- 3.1 to 3.4: the six library crates and the craft-ui facade, with a dependency
diagram, the crate table and the craft-win32 unsafe exception; taxonomy covers
every crate.
- 2.1 and 2.3: the gallery profile, cl and the story renderer.
- 3.6: layout memory's kept text; the scroll extent lives beside the offset.
- 3.7: the aiongg/egui pin, its 19 commits and what landing needs.
- 7.1: the step rule of closed list controls (Entry::stepped).
- 8.2: the gates that exist.
- 9.7 to 9.9: measured series size, landing rows without the pasted step rows,
the facade and API map in place of deprecated aliases, aliases as a question.
- Appendix B: entries for craft-behavior, each library crate and craft-win32;
where each concept lives.
The generator parsed crates/ui/src/lib.rs, which is a re-export facade over
the six library crates, so it found none of the shared modules. It parses
craft-ui-contract as craft_ui and craft-ui-widgets' layout_components, and
reads a path into any library crate as the craft_ui path the facade gives it.
The surface and story layout tables name the library's types under
craft_ui, but since the split a type's crate target is the crate that
declares it, so no library row matched: Slider's value groups lost their
form rows and every library type sat on the panel surface. The table key
reads a type of any library crate as craft_ui, as the level rows already
read a library module path. The stale-row check runs in the craft-ui
facade's trials, where every crate of the library is linked, and checks
the library's rows whenever one of its crates is.
A record of layout memory holds a measure or a text. emit_text hands a
text forward through the same transcript as emit_record: a body that
takes its kept measurement gives the text its measuring run found, where
the content that found it does not run, and a real run's text other than
the kept one shows the transcript changed. The gallery card emits each
story's value text in its shell and the form rows read it beside the
story, so a settled Slider card runs one pass per frame.
AppShell::content_height sizes an embedded shell from its body's measure,
which its body keeps while its inputs hold. The body pane is a pane at its
rect as tall as its content, taking its height in the scope's flow. The
gallery card drops its two measure_in runs and its detached shell record;
the value text and the dialog record come from the shell's real run, and a
story whose shell showed a dialog gets the card's whole room.
The extent that bounds a body's offset lived in shown memory (`BodyView`),
which is evicted soon after a body stops showing, while the offset and its
anchor live in session memory (`ScrollState`). A body shown again after a
longer hide clamped its kept offset against an extent of 0 and came back
at the top. The extent now lives in `ScrollState` beside the offset, with
the same lifetime; `BodyView` keeps the view rect.
The anchor rule runs only in the run that lays the content out for the
view: a measure run places no Lazy children, so it would clear the anchor.
A Lazy child no pass showed yet takes the estimate its app gives
(`Lazy::new(key, estimate)`): that long along the axes its scroll area
scrolls, nothing across them, as a collection places rows it never built
by its item layout's estimate. The never-shown "always run" branch of
`lazy::decide` is gone. Gallery cards estimate their height from their
story rows.
A scroll position the framework holds is also named by content: the
first child at or after the view's start (`lazy::Anchor`, kept with the
offset in `ScrollState` and a collection's `Scrolled`). When extents
before it change, the offset follows the child and the pass runs again,
so the content in view does not move. Bound offsets keep no anchor.
Entries of the out-of-view store record their scroll area. A frame drops
only the entries an area that ran did not reach, so a hidden area (a tab
switched away) keeps its children's extents.
Trials: first show, shown again after a hide, content in view stays when
an estimate above it resolves, collection rows stay when a row above them
changes height; the gallery trial asserts each level first runs the cards
in view.
cl test failed for craft-ui-panels, canvas, patterns and shell: their canonical set named craft-ui-contract/snapshot, and cargo lets a command name a feature only of a direct dependency, which the runtime is not for crates above widgets. craft-ui-contract's testing now includes snapshot (the resolved feature set of the runtime in a test build is the same three features as before, so no unit changes), and the canonical sets drop the snapshot entries. cl doctor resolves every crate's canonical set against cargo metadata, with no build: a plain feature must be the crate's own, a dep/feature must name a direct dependency and a feature it has. docs/development.md states what CI covers (ui.yml for the seven ui crates, theme.yml, docs.yml, path-filtered workflows for some leaf crates) and what it does not (the domain crates on the framework, the gallery, the renderer, the umbrella crate, any workspace-wide build).
