94749b4f3e1406884ac1e50b68409b436aa28322
341
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
  adminandClaude Opus 5.5
|
94749b4f3e |
Let a project keep named snapshots and restore them (#231)
Autosave only keeps the latest revision. Before a risky change there
was no way to put a known good state aside and go back to it later.
The main menu now has "Snapshots..." to create, restore and delete
named copies of the open project.
- Server: SnapshotStore.php keeps each snapshot as a copy of the
project JSON under data/snapshots/<project>/ (git-ignored). Four
routes under projects/{id}/snapshots use the same owner check as the
project, so an admin also sees foreign snapshots.
- Create writes the autosave first. With unsaved edits (play with open
edits, failed save, conflict) no snapshot is made and a message says
why. The rule is the pure function in src/editor/snapshots.ts.
- An empty name falls back to the local date and time.
- Restore asks first, then keeps the current state as "Before
restoring <name>". It writes the snapshot back as a new revision and
reloads the project, so the undo list is empty. A restore can itself
be undone by restoring that snapshot.
- Deleting a project deletes its snapshots too. Images, sounds, fonts
and pixel art are content-addressed and never deleted, so a snapshot
can always reach them.
- Client: SnapshotStore and HttpSnapshotStore in src/core/store, the
SnapshotsDialog, and the flows in ProjectSession. A stale revision
(409) becomes StaleRevisionError.
- CI runs the new PHP test in the gate job next to accounts.
Not covered by e2e: the lock during play and the conflict case. Both
are unit tests only.
Tests:
- check.sh: 164 files, 2401 tests, CHECK EXIT 0.
- tests/php/snapshots.test.php: 14 checks, all passed.
- tests/editor/snapshots.test.ts: 5 tests for the time format, the
blocker rule, the routes and 409.
- New e2e module 50 on a throwaway project: create, rename, restore,
undo the restore, delete, then delete the project and get 404 on
its snapshots. 46b expects the new menu entry. e2e 46,50: 20 checks
passed. The check sum goes from 1385 to 1393 (computed).
- Counter-checks, originals restored and compared with cmp: blocker
always allowed fails 1 unit test. No sorting fails 2 PHP checks.
Restore without keeping the old state fails 50c and aborts 50d.
- The first sort test with 2 snapshots passed by chance half the time
because of random ids. It now uses 8.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
||
|
adminandClaude Opus 5.5
|
17a6fd0339 |
Move the shader test sprite off a name plate so CI passes again
Every suite run since |
||
|
adminandClaude Opus 5.5
|
21d7149756 |
Let a nested behavior without a start node run from On start (#230)
A behavior built as a plain behavior, with On start and no Behavior start, did nothing when another behavior called it through Run behavior. The node only showed a warning. Such a behavior should run as a call too, without the author having to rebuild its entry. - New callEntries() in nested.ts returns the Behavior start node or, when there is none, every On start node. callProblem uses it too. - GraphInterpreter.call() follows all entries one after the other, in graph order, in the same frame and with the same variables. - Behavior end still returns what reaches it. A call through On start has no inputs, because only the start node declares them. - It is a fallback only. With a Behavior start present, On start nodes stay out of the call. - The node warning now reads "No start or On start node - does nothing" and only shows when neither exists. - Keep and local state machines behave as before. Tests: - ./check.sh: 163 files, 2396 tests, CHECK EXIT 0 (before 2392). - 4 new unit tests in tests/core/nested-behavior.test.ts: the flow after On start runs, the return value arrives, several On start nodes run in graph order, a start node keeps On start out. - e2e 19: ALL PASSED, 35 checks. Only the expected warning text changed, the check sum stays at 1385. - Counter-checks: without the fallback 3 tests fail. With only the first On start exactly the order test fails. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
454d202109 |
Let each variable of a behavior choose whether calls keep it (#229)
The first version of #229 put a Keep state switch on the Run behavior node. The user wanted it where the variable is declared instead. Whether a value lives on belongs to the variable, not to every call site. - The node switch keepState is gone again. RunBehavior, CallResult and nested.ts are back to what they were before #229. - BehaviorAsset.keptVariables lists the kept names. It lives on the asset and not in PropertyDef, because that is the schema of every inspector row. - A variable added from now on starts with Keep on. Variables in existing projects start with Keep off, so old projects run unchanged. - SetVariableKeptCommand flips the switch as one undo step. Add, remove and rename carry the list along. Loading drops names that are not declared. - The interpreter keeps only the listed variables, per Run behavior node and per object as before. The others start at their defaults. - Local state machines in a called behavior start empty on every call again, as before #229. Global and scene machines are unchanged. - The inspector shows a Keep button in each variable row of a behavior. Its state is part of the asset signature, so a click redraws the row. Without that the button kept its old look until the inspector was reopened. Found: the project Garden of Sacrifices still has keepState = true on the call Player -> Character Movement. It is ignored now. Direction there predates the switch, so Keep is off until set by hand. Tests: - ./check.sh: 163 files, 2392 tests, CHECK EXIT 0 (before 2388) - tests/core/keep-state.test.ts rewritten, 15 tests - e2e 19: ALL PASSED, 35 checks. 19c-ii now clicks the Keep button of a new variable: on when added, off by a click, on again, off after undo. - Counter-checks, each fails its own test: keep list ignored (2), states kept (1), new variable not kept (8), rename drops the list (1), no Keep button (e2e 19c-ii). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
7c70245c79 |
Let a nested behavior keep its variables between calls (#229)
A Run behavior call started with fresh variables every time, so a nested behavior could not count, remember or keep a local state machine across calls. A new "Keep state" switch on the node lets the callee resume where its last call left off. It is off by default, so existing projects run unchanged. - RunBehavior gets the boolean parameter keepState and its node description explains it. - GraphInterpreter keeps the state in keptCalls, keyed by the call path, the chain of Run behavior nodes. Two nodes calling the same behavior each keep their own state. - One interpreter exists per behavior instance, so every object gets its own state without further code. - Local state machines of the callee are kept like its variables. - resumedVariables() matches kept values against today's declarations. A new variable starts at its default, a removed one drops out, and one whose kind changed starts over. - The kept state lives as long as the object in the run. It is not written to save games. Limit: a callee input named keepState collides with the switch. The same already holds for the name behavior, so it is not caught. Tests: - ./check.sh: 163 files, 2388 tests, CHECK EXIT 0 (before 2377) - 10 new unit tests in tests/core/keep-state.test.ts and 1 new node catalog case. - e2e 19: ALL PASSED, 35 checks, with the new check 19c-ii for the checkbox, its click and undo. - Counter probes: ignoring keepState fails 5 tests, state per behavior instead of per node fails 1, dropping the outer path fails 1, skipping the kind check fails 1, and without the node parameter the e2e check fails. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
03e7a78c3d |
Keep a tilemap with outlined solids at 60 fps in scene and play
A project with three tilemap layers, two of them with collision on water and hill tiles, ran at 33 to 52 fps. A CPU profile showed that almost all of the time went into the collision geometry of the map. It was rebuilt from scratch every frame although nothing had changed. The animated water makes the views redraw every frame, so the cost was never skipped. - New module tilemapMemo keeps what is derived from the cells: the tiles of a layer, the blocking cells, the rectangles and polygons. It is reused as long as the content stays the same. - A brush stroke paints into the existing cells object. The memo therefore compares the content cell by cell, not the identity. The comparison walks two arrays in insertion order. - Clipped outlines are kept per outline list. The model always replaces that list instead of changing it. - PhysicsSystem no longer rebuilds the shapes of a tilemap each tick while its cells, placement and settings are unchanged. Before, it also built a JSON key from all world shapes every tick. - The overlaid layers after Set tile in a run stay stable. Otherwise the memo would miss on every call after the first edit. The #184 work only helped Move and raycasts, not this path. Measured on stein2 (i5-7500T, HD 630), script time per frame: - Scene, nothing selected: 17.7 ms, 52 fps -> 5.8 ms, 60 fps - Tilemap selected: 26.2 ms, 38 fps -> 8.6 ms, 60 fps - Play: 29.9 ms, 33 fps -> 8.7 ms, 60 fps About 3 ms per frame remain. The content comparison costs roughly 60 us per 700-cell layer and runs about 40 times a frame. The mesh geometries in TilemapField are still rebuilt every frame. Tests: - ./check.sh: 162 files, 2377 tests, CHECK EXIT 0 (before 2370) - e2e 1-5 and 39-43: ALL PASSED, 289 checks - 7 new unit tests in tests/core/tilemap-memo.test.ts - 5 counter checks (content ignored, no memo, outline by kind only, physics always rebuilds, layer settings ignored) each fail their test. All originals are back, checked with cmp. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
021a6dcb56 |
Zoom the tile tab in the same steps as the scene view
The tile tab zoomed freely, so the pixels of a tile could end up
unevenly wide at odd factors. Scene view, container tab and animation
preview already snap to FREE_ZOOM_STEPS. The tile tab now does the
same, so every pixel editor zooms alike.
- The wheel steps through FREE_ZOOM_STEPS, counted in device pixels.
