a space from a file, leave the primaries unscaled /
give the identity; the constant Bradford and OKLab matrices fall back
to the identity, which can't happen (round trips are tested).
- pipeline geometry: sample() matches (mode, warp) instead of
expect("warp"); a warp plan without a warp (sample_plan never makes
one) samples through the equivalent affine.
- color, geom, raster, develop, scenes, preview and pipeline now #![deny]
unwrap/expect/panic/unimplemented/todo/unreachable; clippy.toml as in
the steering PR.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
aiongg