An arrow's step is read from the focused node's number: Increment when it rose,
Decrement when it fell, and the arrow's own direction only when the node reports no
number on both sides of the key. The step counts only when it was declared before
the key, the state an AccessKit user acts from. A Select holding its first entry
(story select_first) steps forward on ArrowDown, which is its Increment.
A step over a closed control stops on the enabled, stepped entries only. A value on
none of them (mixed, or an entry only the open list offers) has no place: it reports
no number, so both steps are offered, and a step forward lands on the first stop and
a step back on the last. Select and ComboBox step with the shared step_stops and
step_stop.
LutPicker marks Load 3D LUT…, the folder files and the loaded LUT as not stepped, in
place of its filter on step events. The loaded LUT then offers both steps, which the
craft-color activation suite asks for.
`Composition::aaf_mob_id` is documented as a read identity: writers generate
fresh format identifiers and never preserve it. The roundtrip test compared
whole compositions after the second write, so it broke as soon as #1098
exposed the field. Compare them with the id cleared, as the surrounding
comment already described.
craft-tiff caps chunk descriptors at the file's length (storytold/craft-libs#827), so a grid of tiles sharing one payload lists fewer chunks than the image declares. Comparing that capped list against the grid refused a file shorter than its tile count for "missing chunks" instead of for the data it actually holds. Compare the offsets table, which is the source of truth, and cover the short-file case.
* craft-drawingml preset: the extra outlines dispatch on a Name enum
Ported from storytold/deckcraft#739 by @Bryson14.
* craft-drawingml preset: the 3- and 4-bend elbow and the curved connectors follow presetShapeDefinitions
Ported from storytold/deckcraft#1851 by @Hamza0101.
* craft-drawingml warp: text on a guide sets each glyph down whole
Ported from storytold/deckcraft#1799 by @leobasilio.
* craft-drawingml preset: the Name dispatch test's digests as this repository builds the outlines
The digests in storytold/deckcraft#739 were taken on DeckCraft's fc35ad67. Here two curved arrows
digest differently both before and after the refactor (the preset sources are identical), so the
test now holds the values from craft-libs 2b35d243, before the refactor.
* craft-drawingml preset: every preset's text rectangle from presetShapeDefinitions
Ported from storytold/deckcraft#650 by @NikitkaRa. Its preset::defaults is #1851's
adjust_defaults, which this repository already has.
* craft-drawingml custgeom: connection sites (a:cxnLst) read in the shape space and written back
Ported from the format part of storytold/deckcraft#1748 by @jagjordi.
* craft-drawingml color: behaviour-preserving cleanups
Ported from storytold/deckcraft#1610 by Cédric Legrand.
* craft-drawingml: resolve current-main API and test conflicts
* craft-pptx: use DrawingML writer defaults for custom paths
* craft-drawingml: strengthen geometry regression checks
---------
Co-authored-by: Bryson Meiling <bryson.meiling@gmail.com>
Co-authored-by: Hamza <mail.hamzanaveed@gmail.com>
Co-authored-by: Leonardo <leobasilio@gmail.com>
Co-authored-by: Nikita Rakitin <nnnrrr93@gmail.com>
Co-authored-by: Jordi <jordiag@kth.se>
Co-authored-by: Cédric Legrand <cedric@legrand-tech.fr>
* prepress: printer's marks
Marks::new lays out crop marks, bleed marks, registration targets and
process color bars around a trim box, with the sheet that holds them and
room for a page-information line. Marks::draw paints them onto a krilla
surface: lines in the registration color (Separation /All), bars in
DeviceCMYK. ISO 32000-2 14.11.3, 14.11.2, 8.6.6.4.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* prepress: keep the info line clear of the bottom target
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* prepress: clear targets only when drawn; cap line width at a third of the mark
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: KenM76 <KenM76>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>