The board position is rounded to whole device pixels.
- New wheelSteps() in src/viewport/view.ts turns wheel events into
steps: one notch is one step, a trackpad gathers 50 px per step.
FreeViewport uses it too, so the rule exists only once.
- Fit takes the largest step at which the whole board fits. Refitting
on a new tab or board size stays as before.
- panAndZoom takes an optional zoom rule instead of limits. Graph and
node board keep zooming freely.
- A saved view from before the steps snaps to the nearest step.
- The cap parameter of frameBoxes from
|
||
|
adminandClaude Opus 5.5
|
4027afc228 |
Let Fit fill the tile tab instead of stopping at 100 percent
Fit in the tile tab left the board small in the middle of the tab. It used frameBoxes, which never zooms past 1 so that two graph nodes do not blow up to fill the screen. The tile board is measured in image pixels, though, and with 16 px tiles it is only about 345 x 186 px. - frameBoxes takes a new cap parameter, default 1. The graph and the node board keep their behaviour. - The tile tab passes its zoom limit of 6, so Fit fills the tab up to 600 percent. - TileCanvas remembers whether the view is still the one Fit set. While it is, a change in the size of the tab or of the board fits again. After a pan or zoom by hand the view stays until Fit. Found on the way: switching the image to another tile size (48 to 16 px) kept the old zoom, and the board shrank to a third. The refit above covers that too. Tests: - ./check.sh: 161 files, 2370 tests, CHECK EXIT 0 (was 2369). - New unit test in graph-framing: with a raised cap Fit fills the view up to the limit. The graph test for "not past 100 %" holds. - e2e 39: ALL PASSED, 31 checks (was 29). Fit fills the tab at 48 px (zoom 0.41) and fits again after the switch to 16 px (zoom 1.24). - Counter probes: a cap of 1 and no refit on a new board size each fail exactly the 16 px check (zoom 1 and zoom 0.41). - e2e sum in the README is 1384, computed, not yet seen in CI. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
50bddf68d4 |
Keep scene view names readable on plates and in rows of their own
Symbol names in the scene view were plain grey text. On a background of a similar colour they could not be read, and names of symbols close together were drawn over each other. - Each name stands on a rounded, nearly black plate at 70 % opacity, 3 px padding left and right, 1 px top and bottom. - The text is light (viewportLabel) instead of muted grey. - The plate of a selected object gets an accent border. - New rule stackLabels in src/viewport/labelRows.ts. A name that overlaps another name or an icon moves one row down, then one up, then two down and so on, until a row is free. - Names are placed top to bottom, so the upper one keeps its row. A name only moves vertically and stays on the side of its icon. - Names outside the view still claim their row. Otherwise the visible names would jump while panning. - New hook __gm.sceneLabels() for the e2e check. The frame names in the texture editor already had a plate, so they are unchanged. Not covered by tests: the accent border on selection. Readability on a light background was only judged from a screenshot on a dark one. Tests: - 5 unit tests in tests/viewport/label-rows.test.ts: free names stay, the lower of two colliding names moves, the stacking order, names avoid icons, 30 crowded names never overlap. - e2e 5c-ii: 8 symbols on one spot give 8 plates, no overlap with each other or with icons, each name on the side of its icon. - Counter-checks: without the overlap test in the rule 4 of 5 unit tests fail. With the scene view ignoring the rows exactly 5c-ii fails. - ./check.sh: 161 files, 2369 tests, CHECK EXIT 0. - e2e 1-5: ALL PASSED, 140 checks. The e2e sum is now 1382 (computed). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
af6ac12ab5 |
Show sharp tile preview labels and paths with tunnels and crossings
The mask labels under the tile preview sat inside the zoomed board. Pixi rendered them at 10 px and stretched them with the zoom, and each label was shrunk to the width of its tile on its own. They came out blurry and in different sizes. The example map also only showed a ring, so narrow paths and crossings were never previewed. - The labels live in their own layer in screen space and are re-rendered at the fitting font size on every zoom. - All labels share one size, at most the 10 px of the theme. Below 6 px they are hidden. - A label has two lines: the sides on top, the corners dimmed below. - The example map keeps the ring cell for cell and adds a network of 1 tile wide paths below it. It has straight 1xN and Nx1 pieces, curves, two T junctions, a crossing and a dead end. - The map stands next to the grid in a wide panel and below it in a tall one. Fit picks the layout with the larger tiles, and only Fit changes it, so zooming does not make the board jump. - The new hook __gm.tile.captions() reports size, scale and lines. Found on the way: the taller map made Fit zoom out further, and 3 pixel checks on the 16 px terrain tiles failed. The side by side layout fixed that. Tests: - ./check.sh: 160 files, 2364 tests, CHECK EXIT 0 (was 2363). - e2e 39: ALL PASSED, 29 checks (was 28). - New unit test: the example map has straights, a curve, a crossing with four notched corners, a T junction and a dead end. - Counter check: with the labels back in the zoomed board the new caption check fails (scales [0.648, 1]). - The e2e sum in the README is now 1381. It is calculated only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
5dbac3a8df |
Add a Select node that picks one of two values by a condition
Choosing between two values meant a Branch with the whole action twice behind it, for example two Play animation nodes that differ only in the animation. Every later change had to be made in both. Select picks the value, so the action stays single. - Select is a data node with the inputs condition, false and true and the output value. It passes on "true" if the condition holds, else "false". An unconnected condition counts as false, and an empty picked input gives null. - The ports are named false and true, not a and b, so nobody has to guess which input is which. The order false, true follows Unreal's Select and Blender's Switch. - false, true and value share one type through `sharedType`, like First not empty. With two Animation literals plugged in, value takes their type and fits Play animation directly. - Both inputs are always evaluated. Data nodes have no side effects, so that does not matter. Tests: - ./check.sh: 160 files, 2363 tests, CHECK EXIT 0 (before 2356). - 5 cases in tests/core/node-catalog.test.ts and 2 in tests/core/shared-port-type.test.ts. - Counter checks: with the inputs swapped all 5 catalog cases fail, without sharedType exactly the type test fails. - No e2e change, the sum stays at 1380. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
89048413f1 |
Let a texture frame set where a sprite stands for Y sorting (#228)
A sprite sorted by Y stood on the bottom of its box. With a shadow or empty rows under the feet that is too low, so a figure slipped in front of a tree it should stand behind. The line belongs to the frame, not the texture, because one texture can hold several sprite sheets. - New field `sortLine` on a frame: whole image pixels up from the frame's bottom edge, 0 to the frame height. 0 is stored as nothing, so old projects stay the same. Shrinking a frame keeps the line inside it. - `feetOf` in drawOrder.ts lifts a sprite by the same share of its box. Scene view, game view, play and picking all use it. In an animation the line of the frame shown right now counts. - The texture tab draws the selected frame's line dashed in yellow with a knob left of the frame. At 0 the line shares the bottom edge with the nine-slice border and a resize handle, so the knob is tested first. A line inside the frame can be dragged directly, tested after the nine-slice lines. - The inspector has a new "Sorting" section with a "Sort line" field. Dragging or typing is one undo step each. Limit: frames of one walk cycle need the same value, set by hand. Otherwise the figure changes its place in the order while walking. Tests: - ./check.sh: 160 files, 2356 tests, CHECK EXIT 0 (before 2350). - 6 new unit tests in tests/viewport/sort-by-y.test.ts. - e2e 37 and 39-43: ALL PASSED, 162 checks. New check 37b-ii drags the knob up 6 px and undoes it, plus the inspector row. - Counter checks: if feetOf ignores the line, exactly the 3 order tests fail. If the knob cannot be grabbed, 37b-ii fails. - The e2e sum in the README is 1380, counted, not yet seen in CI. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
50854389bb |
Let a tilemap layer sort its rows by Y with the objects (#228)
A tilemap layer was one piece in the draw order, so a figure stood either in front of a whole layer of trees or behind all of it. For a top-down game the trees in front of the figure's feet have to cover it and the ones behind must not. - A new "Sort by Y" box under the layer list sets `sortByY` on the painted layer. Every change is its own undo step. - Such a layer comes as one draw slot per row. On its `z`, the rows and all objects on that `z` are sorted by where they stand: the bottom of an object's box, the bottom of a row. Ties keep the hierarchy order, and other `z` values are untouched. - A layer drawn as one piece and the map's own node stay behind the sorted ones on the same `z`, since neither has a single place to stand. Ground belongs on a lower `z` or stays unsorted. - A new tile property "Sort offset" (whole rows, -16 to 16) lets a crown sort with the trunk below it. - `TilemapField` draws a container per row for such a layer, and picking takes the row into account, so a click picks what is drawn in front. Scene view, game view and play share the same rule. Limits: on a rotated map "down" is world Y. A layer sorted by Y costs about one draw call per row and tile instead of one per tile. Tests: - ./check.sh: 160 files, 2350 tests, CHECK EXIT 0 (before 2340). - 10 new unit tests in tests/viewport/sort-by-y.test.ts. - e2e 39-43: ALL PASSED, 148 checks. New check 40p-iii ticks the box and finds the slot of row 0, undo brings back the whole layer. - Counter checks: without the Y sort 5 unit tests fail, with the offset ignored 5, with no row in picking 1. If TilemapField draws no rows, exactly 40p-iii fails in e2e. - The e2e sum in the README is 1378, counted, not yet seen in CI. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
