Main now carries #9031 (Halftone) and #8923; this branch keeps only what it adds on top: the guard for a setting beside a Gallery stack's effect list, its test, and the PSD open warning for patterns that paint nothing.
Main reorganised the source tree (command modules under engine/src/cmd, tools and dialogs under ui-egui/src, large files split into modules, inline test modules moved to their own files). This branch's changes to the moved and split files are ported to their new locations.
The per-pixel reference mixed a type layer's fully covering pixel through
the type gamma round trip, which leaves white one float step short of no
ink (1 - 0.99999994). Vivid Light jumps exactly there (a source at full
brightness gives full brightness, anything below over no brightness gives
none), so a white glyph clipped onto a Vivid Light group over magenta
became full magenta in the reference, while the compositor (which keeps
covered pixels exact, like normal_gamma) gave no ink. The reference now
treats a fully covering source like the compositor and the identity it
is: the source's ink exactly.
Adds a named regression test with the shrunk document and a neighbour
test over Vivid/Linear/Pin Light, Hard Mix, Color Dodge/Burn, Normal,
Multiply and Screen with type or pixels clipped, plus the proptest seed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A Pass Through group composites its children straight into the
backdrop, so an Invert inside it changed the layer beneath even when
the group had Blend If. Photoshop renders such a group isolated, as it
does below 100% Fill: the backdrop stays (64,128,192) under a group
holding an Invert with Red Underlying white at 110. The CPU and GPU
pass-through shortcuts are now skipped for a group with Blend If.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Blend If on an adjustment layer judged This Layer by the composite after
Opacity and the mask had mixed the result into the backdrop. Photoshop
judges the adjustment's result itself: Invert at 40% over (64,128,192)
with Red This Layer black 150 keeps the effect, (115,128,140), as
Invert's red 191 passes 150. The CPU and the GPU now apply the
adjustment to the backdrop and judge that.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Gray was Rec. 601 luma, unrounded, so mixed 8-bit colours near a hard
Gray point landed on the wrong side: (123,126,151) read 127.95 and was
hidden by a black point at 128. Photoshop's Gray, measured on pure
channels at three cutoffs, is (30R + 59G + 11B) / 100 rounded half down
on whole levels; it keeps both of the reporter's colours.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore: resolve SonarCloud maintainability issues
* fix: format imports and fix effect_entry argument type
* fix: restore test closure and wildcard import that broke compilation
* refactor: split composite_to_surface to reduce cognitive complexity
A hard Blend If point at level c sits on Photoshop's 16-bit step
round(c * 32768 / 255), inclusive for black and white points (measured
at 1, 64, 128 and 200). blend_if_level snapped only within 1e-4 of a
whole level, so RGB16 32895 (level 128's step) read 127.996 and was
hidden by a black point at 128. A value on its nearest level's step is
now that level, on the CPU and the GPU; 8-bit levels are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Below 100% Fill, Photoshop blends Color Burn, Linear Burn, Color Dodge,
Linear Dodge, Vivid Light, Linear Light, Hard Mix and Difference with
the source pulled toward the mode's neutral colour, at Opacity. Pixel
layers already did (psblend::composite_fill); adjustment layers used
Fill as coverage on the CPU (blend_back) and the GPU (AdjMix), so a
40% Fill Color Dodge gave (90,179,217) instead of (80,160,240).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adjustment layers (compose blend_back), destructive edits
(mix_through_selection) and plug-in filters mixed their result back as
o + (r - o) * k. In f32 the difference had already rounded a small
result away against a huge backdrop, so a Color Lookup's 0.1 over 2^20
came out 0.125, and 0.5 over 1e8 came out 0. They now compute
o * (1 - k) + r * k, which is exactly r at full coverage.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The feather blur clamped every one of its box passes at the plane's
edge. Where the canvas cut a feathered mask off, each pass repeated the
already blurred values there, brightening the edge (12 levels at a
32 px feather) and making the result depend on the canvas size.