ffb5bbe039 |
Stop a key getting stuck when Shift changes it while held (#227)
Input remembered keys under KeyboardEvent.key. Pressing a, adding Shift and letting go sent "A" on release, which matched nothing, so "a" stayed down until the next release without Shift. - New rule keyName in src/core/keys.ts lowercases a single character. So a and A are the same key, also under Caps Lock. - Input remembers per physical key (e.code) under which name it went down and releases exactly that name. So 1 released as ! is freed too, which lowercasing alone does not cover. - On key down, up and held, Key is down and Axis input compare with the same rule. A node that already stores A reacts to a. - Recording a key in the inspector stores Shift+A as a. - Limit: holding Shift before pressing 1 still sends !, and a node on 1 does not react. Nothing gets stuck by it. Tests: - 3 new unit tests in tests/core/behavior-events.test.ts: the flow from the issue, 1 released as !, and a node on A hearing a. - Without lowercasing exactly the issue test and the A test fail. - Without the code memory exactly the 1/! test fails. - ./check.sh: 159 files, 2340 tests, CHECK EXIT 0 (before 2337). - No e2e change, the check sum stays at 1377. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
c1a6d63edb |
Cover which tilemap layer and texture frame get a collision fill
|
||
|
adminandClaude Opus 5.5
|
8f29b88cc9 |
Make the actor smoke reach every sub pixel edge position again
Since #225 the game view opens at 100 % instead of fit. The actor walks 150 px/s, so its top edge only lands on .0 or .5 of a pixel. The check for the frame top edge needs 0.6 to 0.9 and gave up in 6 of 7 CI runs. The picture itself was fine: the pixels above and at the edge were the same in every red run. - While nudging the actor into place the smoke slows time to 0.08 with __gm.setTimeScale. The actor then moves 0.2 px per frame and can reach any position inside a pixel. - Time scale goes back to 1 before the edge is measured. Tests: - Old smoke on the same seeded project fails locally like CI (fraction 0.08). - New smoke passed three times in a row (fraction 0.80, 21 checks). - The actor project file was saved first and restored byte exact. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
f4cc681df4 |
Fill texture and tilemap collision only where it is being edited
The faint fill on every frame outline and on every tilemap made the scene and the texture editor look green all over. Like the collider before, the fill now shows only on what is selected, and a bit stronger, so it marks the thing being worked on. - Texture collision: all frame outlines stay visible. Only the outlines of the selected frame are filled, at alpha 0.3 instead of 0.15 on every frame. - Tilemap: the outline still shows where the whole map blocks. The fill appears only while the tilemap alone is selected, and only for the part the layer chosen in the panel adds. Solid is 0.3 and one-way 0.15, before 0.12 and 0.06 on every map. - New query layerSolids(layerId) in items.ts. It builds the same solids for one layer and runs only for the selected map. - The small landing wedge on one-way edges stays filled. It belongs to the edge marker, not to the area. Not covered by tests: the texture canvas and the scene view wiring that picks the map and the layer. Neither was looked at in a browser. Tests: - ./check.sh: 159 files, 2337 tests, CHECK EXIT 0 (before 2334) - e2e 1-17,39-43: ALL PASSED, 668 checks - 3 new unit tests in tests/viewport/tilemap-fill.test.ts - Counter-check: with layerSolids ignoring the layer, exactly the layer test fails. Original restored, checked with cmp. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
495d87847d |
Tint a collider's area only while the collider itself is selected
A collider filled its area with a faint green in the scene view all the time. It lay over every sprite a collider sits on and made the art look washed out. The outline alone shows where a collider is. The area only matters while one works on that collider. - drawItem takes a selected flag. Only a selected collider is filled, every other one is drawn as its outline. - The fill is stronger than before, alpha 0.3 instead of 0.12, so the selected area is easy to see. The value is COLLIDER_FILL_ALPHA in src/viewport/symbols.ts. - The scene view says what is selected through isSelected. The game view keeps the default and never fills. - Only the collider object itself counts. Selecting its body leaves the area unfilled. Tests: - 2 new unit tests in tests/viewport/collider-color.test.ts read the Pixi instructions: outline only when not selected, fill at 0.3 and outline when selected. - Counter check: with the fill drawn regardless of selection the outline-only test fails. The original is back per cmp. - ./check.sh: 158 files, 2334 tests, CHECK EXIT 0. - e2e 1-17,28-30: no FAIL. e2e 39-43: ALL PASSED, 145 checks. A first run left out 39, which 40 builds on, and failed 8 checks in 40 for the missing tile only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
4b7e8614a8 |
Keep play start and game view scale in the project (#226)
Both settings lived in the browser's localStorage, so they did not travel with the project and differed per machine. They now belong to the project like the symbol scale. There is no migration, old projects load with the defaults. - New module src/core/model/playSettings.ts reads playFrom and gameScale on load. Unknown or missing values fall back to the start scene and 100%. - src/editor/settings.ts is gone. - Play starts in the project settings dialog is an undo step (SetPlayFromCommand). It is only recorded on a real change. - The game view scale is saved with the project but is no undo step, so Ctrl+Z never undoes a zoom. A scale picked during play is written to the stored project and survives Stop. - Save states of persistent Variables assets (saves.ts) stay in the browser on purpose. They are runtime data of the game, not settings. Tests: - 6 new unit tests in tests/editor/play-settings.test.ts: defaults, save and load, broken JSON, undo of Play starts, scale without undo step, scale picked during play survives Stop. - e2e 03g checks the value in the project, nothing in localStorage and no new undo step. 35f also checks project.playFrom. No new checks, the sum stays at 1375. - Counter checks: no write to the stored project in play, no read on load, and an undo step for the scale each fail exactly their test. - ./check.sh: 158 files, 2332 tests, CHECK EXIT 0 - e2e 1-17,28,30,35,38-40,46: ALL PASSED, 752 checks Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
5afb70a92e |
Let the project set the size of scene view symbols
Pixel art projects draw at very different sizes, so the fixed 18 px icons of behaviors, containers and rigidbodies were too big in some projects and too small in others. The size now lives in the project settings and travels with the project. - New field Symbol scale in the project settings dialog, below Play starts. It takes 0.1 to 4 with two decimals and defaults to 1. - The value is stored as project.symbolScale. Older projects load with 1, and values out of range are clamped on load. - Every change is its own undo step through a project command. - The factor scales the icon size, the spacing of icons that share a spot, the label offset and the selection frame of icon-only symbols. - The footprints in bounds.ts that Fit uses stay as they were. The hit area follows the icon, so clicks match what is drawn. - uiHooks reports the icon size so e2e can measure it. The e2e sum in the README is 1375. That is 1374 plus one new check, and 1374 itself has not been confirmed by CI yet. Tests: - ./check.sh: 157 files, 2326 tests, CHECK EXIT 0 (before 2320). - New tests/core/symbol-scale.test.ts with 5 tests (default, save and load, clamping, undo, icon frame), one more in project-settings.test.ts. - e2e 46f: dialog opens with 1, 0.5 plus Apply gives a 9 px icon, undo gives 18 again. Module 46: ALL PASSED, 13 checks. - Counter checks: without loading the value the save and load test fails. With the icon ignoring the factor exactly 46f fails. Both originals restored per cmp. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
5399ef17e5 |
Snap the scene view zoom to pixel exact steps
The wheel zoomed smoothly and fit computed any factor, so the scene view almost never showed pixel art with even texels. Under Windows scaling even zoom 1 was off. The zoom now only takes steps that give every texel the same number of screen pixels. - The zoom counts device pixels per world pixel. freeView divides by the renderer resolution, so zoom 1 is one screen pixel per texel. - FREE_ZOOM_STEPS: 1/8, 1/4, 1/2, then 1 to 6, 8, 12, 16 and 24. - The wheel moves one step per notch. A trackpad collects its travel and steps every 50 px. - Fit rounds down to the next step. A saved zoom from before snaps to the nearest step on a log scale. - Pan, drag, grid and resize use CSS pixels per world unit, so they still follow the pointer under Windows scaling. - Always on, no toggle, as decided in the session. Tests: - 4 unit tests: zoom 1 at 125%, whole steps, wheel steps and ends, fit rounding and snapping of a saved zoom. - Counter-check: without the ratio in freeView the 125% test fails. - check.sh: 156 files, 2320 tests, CHECK EXIT 0. - e2e 1-17,28,30,38-40: 726 checks passed on the steps before the final values (up to 8). Not rerun on the final steps. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
22a89ed0f0 |
Keep game view steps pixel exact under Windows display scaling
At 125% Windows scaling the game view drew 100% as 1.25 screen pixels per camera pixel, so pixel art showed texels of uneven width even at the round steps. The steps now count device pixels instead of CSS pixels. - gameScaleFactor, cameraFactor and cameraView take the device pixel ratio. 100% is one screen pixel per camera pixel, 200% two. - Auto picks the largest step that fits in device pixels. Fit reports its factor in device pixels too, e.g. 156% at 125% scaling. - Fixed steps start the frame on a whole device pixel. - The game view reads the ratio from the renderer resolution. Limits: main.ts caps the resolution at 2, so scaling above 200% is stretched by the browser again. A changed ratio at runtime only counts after a reload, as before. Tests: - 2 unit tests at 125%: 100% is one device pixel on a whole pixel, auto and fit count device pixels. - Counter-check: without the conversion the 100% test fails. - check.sh: 156 files, CHECK EXIT 0 on the combined working tree. - No e2e: the harness runs at devicePixelRatio 1, same math as before. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
2592873d76 |
Pull nearest tilemap uvs a tenth of a texel in, 0.02 still showed lines
At 0.02 texel the lines between tiles of a nearest tilemap were gone at the zooms tried first, but still showed at others. The user raised the inset to 0.1 by hand and checked it in the browser. - A tenth of a texel is still far below half a texel, so the texel picked at every pixel centre stays the same. - The comment names the new value. Tests: - tilemap.test.ts follows the value: the inset test expects 0.1/96, the sliver test allows up to 0.25 texel from the edge, and the quad test checks its uvs to 2 digits instead of 3. - ./check.sh: 156 files, 2314 tests, CHECK EXIT 0. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
5bf1ee7a4e |
Let Equals match a number against a typed number like 130 (#224)
The float tolerance from the last commit never reached the case it was built for. The inputs of Equals are of type any, and their inline field is a text field. A typed 130 therefore arrives as the string "130", and the tolerance only applied when both sides were real numbers. The node fell back to comparing "129.99999999999997" with "130" as text. - A number against a text that parses as a number now uses the same 1e-9 rule as two numbers. - An empty text does not count as 0. - Two texts are still compared as text, so "1.0" and "1" stay different. - Compare was not affected, its inputs are real number fields. Tests: - 3 new cases in node-catalog.test.ts: 129.99999999999997 vs "130" is equal, 0 vs "" is not, "1.0" vs "1" is not. - Counter-check: with the old rule (two real numbers only) exactly the "130" case fails. Original restored and compared with cmp. - ./check.sh: 156 files, 2314 tests, CHECK EXIT 0 (before 2311). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
9d6a0aebd3 |
Let Compare and Equals treat float noise as equal numbers
A graph that checks "time - late == 6" after On timer (#224) failed, because the difference comes out as 5.999999999999972. Both values are sums of measured frame times, and they round along different paths. The only workaround was a Round node in front of every comparison. - New helper nearlyEqual in values.ts. Two numbers count as equal when |a - b| <= 1e-9 * max(1, |a|, |b|). The tolerance grows with the size of the values, so it holds for hours of summed dt. - Compare uses the rule for all six operators. "<" and ">" only hold when the values are not nearly equal, "<=" and ">=" also hold when they are. So "==" and ">" are never true at the same time. - Equals uses the same rule when both inputs are real numbers. Objects still compare by identity, everything else as text. The node description says so. - The tolerance is fixed, there is no parameter on the node. Physics values come from float32 in Rapier and are not covered on purpose. - The log display is unchanged and still prints the raw number. Tests: - ./check.sh: 156 files, 2311 tests, CHECK EXIT 0 (before 2300). - 11 new cases in node-catalog.test.ts, including the value from the log against 6, growth with magnitude, 6.001 != 6 and 1e-6 > 0. - Counter-check exact compare: exactly the 8 noise cases fail. - Counter-check "<"/">" without the exclusion: exactly 2 cases fail. - No e2e change, the check sum stays at 1374. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
6fecb2e841 |
Tell On timer how late its frame is, so time minus late is exact (#224)
A timer can only fire on a frame. It fires on the first frame past each due time, so Time.time at that moment is the due time plus up to one frame. On a display that is not exactly 60 Hz this delay creeps up by a millisecond every few seconds and falls back to zero once it reaches a whole frame. The issue saw 82.009 after 82 ticks. The timer itself does not drift, it keeps the remainder when it fires. - On timer gets a number output late. It is what is left of the timer's clock after taking one interval, so time - late is exactly the due time. - When one long frame catches up several ticks, each gets its own late. The first one was due longest ago and is the latest. Tests: - New unit test in behavior-events: late 0.125 after a frame past 1.0 s, 0 on the dot, 1.5 and 0.5 for two ticks caught up in one frame. - Counter-check: late taken before the interval is subtracted fails exactly that test. Original restored with cmp. - ./check.sh: 156 files, 2300 tests, CHECK EXIT 0 (was 2299). - No e2e change. The e2e modules only wire the exec of On timer. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
aef1eba917 |
Stop thin lines between tiles of a nearest tilemap at odd zooms
At zoom levels that do not land quarter edges on whole screen pixels, a tilemap whose texture uses Nearest showed one-pixel lines in the wrong colour. The float uv of a pixel centre right at a quad edge could round onto the first texel of the neighbouring quarter in the sheet, which is usually a different tile. Terrain sheets showed it most, because every tile is four quads there. - Nearest now pulls the uvs 0.02 texel inside each quarter. That is far below half a texel, so every picked texel stays the same and pixel art keeps its size. - Linear keeps its half-texel inset. - The first draft used 0.01. The user raised it to 0.02 after checking the result in the browser. Tests: - ./check.sh: 156 files, 2299 tests, CHECK EXIT 0 (before 2298). - New unit test: every nearest uv sits strictly inside its quarter, by less than 0.05 texel. Two older tests now expect the inset. - Counter-check: with no inset for nearest the new test and the inset test fail. - No e2e check. The user checked the result in the browser. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
9ef360ba1b |
Let the game view show the picture at a chosen scale (#225)
The game view always stretched the camera picture to fit the pane.
Pixel art came out blurred at odd factors, and there was no way to
see the game at its real size. The view now has a header with a
scale drop-down, and 100 % is the default.
- The header of the new GamePane offers Auto, Fit, 25 %, 50 %,
100 %, 200 % and 400 %. Auto and Fit show the factor they land
on, e.g. "Auto · 100%", and follow the pane when it resizes.
- 100 % means one screen pixel per pixel of the camera resolution
(width x height). The camera zoom scales the world on top of
that, so a zoomed camera still fills the same frame.
- Auto takes the largest fixed step that fits the whole picture
and falls back to 25 % when nothing fits. Fit keeps the old
behaviour.
- A picture larger than the pane is cropped around the camera
centre. Fixed steps start on a whole pixel so pixel art stays
sharp at 100 %.
- The choice is kept in the browser like "Play starts". Nothing
is written into the project.
Found on the way: the first draft applied the step to the world
instead of the camera resolution. The camera zoom then had no
effect at 100 %, which the existing check 12c caught (zoom 1.37
measured as scale 1).
The CI smokes were not run locally, because that would reseed
projects the user works in. All three smoke cameras are 640x360
and fit the 670x383 harness view at 100 %.
Tests:
- ./check.sh: 156 files, 2298 tests, CHECK EXIT 0 (before 2290).
- 8 unit tests in tests/viewport/view.test.ts for whole-pixel
steps, cropping, Auto and Fit with fallback, and zoomed cameras.
- 7 e2e checks in 03g. e2e 1-17,28,30,38-40: ALL PASSED, 727
checks.
- Counter-checks: no cropping, no rounding, no factor reported to
the header, and no zoom share each fail their own tests.
- Check total: 1367 (CI at
|
||
|
adminandClaude Opus 5.5
|
c1a702bc39 |
Add a terrain autotile kind for 6x4 sheets like TD Plains
Sheets such as "TD - Tileset - Plains" draw an autotile as a 6 x 4 block: a 3 x 3 nine-slice, a vertical and a horizontal strip and a single tile on the left, the inner corners on the right. The floor kind only knew the RPG Maker 2 x 3 layout, so these sheets could not be used as autotiles at all. - New TileKind "terrain" with 16 px tiles in a 6 x 4 block. All 47 neighbour combinations are resolved per quarter, so the renderer, the collision and the palette stay untouched. - Whole tiles are used where the sheet draws them: the 16 side cases, the four single notches and the two diagonal pairs. Hand-drawn strips and single blocks therefore look as drawn. - Only the missing cases are composed, for example an edge with a notch or two notches side by side. They take the side tile and put the notch quarters in. - A notch also takes the quarter above or below from its notch tile when fill would stand there. Tall cliff notches keep their wall. On ground and stones fill meets fill, so nothing changes there. - The label rule for floor and terrain now runs through one notch function. Floor still passes the MZ table test. Not touched: the tile board only refits on open and on Fit. Switching an open tile to a kind with another tile size leaves it small until Fit is pressed. That was the case before. Tests: - ./check.sh: 156 files, 2290 tests, CHECK EXIT 0 (before 2283). - e2e 39,40: ALL PASSED, 90 checks (before 86). 39j checks 47 tiles, pixels of single, notch, pair and composed tiles, and undo to floor. - Counter-checks: without whole notch and pair tiles one unit test and one e2e check fail. Without the taken quarter one of each fails. Originals restored by cmp and rebuilt. - Check sum: 1363 measured by CI plus 4 is 1367, computed only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
32cb5b4ba5 |
Size, turn and scale a selected sprite or collider in the scene (#223)
A sprite or collider could only be sized, turned and scaled through
the inspector. The scene view now draws a size gizmo on the selection,
so the common edits happen where the object is.