The blur now pads the plane with its edge values by the passes' reach,
blurs, and keeps the original area, which matches Photoshop within one
level both for a mask black at the edge and one white up to it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Merge #9104 (raw EngineData strings are written back as the same bytes) so a style-only type edit does not re-encode Photoshop's raw strings as UTF-16, with tests for the numbered and the legacy restyle paths. The over-budget pattern warning is written without a panic path and explained where the lock is held.
effect_maps_with_nested_rayon_work_do_not_deadlock rendered its document
101 times and failed when the whole run took longer than 120 s. On a
machine busy with other builds the renders slow down far enough to pass
that limit while every render still finishes, and the test then fails
with "effect-map rendering deadlocked (#276)".
The render thread now sends a message after each render, and the test
fails when 120 s pass without one. A render that never finishes still
fails the test after 120 s.
A pattern past the per-call budget is converted under one lock and shared while in use, with a one-time warning, and the desktop budget keeps every pattern a call uses (1 GiB). The web build caps a pattern's working copy at 128 MiB and a call at 256 MiB, and the pattern readers reserve their memory fallibly on every target. The legacy text restyle writes and compares BaselineShift, FontCaps, FontBaseline, UnderlinePosition, StrikethroughPosition and the ligatures per run, keeps an automatic leading's own value, and import reads the legacy AutoKern; a corpus test edits the size in psd-tools' text.psd and checks the rest of the legacy object is unchanged.
Align measures shape extents at the document's own sample type, a render keeps two of the largest patterns converted per call instead of rebuilding the second for every tile, a pattern too large to paint is named by every command that would use it, by pattern.list and by a PSD open warning, and a style-only edit restyles the older Photoshop's legacy text objects too, checked against the real psd-tools text.psd.
Return the pattern code to the state before the earlier 8000 x 8000 attempts. A pattern that fits the compositor's 256 MiB cap is expanded as before; a larger 8-bit or 16-bit one is held as straight RGBA8 or RGBA16, exactly its values at 4 or 8 bytes a pixel, so Define Pattern's 8000 x 8000 paints at 8 bits. Define Pattern refuses an area the compositor could not hold with its limit, patterns whose RGBA32F texture passes 256 MiB paint on the CPU, shape bounds for Align keep full alpha precision, Define Pattern judges opacity on the stored samples, and a Halftone Gallery stack with an unknown setting beside its effect list stays verbatim.
Define Pattern accepts 8-bit patterns up to 8000 x 8000: an 8-bit pattern is held at 4 bytes a pixel by the compositor and the GPU, read in bands, and what is paintable follows the working-copy bytes. Shape bounds for Align keep full alpha precision, Define Pattern judges opacity on the stored samples, and a Halftone Gallery stack with an unknown setting beside its effect list stays verbatim.
Main split photocraft-cli into modules, so PR #8807's skip of Deselect with nothing selected is applied in document.rs and batch.rs, with its two tests in tests/it/cli.rs.
The CMYK proptests failed on main now and then (the_split_route_agrees_with_the_exact_path
after #8663). Three gaps, each found by the proptests at 5000 cases:
- cmyk.rs: a clipped Dissolve layer blended like Normal; it now shows each pixel fully or not
at all by the compositor's noise, as the RGB compositor does since #3052.
- cmyk.rs: a Dissolve adjustment layer mixed by its opacity; opacity and mask are now the chance
a pixel takes the adjustment, as on the RGB compositor (#3050).
- native.rs: the display/Flatten ink route didn't round to the document's levels after an
adjustment layer, where the RGB compositor and cmyk.rs do; coverage in a pass-through group
then differed by up to half a level.
The test reference model had the same two Dissolve gaps (ref_clip passed no position, so
Dissolve fell back to Normal), which is why random_documents_match_the_reference never caught
the first one. Three named regression tests cover the minimal cases.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(curves): 32-bit Curves continue past 0..1 like Photoshop
In a 32-bit RGB document an edited Curves clipped every sample to the
curve's table, so HDR values above 1 (and negative ones) were flattened.