- Edges and corners set Width and Height. The opposite side stays
where it was, so the position moves by half the change. Sizes land
on whole units, at least 1.
- A circle collider has edge handles only and stays round. A polygon
or frame collider keeps its own corner editing and gets no size
handles.
- One knob on a stem above the top edge turns or scales. The first
5 px of the drag decide which: out from the centre scales, sideways
turns. The choice holds until the release, so one drag never does
both. Turning and scaling share one handle by the user's choice.
- Shift snaps a turn to 15 degrees. The turn adds up the swept angle,
so it goes on past a half circle instead of jumping back.
- The values are overrides of this placement, like "Fit to sprite".
A value equal to the asset's own clears the override. One drag is
one undo step.
- A handle reaches 7 px outwards but only half that into the object.
Otherwise a press near the edge of a small sprite sizes it instead
of moving it.
- The rules live in src/editor/sizeGizmo.ts without Pixi. The handles
are drawn and driven by src/editor/viewport/sizeGizmoEditor.ts. The
test hook __gm.sizeGizmo() reports them in screen pixels.
Found on the way:
- The first e2e run failed the play-mode drag check in 41b. It grabs
the 40 px crate 6 px inside its bottom edge, and the full grip zone
took that as a resize. The half zone inside fixes it.
- A first gizmo check failed because the sprite's right edge lay
outside the scene view (x 940, view ends at 925). The test was
wrong, the sprite now sits further left.
Tests:
- ./check.sh: 156 files, 2283 tests, CHECK EXIT 0 (before 2273).
- 10 unit tests in tests/editor/size-gizmo.test.ts.
- 6 new e2e checks in 29j: handles present, edge resize with the
opposite edge fixed, undo, knob turn, knob scale that ignores a
later sideways pull, circle radius.
- e2e 1-16,29,39-43 on the final state: ALL PASSED, 663 checks.
- Counter-checks: deciding the knob mode on every move fails the
"does not turn it" check. Not moving the centre on an edge drag
fails both edge checks. A full grip zone inside fails the new unit
test. Originals restored.
- Check total 1357 + 6 = 1363, computed. CI measured 1357 at
|
||
|
adminandClaude Opus 5.5
|
d257298bc7 |
Keep a dragged rigidbody on the mouse while the game plays
Dragging a sprite with a dynamic rigidbody in the scene view during play barely moved it. Only a quick drag and release got through. A dragged object should follow the mouse as if it had no rigidbody and start from rest when it is let go. - The scene view reports the dragged objects as held through EditorState. The game tick no longer writes its own position over a drag position that has not been read in yet. This affected every object in play, not only rigidbodies. - A held dynamic body turns kinematic and is set straight to the drag target. Gravity and velocity no longer pull it away between two mouse moves, and the "moved directly" warning stays quiet. - A body counts as held when the object itself, an object above it or its rigidbody is dragged. - On release the body turns dynamic again with linear and angular velocity set to 0. Not changed: a press on an unselected crate still lands on its same-sized collider (rule from #203). Dragging while play is paused and in the editor run is untouched. Tests: - ./check.sh: 155 files, 2273 tests, CHECK EXIT 0 (before 2271). - e2e 41 alone: 21 checks. e2e 1-16,39-43: 641 checks, all passed. - 2 unit tests: a held body stays where the drag puts it and starts from rest, also for a grip shorter than one physics step. The tick does not overwrite a drag position that is not read in yet. - 1 e2e check in 41b: the crate is lifted 150 units with the real mouse, holds exactly at -3490 for a second, then falls from rest (1.3 px, then 5.1 px) with an empty log. - Counter-checks: without holding in physics the crate falls back to the ground and the old warning shows. With the tick writing back the unit test fails. Without zeroing, a short grip keeps 490 px/s and the test fails. - Check total 1356 + 1 = 1357, calculated. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
d444b27fae |
Let behaviors read and set tilemap cells while the game runs (#135)
A behavior could move sprites around but had no way to look at or
change the tilemap it was standing on. Doors, breakable blocks and
"is there ground ahead?" checks all need that.
- New nodes Tile at, Set tile, World to cell and Cell to world. The
conversion honours position, rotation, scale and mirroring of the
placed map. A small Tile literal node lets a graph compare what it
read against a fixed tile.
- The layer is picked by name from all tilemap layers in the project
through a new field kind. "First layer" takes the map's first one.
Because layer is an input, a text can also be wired in.
- Each placement keeps only its edited cells in liveTilemap.ts,
layered over the asset's cells. A second placement of the same map
stays untouched, and the asset itself is never written.
- A read sees a Set tile from earlier in the same flow. Drawing picks
up the edit at once and neighbouring autotile edges adapt.
- Solids see the edits straight away, so a Move in the same frame
already hits the new cell. Rapier rebuilds its shapes on the next
step.
- After Stop the map is back to what was painted. The editor run shows
the edit in the scene view while it lasts and does not touch the
project.
- An unwired tilemap input means no map, like Get position. Cells
outside the grid are refused and done reports false.
Tests:
- ./check.sh: 155 files, 2271 tests, CHECK EXIT 0 (before 2245).
- tests/core/tile-access.test.ts (14), one physics test for a floor
set at runtime, catalog entries for the five nodes.
- e2e 39,40: all passed, 86 checks (before 81), five new in 40s.
- Counter-checks: collision ignoring edits fails 3 unit tests, the
solids cache missing same-tick edits fails only the "at once" test,
drawing ignoring edits fails the 2 drawing checks, the editor run
not reporting a new tile fails only its check.
- 40s uses a still tile. The animated fixture tile redraws every frame
and hid the last counter-check at first.
- Check total 1351 + 5 = 1356, computed. CI confirmed 1351 at
|
||
|
adminandClaude Opus 5.5
|
cb32b4bc23 |
Draw scene symbols as white panel icons that make room for each other
The scene view drew its symbols from its own SVGs, each in a type colour and some on a plate. They looked heavier than the panel icons and were hard to read on busy ground. Symbols of different objects also sat on top of each other whenever their positions met. - Every symbol is now the panel icon of its kind, one file for both places. The scene-only SVGs in src/viewport/icons are gone. - Symbols are white in a thin dark outline with no plate behind them. haloSvg widens each shape of the file by its own stroke and draws that silhouette dark under the icon, so no second file is needed. - A butt stroke end would leave its end without outline. The halo caps such strokes square, and the rigidbody arrow shaft is now a filled rect so its top edge is outlined exactly. - Symbols are 18 world units across, stay upright and are clickable. A click selects the object, a drag moves it. - symbolLayout places every symbol on its object's position. Symbols that overlap stand on a ring round the middle of their positions, evenly spaced one icon apart. A ring that reaches into another symbol takes it in until nothing overlaps. Ownership plays no part. - Names sit beside their symbol, on the outside of a ring. Selection frames and parent links follow the symbol. - A collider icon no longer shows trigger or invalid state. The outline colour of the shape still does. - A lone collider sits on its sprite's middle again. A press there on an unselected sprite drags the collider, so e2e 02 and 16 now grab the sprite beside its middle. Hook __gm.container.objectRect helps 16 find that spot. Tests: - check.sh: 2245 tests, CHECK EXIT 0 (+8 for layout and halo) - e2e 1-16,31,39-43: 670 PASS, 2 FAIL. Both were sampling too strictly for the thinner, smaller icons. 5c looked for the exact outline colour, which the thin outline only reaches blended at this zoom. 31n started the ring scan one pixel past the smaller ring. Both are fixed, and e2e 5,31 then passed with 48 checks. - 5c checks white with an outline and none of the old colours, 41b checks the ring positions and that a click on a symbol selects its object, 31n checks the now white particle ring. - Counter-checks: without the halo only 5c fails, also with the looser outline match. Without the layout the ring and click checks in 41b fail. Originals restored by cmp. - The e2e check total stays 1351. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
e3d47ca9c1 |
Show a clickable symbol for colliders in the scene view
Colliders were the only placed part without a symbol in the scene view. A collider that sits exactly on its sprite was hard to find and could only be picked by its outline. - The symbol is the box with four corner handles from the asset panel. It stays upright and has a fixed size in world units. - colliderColor() holds the colour rule of the outline (green, trigger colour, red for a shape that collides with nothing). The outline and the symbol share it. - The symbol sits 28 units left of the collider centre. The sprite centre stays free for grabbing the sprite, and a rigidbody symbol on the same spot is not covered. - The symbol has its own 26x26 hit area and the outline's handler. A click selects the collider and a drag moves it. sceneStack counts it as a hit under the pointer. - The hook __gm.sceneIcon(id) also returns x, y, color, screen and clickable. Findings: - With the clickable symbol on the sprite centre, four checks in 01 and 09 dragged the collider instead of the crate. That is why the symbol moved beside the centre. Tests: - ./check.sh: 153 files, 2237 tests, CHECK EXIT 0 (was 152 / 2234). - New unit test for colliderColor. svg-icons.test.ts lists the collider symbol. - e2e 41b: 3 new checks (beside the centre without and with a rigidbody, click selects the collider). - e2e 1-16,39-43: ALL PASSED, 635 checks. e2e 41 alone: 20 checks. - Counter-probes: no symbol fails both placement checks. A symbol without hit area fails only the click check. - Check total 1348 + 3 = 1351, computed, not yet measured by CI. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
abf3b1f618 |
Pick an existing collider or rigidbody from the object menu (#222)
"Add collider" and "Add rigidbody" could only create a fresh asset.
Reusing one meant creating a copy by hand, and a new box collider
never matched the sprite it was added to.
- "Add collider…" and "Add rigidbody…" open a second menu at the same
spot: a search field, "Create new …", a separator, then every asset
of that kind sorted by name. Enter takes the first match.
- An existing asset is placed as a shared child, not copied. An asset
the object already carries is greyed out.
- The rule for when the entries are available lives in one place in
the collider commands. An object without a resolvable asset is
greyed out again, as with the old canAddCollider.
- "Fit to sprite" in the menu of a placed collider sizes it to its
sprite and centres it (position 0,0, z kept) as one undo step.
- Fit writes overrides of that placement, not the asset, because a
collider can now sit on several sprites. A value equal to the asset
value clears the override instead.
- Fit removes the collider's own scale, swaps width and height near
90 degrees and gives a circle the radius that fits (min/2). It is
greyed out for polygon and frame outlines and off a sprite.
- Menu has no submenus, so a second menu opens instead, similar to
"Add Component" in Unity. That costs one click more than before.
Tests:
- ./check.sh: 152 files, 2234 tests, CHECK EXIT 0 (before 151/2228).
- 6 new unit tests in tests/core/collider-parts.test.ts.
- 3 new e2e checks in 08a: submenu with search and greying, an
existing collider via search and Enter, Fit on oc2.
- 09, 40 and 41 follow the new labels.
- e2e 1-16,39-41 on the final build: ALL PASSED, 596 checks.
- Counter-checks: no greying, existing asset ignored and fit without
centring each fail at their own check. Breaking the override rule
fails 2 unit tests.
- CI confirmed 1345 checks at
|
||
|
adminandClaude Opus 5.5
|
2cf72022eb |
Ping an asset from its inspector link instead of selecting it (#221)
A single click on an asset link used to select the asset. That swapped
the inspector over to the asset, so the object being edited was gone
and the second click of a double-click landed on a different row. A
click now pings the asset in the asset panel, like Unity does in its
Project window, and a double-click opens it.
- state.pingAsset(id) emits a new `ping` event. It runs no command and
leaves the selection alone, like openAsset.
- ListView.ping(id) draws an accent overlay on the row. It stays full
for 200 ms and is gone after 1 s, driven by onFrame. destroy cancels
the animation.
- The asset panel expands the group, scrolls to the row and pings it.
- The capture listener and followedLink in the inspector are gone.
The inspector is no longer rebuilt, so pressCount on the label is
enough. followsPress and pressOf are no longer exported from ui/.
- The "Asset" and "Part of" rows of a scene object were a ghost button
that selected the asset. They are now a link label with the same
ping and open behaviour, next to a read-only field with the name.
- The link logic lives in linkLabel() in fieldFactory.ts. createFieldRow
and the asset row share it.
- TextField gets a readOnly option: no focus, no caret, no hover.
- New hook __gm.assetPing(). inspectorLink also reports the field
next to the label (value, readOnly).
Findings:
- Renaming an asset while its object was selected left the old name
in the inspector. This was already the case with the old button.
The refresh now updates the name.
Tests:
- ./check.sh: 151 files, 2228 tests, CHECK EXIT 0 (before 150 / 2226).
New: tests/ui/list-ping.test.ts for pingAlpha and pingAsset.
- e2e 1-16,28 on the final build: ALL PASSED, 575 checks.
28l-ii has 9 checks instead of 5.
- Counter-checks, each failing exactly at its check: ping also selects,
no follow-up frame (alpha stays 0.55), inspector never opens, field
takes focus, name not refreshed, label not a link.
- Check total 1345, calculated. CI last counted 1336 at
|
||
|
adminandClaude Opus 5.5
|
1aae97b335 |
Open a sprite's texture or shader from the menus and inspector (#221)
A sprite's texture and shader were only reachable by hunting for them
in the asset panel. Every place that shows such a reference now leads
straight to it.
- Object menu (hierarchy and scene view): "Open texture" and "Open
shader" sit right below "Open asset" when the sprite has one. The
object's own values count, so an overridden shader opens the
override.
- Asset context menu: the same entries at the top.
- Inspector: every asset field with a value shows its label as a blue,
underlined link. One click selects the asset, a double click opens
it. An override stays gold and gets the underline too.
- The first click rebuilds the inspector, so the second one lands on a
different row. The inspector catches that second click before any
row sees it.
- New pure rule `references.ts`: the asset behind a reference field. A
frame leads to its sheet. Kinds without a tab (palette, font, sound)
get no menu entry.
- `followsPress` in `clicks.ts` keeps the double-click window and
tolerance in one place.
- `activateAsset` in `actions.ts`: opening means the same everywhere
as a double click in the asset panel, so a scene also becomes the
active one. The asset panel uses it too.
Tests:
- ./check.sh: 150 files, 2226 tests, CHECK EXIT 0 (before 148/2214).
- 12 new unit tests: 6 for the reference rule, 4 for the link in
createFieldRow, 2 for followsPress.
- 5 new e2e checks in 28l-ii: both menus, the link's look, click
selects, double click opens.
- e2e 1-16,26,28,35,39,40 on the final build: ALL PASSED, 732 checks.
- Counter-probes: no "Open ..." entries, no underline and a second
click never recognised each fail exactly at their checks.
- Check total 1341 (CI counted 1336 at
|
||
|
adminandClaude Opus 5.5
|
9b62d572b4 |
Stop the bead spacing check failing on one slow frame
CI task 261 at
|
||
|
adminandClaude Opus 5.5
|
fd6af84474 |
Keep the texture tab smooth with many collision outlines (#220)
With a large sheet the texture tab stalled on every mouse move. A
profile over a pan with 1728 frames showed two causes. Half of the time
went into convexPieces, which split every outline into convex pieces
only to pick its colour. And every pointer move during a pan rebuilt
and re-tessellated the whole overlay geometry.
- TextureCanvas draws on two layers. `sheet` holds every frame except
the selected one and is rebuilt only when the zoom, the selection or
one of those frames changes. `overlay` holds the selected frame, its
handles and the rubber band of a new frame.
- Both layers are drawn without the pan. A pan only moves the
container and places the name labels again.
- SheetCanvas has a new hook `panned()`. By default it calls `redraw()`
as before, so Pixel Art and Vector Art are unchanged.
- The validity of an outline is remembered per outline and comes from
polygonProblem, with the same result. This works because the model
replaces an outline list on every change and never edits it in
place, the same as frameOutline.ts already relies on.
- The name plates stay below the outlines and the texts on top. Checked
on a screenshot: the corner handle next to a label sits on top.
- New debug hook `__gm.texture.drawInfo()` with the rebuild count, the
outline count and the screen positions of outlines and names.
Measured with a throwaway probe, 1728 frames, 120 moves, main thread
time per move, before and after in the same browser:
- Middle-click pan: 76.9 ms -> 4.7 ms (4.4 ms without frames)
- Dragging a frame: 51.7 ms -> 10.1 ms
- Wheel zoom: 52.2 ms -> 30.9 ms
- Selecting another frame: 67.6 ms -> 41.0 ms
Findings:
- Wheel zoom and selection still rebuild `sheet` once on purpose. That
stays a single hitch per notch or click on very large sheets.
Tests:
- ./check.sh: 148 files, 2214 tests, CHECK EXIT 0.
- e2e 1-16,37,39-43 on the final build: ALL PASSED, 641 checks.
- Three new checks in 42c-ii. A corner drag does not rebuild the other
frames. An undo on an unselected frame reaches the picture (1 -> 2
outlines). A pan moves outlines and names by exactly 40/25 px without
a rebuild.
- Counter-checks: a missing outline in the cache key and a pan that
does not move each fail at their check. The selected frame in the
key fails alone with 9 rebuilds during an 8-step drag.