Photoshop extends the curve beyond its end points along its end segments:
the straight curve [[0,0],[255,128]] maps -0.5, 0.5 and 2 to
-0.250980407, 0.250980407 and 1.003921628 (measured in Photoshop 27.11 on
the issue). The CPU compositor, Image > Adjustments and the GPU kernel
now do the same, with the end slopes measured over 16 table steps.
8/16-bit documents and unedited curves are unchanged.
Fixes#7157
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(curves): extend 32-bit curves in RGB documents only
The curve's space is not the document's: a grayscale, CMYK or Lab
32-bit document with an RGB-space curve keeps the table's ends. The
document mode now reaches the decision on the CPU (apply_in, the
compositor) and the GPU (adjustment_program).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The grey was clamped to 0..1 at every depth, so a 32-bit (2, 2, 2) came
out as (1, 1, 1). In 32-bit documents it now clamps to 0..16, as
Photoshop 27.11 does; 8 and 16-bit still clip at 1. CPU and GPU both.
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(blend): Soft Light uses the W3C curve over dark backdrops (#8804)
Photoshop's Soft Light moves the backdrop towards D(cb) in the upper half,
and D(cb) is the cubic ((16·cb − 12)·cb + 4)·cb below a quarter, not
sqrt(cb). With sqrt the Soft Light cell of psd-tools rgb-blend-modes.psd
was up to 20 levels too light over the dark rainbow; with the cubic it is
within 4 of Photoshop's merged image.
Same change in compose.wgsl so the GPU compositor stays at CPU parity.
Krita's own soft_light (the sqrt curve) now opens as the approximation and
svg:soft-light / krita:soft_light_svg as the exact match, without warning.
Fixes#8804
* compose: the see-through Soft Light test expects the W3C curve over a dark backdrop (#8804)
The #6232 test computed Photoshop's Soft Light of yellow as sqrt(b) for red and
green. Its backdrop #04170F is below a quarter in both channels, where the W3C
curve (Photoshop's, measured in #8804) uses ((16b - 12)b + 4)b instead.
* security(compose): bound dense working sets and repeated group evaluation
Smart-filter mixing and stroke fitting work tile by tile instead of over the
union of sparse surfaces; render_layer admits its rect like render; hidden
layers no longer disable the proxy and uniform-area shortcuts; Blend If
groups nested deeper than two are judged by their result, so nesting costs
polynomial instead of exponential work; effect builds past an absolute byte
cap are skipped.
* CodeRabbit #8949: exact tile skips in filter_mask, allocated-tile scan in fit_stroke, contrasting backdrop in the nested-group test
- filter_mask mix/unmix: bare tiles (no tiles on either side) skip for
non-alpha formats too when the two surfaces' default pixels are equal;
differing defaults still get processed, so the clone's default pixel is
never left wrong under a non-unit mask
- shape_split fit_stroke: scan the union of the three surfaces' allocated
tiles (exact skip when all three defaults are transparent) instead of
every tile of a 300k-canvas document; non-transparent defaults fall
back to the full scan
- tests: the nested-groups assertion now uses a black backdrop so the
asserted white can only come from rendering through the group chain
fixes CodeRabbit review comments (major/minor) on filter_mask.rs:27,29,
shape_split.rs:126, tests.rs:2354
* compose: skip a BTreeSet in fit_stroke — a sorted Vec compiles under the wasm toolchain (E0599 on ::drain there)
Two places turned an RGB or Lab colour into ink with the built-in coated
CMYK instead of the document's embedded profile: the exact ink path's
tiles (render_tile runs on rayon workers, which never inherit a caller's
with_cmyk_space scope), e.g. a solid fill picked in RGB, and the split
path's stand-ins (fill and effect colours, converted on the calling thread
outside any scope). Both now enter cmyk_space(doc). A test with a synthetic
uncoated profile checks a solid fill and a Color Overlay get the profile's
inks; each half fails on its own without its fix.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>