- Check total 1333 -> 1336 (confirmed by CI at
|
||
|
adminandClaude Opus 5.5
|
ceaa44e5b8 |
Add roles and a user management tab for admins (#180, #179)
Until now every account could do the same, and managing accounts meant running make-user.php on the server. Accounts now carry a role, and admins get a tab in the editor to create, rename, re-role, reset and delete accounts. Both issues share the same server files, so they land in one commit. #180 Roles: - Every account is admin or user. A missing or unknown role counts as user, on the server and in the client. - auth/session and auth/login return {user, role}. - Auth::requireAdmin() answers everyone else with 403. - make-user.php takes the role as a third argument. Without it an existing account keeps its role, so a password reset from the CLI never demotes an admin. - Admins see and save other users' projects like their own. Saving keeps the original owner. #179 User management: - The account menu gets "User management" (admins only) and "Plugins" (everyone). Both open the same closable tab, accounts on top and plugins below. The fixed Plugins tab is gone. - Per account: Rename, Role, Password (no old password needed) and Delete with a confirmation. "New user..." creates one. - Each account lists its projects with an Open button. - Delete is refused while the account still owns projects. - Renaming moves the projects along and ends every session of the account. transfer takes the same lock as save. - On their own row an admin can only set the password. The last admin cannot be demoted or deleted. The server enforces both. - The account rules live in AccountRules.php and accountRules.ts as pure functions, so they are testable without a server. Findings: - At 1280 px the Delete button reached past the column edge and was not clickable. The name column is flexible now, and module 49 checks the layout. - The server lets an admin delete other users' projects, but the editor has no button for it yet. Left open in the issue. Tests: - ./check.sh: 148 files, 2214 tests, exit 0 (was 147 / 2204). - New tests/php/accounts.test.php, 51 checks, plain PHP. CI runs it as its own step in the gate job, the host has no PHP. - New e2e module 49 with 18 checks. CI creates the e2e-admin account in serve.sh. Modules 14, 38, 46 and the scaffold follow the new tab. - e2e 1-16,37,38,45,46,49: all passed, 581 checks. - Counter-checks: owner on save, admin view, default role and self delete in PHP fail 7, 5, 5 and 1 checks. Unchecked requireAdmin, a wider column and the caller as owner each fail exactly their e2e check. - Check sum 1314 -> 1333 (+1 in 38, +18 in 49), computed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
30fb3a2316 |
Dragging previews once per frame instead of on every move (#213)
Scene view, texture tab and graph called state.preview() on every pointer move. Each call rebuilds panels, the editor run and generated images, so a drag in a busy project cost up to 34 ms per move. The model is still written on every move. Only the preview now waits for the next frame. - New FramePreview in src/editor/framePreview.ts. request() goes into the move handler, cancel() into the release. The rule lives in one place instead of six copies. - Scene view: object drag, polygon corners and tilemap strokes use it. - Texture tab: frames, slice edges and collision corners use it. - Graph: nodes, groups and comments use it through GraphDrags. That covers the state machine, shader and particle tabs as well. - Vector art is unchanged. Its frame already writes the model since #129. Findings: - The texture tab paid twice per move (232 project events for 120 moves). Pixi sends a second pointer event 4 to 8 ms after each real move as a hover update. It now lands in the same frame for free. - Inspector scrubbing and gradient stop dragging still preview on every input. Both lie outside the three canvases of the issue. Measured with a throwaway probe, 120 moves at 8 ms, 120 extra objects: scene 26.2 -> 18.5 ms/move, texture 34.2 -> 17.9, graph 28.8 -> 17.0. Project events dropped from 120/232/119 to 48/48/41. Tests: - ./check.sh: 147 files, 2204 tests, exit 0 (before 146 / 2199). - 5 new unit tests in tests/editor/frame-preview.test.ts, including the real GraphDrags. Counter-checks (direct preview in the graph, no cancel on release, one key per name) each fail the expected test. - e2e 1-16,29,31,36,37,39-43: 708 checks passed. - e2e 1-16,26,28: 634 checks passed. - No new e2e check, the check sum stays at 1314. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
f9d5e8bca5 |
Move the one-way rule and collider geometry out of physics (#215)
PhysicsSystem had 940 lines. The step logic shares its state with the engine handles of parts and bodies, so a separate class would need almost the whole system as its interface. That part stays. Two pieces without own state move out instead. Behaviour is unchanged. - passesOneWay in oneWay.ts is the one-way rule as a pure function. It can now be tested without Rapier. - bodyGeometry.ts holds the collider shapes and the pose maths. - PhysicsSystem is down to 820 lines. Tests: - ./check.sh: 146 files, 2199 tests, CHECK EXIT 0 - 5 new unit tests for the one-way rule without an engine. - New tests/core/body-geometry.test.ts: a capsule turned by 90 degrees went unnoticed before. Both tests fail against that bend. - Counter-checks on landing hold and teleport detection fail in 3 to 6 tests each. - e2e 1-28,31,36,39-43 on the final state: ALL PASSED, 1060 checks - Each of the four #215 commits was checked green on its own in a scratch worktree: 2190, 2192, 2192, 2199 tests. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
ed29b2f3dc |
Split the debug hook into one file per topic (#215)
debugHook.ts had grown to 1097 lines of unrelated test hooks. They now live in seven files under src/editor/debug/, grouped by topic. debugHook.ts only assembles them and is down to 86 lines. - The shape of window.__gm is unchanged for tools/. Measured in the browser before and after: 120 top-level keys and 313 leaves with their arity, byte-identical JSON. - The Text import is explicit in the helpers. Without it the name fell back to the DOM type and instanceof would never have matched. Findings: - The comment for vectorArt sat above pixelArt. It is back in place. - An orphaned duplicate comment on assetRow is gone. Tests: - ./check.sh: 145 files, 2192 tests, CHECK EXIT 0 (unchanged) - e2e 1-16,37,39-43,47,48: ALL PASSED, 734 checks - The shape probe turns red against a deliberately misnested hook. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
b830f9c747 |
Split wiring, drags and the frame layer out of GraphCanvas (#215)
GraphCanvas had grown to 1406 lines. Wiring, dragging and the frame
views each carry their own state, so they now live in their own
classes behind a narrow host interface. Behaviour is unchanged.
- GraphCanvasHost is what the new classes see of the canvas.
- Wiring owns dragging a pin, the palette on a drop into empty space
and "Make transition".
- GraphDrags owns moving nodes and frames, including snapping.
- AnnotationLayer holds the frame views. The test hooks stay on the
canvas and read the views through the layer.
- BeadClock moves into beads.ts and can now be tested without a
browser.
- The canvas builds host, wiring and drags in its constructor because
state is a parameter property and not set during field init.
- GraphCanvas is down to 1019 lines.
Tests:
- ./check.sh: 145 files, 2192 tests, CHECK EXIT 0 (+2 for BeadClock)
- e2e 1-28,31,36: ALL PASSED, 932 checks
- New e2e check in 36k: a frame caught by the rubber band is also
drawn as selected. No check covered that before. Against a bent
build it fails with selected: false.
- Counter-checks for drop target, snapping and beads fail exactly
the expected checks.
- Editor check sum is now 1314 (CI measured 1313 at
|
||
|
adminandClaude Opus 5.5
|
e31db3df1d |
Split tile painting and polygon corners out of SceneViewport (#215)
SceneViewport had grown to 1216 lines. Three blocks carried their own state and only needed a few things from the view, so they now live in their own classes. Behaviour is unchanged. - SceneViewHost is the narrow interface the new classes see of the view: state, coordinate mapping, redraw and the preview trigger. - TilemapPainting owns the paint stroke, the rectangle preview and the flood fill, with press/move/release called in the old order. - PolygonCorners owns hovering, dragging, inserting and deleting the corners of a polygon collider. - TilemapSolids draws the solid outlines of tilemap layers. - SceneViewport is down to 668 lines. Findings: - PaintStroke.holderId was written but never read. It is gone. Tests: - ./check.sh: 145 files, 2190 tests, CHECK EXIT 0 (unchanged) - e2e 1-16,29,31,39-43: ALL PASSED, 679 checks - Counter-checks in the browser: bending the brush, the corner drag and the solid outline each fails exactly the expected checks. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
adminandClaude Opus 5.5
|
487b3aa5b4 |
Painting over identical tiles no longer adds an empty undo step (#212)
A rectangle drawn over cells that already hold exactly that tile
pushed a command that changed nothing. Undo then seemed to do nothing
and the project was marked as modified. PaintTilesCommand.changes()
was meant to prevent that, but nothing called it since #128.
- New paintTiles() in tilemapCommands.ts returns the command for a
finished stroke or fill, or null when no cell changes. It is the
first caller of PaintTilesCommand.changes().
- All three paint paths in the Scene View go through it: brush and
eraser, rectangle (both in endStroke) and flood fill (fillFrom).
- Flood fill was not affected, because floodFill already returns []
for the same tile. The guard is there anyway, as the issue asked.
- A rectangle that changes even one cell stays a normal undo step.
Tests:
- ./check.sh: 145 files, 2190 tests, CHECK EXIT 0 (before 2187).
- 3 new unit tests in tests/core/tilemap.test.ts: same rectangle
twice gives no command, a wider rectangle still does, a fill that
finds nothing gives no command.
- 1 new e2e check in 40d: the same rectangle again, then one undo
must take 9 cells to 0. A redo restores them for later blocks.
- GM_STEPS=39,40 on the final build: ALL PASSED, 81 checks.
- Counter-probe with changes(...) || true: exactly the two unit
tests fail, and in the browser exactly the new check fails with
{"afterUndo":9}. Original restored via cmp (IDENTICAL), rebuilt.
- Editor suite check total: CI measured 1312 at
|
||
|
adminandClaude Opus 5.5
|
0e8585665a |
Load the 500 weight and draw missing glyphs as SVG icons (#214)
The theme asks for Source Sans 3 at weight 500, but fonts.ts only
loaded 400, 600 and 700. The browser synthesised the 500 text. The
issue comment also asked for the symbols the font lacks to become SVG
icons. A scan of src/ against the font's cmap found 20 such characters
in 30 places.
- fonts.ts loads source-sans-3-latin-500-normal.woff2 as well.
- Menus draw their ticks and the "current" dot as icons. The dot is a
new MenuItem.current field and no longer part of the label text.
- Inspector action buttons (reset, apply, fit width/height) are
icons. The menu arrow behind them is the drawn Dropdown chevron.
- Kind marks come from a table in markIcons.ts. Picture kinds use an
icon, # and Aa stay text because the font has them.
- Scene marks in the asset panel, sound play/stop and the transition
arrows are icons too.
- Apply is an icon although the font has the arrow. It sits next to
the reset icon and would look different as a glyph.
- A new Icon widget draws on redraw like Button, so Node tests need no
DOMParser.
- Tests find icons through the SVG id. The new hook
__gm.locateIcons(id) and the harness helpers findIcon/clickIcon
replace locate() on the old characters.
Findings:
- Weight 700 is loaded but the theme never uses it. Left as is.
- The old scene-mark check in module 35 accepted any play glyph on
screen. It now checks the mark on the start scene's row.
- Module 44 needs the textures from module 37 and fails without it.
Tests:
- ./check.sh: 145 files, 2187 tests, exit 0 (before 144 / 2143).
- New tests/ui/fonts.test.ts: every theme weight is loaded, and every
character in src/ exists in every shipped face. The cmap reader
matches fontTools (231 characters).
- Icon tests cover the 19 new files. Each kind mark is either an
icon or text, never both.
- e2e 1-16,22,30,34,35,43,44: 586 green, 0 FAIL, then 44 stopped
for lack of module 37. e2e 1-16,37,44,47,48: 618 green.
e2e 1-14 on the final build: 454 green.
- Two new e2e checks: loaded weights in 14, a reset icon pixel in 04.
- Counter-probes all fail as expected: 500 removed, a play glyph put
back, the struct icon removed (one unit test each). In the browser,
500 removed and reset icon hidden fail exactly the two new checks.
- Editor-suite check sum 1312 (1310 measured by CI at
|
||
|
adminandClaude Opus 5.5
|
056a622524 |
Refactoring round 6: shared sheet canvas, fewer copies (#85)
Recurring cleanup round. Behaviour is unchanged. Duplicated code in
src went from 27 clones with 299 lines to 6 clones with 56 lines
(jscpd, the two identical OFL licence texts not counted).
- SheetCanvas is the new base of the texture, pixel art and vector art
canvases. It holds background, mask, hint, panning, wheel zoom, fit,
view save and restore and the screen mapping. Pixel art keeps its
whole-pixel panning and fixed zoom steps.
- mixColor in src/core/util/color.ts replaces three copies in
animation, gradient and particles. It clamps t like the particle one.
- fingerprint now lives once in generated.ts, next to the revision
contract from #185.
- MoveCommand next to RenameCommand is the base of the three reorder
commands for tile pictures, tilemap layers and vector shapes.
- The two raycast nodes share inputs, settings and filter via rayOf.
- canvasRect() lives once in AssetTabEditor.
- Removed unused code: findPixelArt, isDrawableShape,
GeneratedPictures.has and SnapToggle.snapping. Tests now check the
same things through equivalent paths.
- 38 values used only in their own file lost their export.
- Three comments that told history now describe the current state.
The bundle stayed md5-identical for those edits.
- README lists vector art, pixel art and palette, and links the two
wiki pages.
Findings:
- PaintTilesCommand.changes is meant to prevent empty undo steps but
has not been called since #128. The rectangle tool therefore records
a no-op undo step when it paints the same tile over itself. Left as
is, because fixing it changes behaviour.
- No test covered middle-click panning in any sheet tab before. 13f
and 48a now check it.
- The default values of the ray inputs were untested. Length 1000
could become 10 without a red test. tests/core/raycast.test.ts now
covers it.
- CI measured 1308 editor checks at
|
||
|
adminandClaude Opus 5.5
|
d7c0b7cfdb |
Variables and state machines can reset with every scene (#164)
A shared variable or machine used to live either per object or for the
whole run. Levels need a third lifetime: shared like global, but built
fresh from the declarations whenever a scene is loaded.
- New scope `scene` for Variables assets and state machines. A missing
field keeps the old meaning: `global` for assets, `local` for
machines, so existing projects do not change.
- Returning to a scene that was visited before is fresh as well. There
is no memory per scene.
- A `scene` machine announces its start state again after a scene swap,
so `On state entered` in the new scene hears it. A global machine
stays quiet because it never left its state.
- The rule lives in two pure functions in `sharedState.ts`
(`carriedForward`, `restartedAssets`). BehaviorSystem only calls them.
- The inspector shows a Scope dropdown on Variables assets and a third
step on the machine dropdown, with a note on what `scene` means.
Decisions:
- `scene` and `persistent` exclude each other (variant A). The choice
dialog was cancelled, so this is an assumption to revisit. A
scene-scoped asset hides the Persistent switch, and switching to
`scene` drops the flag (undo restores it). The guard sits in
`isPersistent` itself, so no path can save values the scope throws
away anyway.
Findings:
- `clickText('Scene')` hits the Scene tab, not the menu entry. One e2e
check was red because of that, and its partner was falsely green
because it compared `undefined` with `undefined`. Both use `clickLast`
now.
Tests:
- ./check.sh: 144 files, 2142 tests, CHECK EXIT 0 (before 143 / 2133)
- 9 new unit tests in tests/core/scene-scope.test.ts
- e2e 32 fresh: ALL PASSED, 17 checks
- e2e 1-3,17,18,26,32,35 fresh: ALL PASSED, 213 checks
- 3 new e2e checks (35: scene-scoped asset resets on swap, 32h: dropdown,
note, hidden Persistent switch)
- Counter-checks bite: ignoring `restarted` fails 2 unit tests and the
new check in 35, the old announcement rule fails 1, dropping the scope
guard in `isPersistent` fails 1. Originals restored, cmp IDENTICAL.
- Check sum 1307 (1304 + 3) is computed. Last CI measurement: 1293.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
||
|
adminandClaude Opus 5
|
6e3e8d23c5 |
Combining vector shapes keeps the curves it did not cut (#208)
Until now every boolean operation on vector art threw the curves away. The shapes were flattened into line segments, the operation ran on those segments, and the result came back as a polygon. A circle that nothing even touched came out of an Exclude with 40 corners instead of its four points and handles. Now the result carries the original curves again, and only the edges the operation really drew stay straight. - While flattening, every sample point remembers which curve it came from and where on that curve it sits. The operation runs on the same segments as before, so nothing about the boolean result changes. - When a run of result points walks along one of those curves, that piece is cut out of the original curve again with de Casteljau. This is an exact cut, not a fit through the points. - An edge whose two ends sit on the same curve but whose middle does not follow it stays straight. Without that rule a cut-off cap would bulge back into the shape. - A closed contour keeps at least three points, so a lens shape does not collapse into two. - The rule lives in its own module, src/core/model/vectorCurves.ts. vectorBoolean.ts only hands it the table of sampled points. Findings: - The two intersection points themselves sit up to 0.09 canvas pixels off the original curve. Clipper cuts the flattened segments, not the curves, so this was the case before as well and stays inside the 0.2 px sampling tolerance. - Refining those points with Newton steps on the curve was measured and gave nothing back (0.02952 px instead of 0.02978 px), because the error is in the intersection itself and not in the estimate. The code was removed again. - Theme.ts rowHeight 22 -> 24 is in this commit. The change is not mine, it was already in the working tree; gate and module 47 are green with it. Tests: - ./check.sh: 143 files, 2133 tests, exit 0 (was 2127) - e2e module 47 on a fresh seed: all passed, 59 checks (was 57) - 6 new unit tests: untouched circle, untouched side plus a straight cut edge, a cut in the middle of a curve, a chord crossing one curve twice, two overlapping circles, the three point minimum - 2 new e2e checks in block 47p: point count and handles from the model, plus a pixel pair on the round edge - Four counter-proofs, all bite: no reconstruction -> 5 unit tests and the e2e check fail with 35 points and nothing curved; no chord test -> the cap bulges back to 28.28; no three point rule -> the lens test fails; handles halved -> the pixel check reads background Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |