Patterns could only be heard one at a time, in their own tab or from
a behavior. A song lays patterns and sounds out on tracks along one
time axis and plays them through the scheduler of #332, the third
piece of the music tools of #330.
- New asset type song under the heading Songs, with four empty tracks,
120 BPM in 4/4 and a grid of a bar. Tempo, meter and grid are rows
in the inspector. Tracks, items and the loop are the asset's data,
held to their rules on load and in every command.
- A track has a name, volume, pan, mute and solo. Each is heard on a
bus of its own, so a change reaches what is sounding at once.
- A pattern item repeats its pattern for as long as it is, with a
transpose and another instrument if wanted. A sound item plays its
sound as a clip at its own speed, round and round, through a sound
FX if wanted. Both have a volume on top of their track's.
- Cut back at the left end or split, an item starts as far into what
it plays as the cut is: ticks into a pattern, seconds into a sound.
- The tab shows the heads with M, S, volume and pan, the ruler of the
Pattern tab and the items, a pattern with its notes, a sound with
its waveform, a missing asset red. Patterns and sounds are dragged
in from the Assets panel. Body, right and left end move, lengthen
and cut back, Alt leaves the grid. Ctrl+C/V/D/A, Ctrl+E splits
where playing starts, Delete, arrows. A double-click opens a
pattern item in its tab.
- What is picked is a new selection of its own, so the inspector
shows the item or the track, with a button to delete it.
- The scheduler plans clips beside notes. Where the playhead starts
or jumps back inside a clip, it is taken up that far into its
sound. A clip fades out at the end of its item or of the loop.
- The editor plays one piece at a time: a pattern started stops the
song and the other way round.
- The rules are pure functions in core/music/song, editor/song/
songView and songEdit.
Findings:
- The suites of 4ea752f and 2a3b90b were red at e2e 54: the Pattern
pushed Struct below the fold of the + menu. The module now scrolls
to it, as e2e 47 does for Vector art. It also left the menu open,
which made e2e 55 fail behind it.
- The Assets panel only dragged assets a scene can place. Patterns
and sounds are dragged now as well, and a scene says why it does
not take one instead of calling every asset a scene.
Tests:
- song.test.ts (new, 14) and song-edit.test.ts (new, 9): model,
round trip, sequence of patterns and clips, mute and solo, cuts,
clips on the scheduler, moving, sizing, trimming, splitting,
duplicating, hits, head parts. sound-library.test.ts +5: clip
offset, end fade, pitched clip, track bus, peaks.
- e2e 65 (new) on melektaus-atx: ALL PASSED, 18 checks. e2e 54 and
55 with the fix: ALL PASSED, 33 checks. e2e 4 with 30 asset
types: ALL PASSED, 62 checks.
- Gate on melektaus-atx: CHECK EXIT 0, 250 test files, 3884 tests.
- Counter-checks: no taking up of a clip, no pattern offset, no clip
end each fail 3, 1 and 1 tests.
- Probe on stein2: the tab with four tracks, before and while
playing, looked at in two screenshots.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The wheel and the Fit button used to jump to the new view. Now the
view takes a short way there, 200 ms by default, everywhere the editor
zooms. The zoom steps stay the same, only the change is shown.
- viewTween: tweenView is the way between two views as a pure
function. The scale moves by equal ratios, and the world point that
sits at the same place on screen in both views stays there. After a
wheel step that is the point under the pointer.
- ViewTween drives it on the ticker. A surface hands it reading,
writing and drawing its own view and wraps its wheel and Fit code in
change(). A step during a transition starts from where that one was
heading, so quick steps add up to exactly as many steps.
- Anything else that moves the view ends the transition. A pan jumps
to the target first, and saving a view saves the target.
- Wired into FreeViewport (Scene, Container, animation preview),
SheetCanvas (Texture, Curve, Pixel art, Vector art), GraphCanvas,
panAndZoom (Tile board, plugin node board) and the timeline.
- Menu > General settings... sets the time in ms, 0 jumps at once. It
is kept in this browser like the panel sizes, not in the project,
and it is not an undo step.
- A document that opens still comes up framed at once. The animation
preview frames a new subject itself instead of calling fit().
- The pixel art readout names the step the view is heading for.
Measured in the scene view at 200 ms on melektaus-atx, where every
frame of the transition is a full redraw:
- 500 sprites: redraw 6.7 ms, 16 frames, mean 18 ms
- 1000 sprites: redraw 13.0 ms, 9 frames, mean 27 ms
- 2000 sprites: redraw 32.2 ms, 5 frames, mean 58 ms
Tests:
- The harness sets 0 ms on every load, so the existing modules read
the step and not the way.
- e2e 46i samples one wheel step at 1000 ms a frame later: zoom 1.033
between 1 and 2, the point under the pointer kept at (134, -40). It
lands on the view a 0 ms jump reaches.
- e2e 46 and 49 expect the new menu entry.
- Gate (atx): 234 files, 3563 tests, CHECK EXIT 0 (+15)
- e2e 46,49 (atx): ALL PASSED, 43 checks (+3, sum 1735)
- Three counter-checks fail view-tween.test.ts as expected. e2e 46i
fails when the scene view does not carry its centre along.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Check 56a-ii clicks the read-only value of a private variable and types
into it. A read-only field takes no focus, so typeText found no input
and threw when the module ran alone. In the full run the keystroke went
into whatever input an earlier module had left focused, and the check
was green by luck.
- New harness helper typeBlind sends the keystrokes without checking
where they land. It is meant for keystrokes that should be refused.
- 56a-ii now also checks that no field holds the focus after the click
(__gm.focusedRect() is null), the same way module 28 checks its
read-only Asset row.
- No other module types into a field that is meant to be locked.
Tests:
- e2e 56 alone: ALL PASSED, 10 checks (was E2E EXIT 1 before)
- e2e 55,56: ALL PASSED, 27 checks
- Counter-check with the second lock click left out: 56a-ii fails
with a focused field and timer 9, three later checks follow
- Gate: 229 files, 3514 tests, CHECK EXIT 0
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On a screen scaled to 125 % or 150 % the editor looked soft beside
a page the browser draws itself. The layout counts in whole CSS
pixels, and most of those fall between two device pixels there. The
GPU averages both: text loses its edges, and a 1 px line becomes a
pixel and a half of mixed colour. Now text and boxes are rounded to
device pixels, and a line is one device pixel wide.
- ui/pixelGrid.ts holds the rule. crispText() makes a text with
Pixi's roundPixels, crisp() does the same for a Graphics of
upright boxes, and hairline() is the width of one device pixel.
- Every text of the interface is made with crispText() instead of
new Text().
- crisp() only rounds where the scale is not a whole number. At
100 % and 200 % nothing changes.
- drawControlBox, drawHeaderBar and drawSectionBand draw their
lines a hairline wide and flush with the edge. The widgets that
draw lines of their own do the same.
- Two bands share exactly the line they draw, and a column still
ends on a whole pixel.
- The stage is nudged by 1/32 px where the scale is not whole. At
150 % every odd pixel is an exact half, and without the nudge one
band came out a row shorter than its neighbours.
- Curves, icons and everything of the game stay unrounded. Rounding
their points would bend them.
- The harness takes deviceScaleFactor as an option, and the hooks
deviceColumn, edgeContrast and pixelGrid read the frame as the
screen gets it.
Not in here: the subpixel smoothing a browser gives its own text.
Pixi draws a text into a transparent picture first, and there only
greyscale smoothing is possible.
Measured at 150 %, edge contrast before and after: top bar 31.5 to
47.8, the lists on the left 68.6 to 86.4, inspector 27.6 to 38.4.
At 100 % it is 46.5, 90.1 and 37.9, and the picture there is the
same as before pixel for pixel, apart from the fps display.
Tests:
- Gate on melektaus-atx: 206 files, 3114 tests (3104 + 10), exit 0
- Full run on melektaus-atx: 1655 of 1656 checks passed, smokes
24/19/21. The one that failed was a condition of the new step
62 itself: it scrolled an inspector that had nothing to scroll.
The step now makes the window lower for that reading.
- e2e 61,62 after that fix: 14 checks. The full run was not
repeated, and CI is switched off for now.
- Ten unit tests, nine in tests/ui/pixel-grid.test.ts
- Eight e2e checks in the new step 62, which runs at 150 %
- Eleven unit counter-probes fail tests
- Four browser counter-probes fail their checks. They ran before
the fix to step 62 and were not repeated for it.
- Screenshots looked at for 125 % and 150 %
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
stein2 should stop carrying the full e2e run. The new runner lives in
the WSL2 distro game-maker-ci on melektaus-atx and uses the same image
and fonts. Its GPU sits behind /dev/dxg and Mesa's d3d12 driver, which
headless chromium cannot reach, so the harness gets a second way to
start the browser.
- GAME_MAKER_BROWSER=wsl starts an invisible Xvfb and an X11 chromium
on ANGLE/GL. Without the variable, stein2 keeps the headless Vulkan
launch exactly as before.
- The harness logs the renderer and stops on software rendering.
Under WSL a missing d3d12 driver falls back to llvmpipe, which still
offers WebGL 2 and would pass unnoticed.
- The CI image gains xvfb. The font set is unchanged: fc-list gives
the same 149 fonts with the same hash.
- ci.yml runs both jobs on the new label game-maker-wsl. The stein2
runner keeps game-maker-ci, so moving back is one line per job.
Findings:
- Wayland on weston reached the GPU too, but drew at about 40 fps and
let the editor clock stall in long runs (GM_STEPS=1-17: 2 failed).
- Vulkan through a self-built dzn caps ANGLE at ES 2.0 because
VK_EXT_provoking_vertex is missing, so chromium drops the GPU.
- The flaky check in e2e 41 (#282) fails more often under WSL: two of
three solo runs were red.
Tests:
- ./check.sh: 195 files, 2823 tests passed
- WSL container, GM_STEPS=1-17: ALL PASSED, 553 checks
- WSL container, full run: 1589 of 1590, the one FAIL is #282
- WSL container, smokes: 24 / 19 / 21 checks
- stein2 headless, e2e 39+40: ALL PASSED, 100 checks
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The game did get Tab, but the editor moved the keyboard focus to a
button at the same time. The next Space then pressed that button
instead of reaching the game. The key field of On key down could not
capture Tab at all, so an inventory on Tab could not be built.
- FocusManager asks tabBelongsElsewhere before it moves the focus.
The editor answers yes while the Game tab is in front, so Tab stays
with the game as long as no field has the focus.
- A widget can claim Tab through takesTab. The key capture button
does so while it is capturing, so On key down accepts Tab.
- The harness knows the key Tab.
Found on the way:
- The first version of check 20g left the Behavior tab in front, and
module 21 then failed in the chained run. 20g now goes back.
Tests:
- ./check.sh: 2746 tests, CHECK EXIT 0.
- e2e 20 has two new checks: Tab reaches the game and leaves the
keyboard with it, the key field captures Tab. Both failed against
the old bundle.
- GM_STEPS=19,20: all passed, 44 checks. The full local run with all
three topics of this push passed with 1543 checks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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 056a622, +2).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The test browser keeps Chrome's default profile, disk cache included,
and nginx sends no Cache-Control header. Right after `npm run build` a
run could load the index.html of the previous build and test old code.
It showed up while measuring the scene view pan: a new debug hook
returned NaN although server and bundle already had it.
- signIn() turns the network cache off via
Network.setCacheDisabled before the first page load.
Tests:
- GM_STEPS=5 (fresh seed): ALL PASSED, 10 checks.
- ./check.sh: CHECK EXIT 0, 2080 tests.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CI has been red since 66d6aaf (tasks 221 and 223, two failures each).
Both failures come from the tests added with #130, not from the
editor.
- Module 04 still expected 19 asset types in the + menu. Pixel art
and Palette make it 21.
- Module 48 died with "reading 'assets'". The password change in
module 45 ends every other session of the e2e account, including
the one the tools had cached, so api() got a 401. api() now drops
the cached session and signs in once more on a 401.
Tests:
- GM_STEPS=4,45,48 (fresh seed): ALL PASSED, 102 checks.
- Counter-check: without the retry module 48 dies with exactly the
CI error.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Until now only the editor toolbar could pause a running game. A pause
menu inside the game had no way to stop the world while its own graph
kept running, and there was no slow motion.
- Five new nodes: Pause game, Resume game, Is game paused, Set time
scale and Time scale. Pause and resume take effect on the next frame.
- A game pause stops world time, physics, animations, particles and
shader time. Only behaviors with the new "Run while paused" switch
keep running.
- Set time scale 0 freezes the world but graphs keep running, as
agreed in the issue.
- Play and Stop reset the time scale to 1, Load scene keeps it. The
pause belongs to the scene, so Load scene and Restart unpause.
- The toolbar pause stays separate and also freezes the menu. When the
game pauses itself the status line reads "The game paused itself".
- While paused, collision events are not reported, sounds keep
playing, and the editor run ignores both nodes with a log line.
Findings:
- The first status line was longer than any before and got cut with
an ellipsis. It is shorter now.
- The e2e harness did not know the key "p". It does now.
Tests:
- ./check.sh: 133 files, 1968 tests, exit 0 (before 132 / 1952)
- npm run build: exit 0
- GM_STEPS=18 fresh: ALL PASSED, 33 checks (before 26)
- 16 new unit tests: 8 catalog cases, game-pause.test.ts, editor run
- 7 new e2e checks in block 18i, driven by real keys in the game
- Six counter-checks all bite: behavior filter, collision lock, world
time, editor run, clock reset and the game clock in the browser
- Check total 1217 in README is computed (1210 + 7), CI measures it
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The suite spent 88% of its step time asleep: 564 s of steps, 496 s of
them in sleep(n) - fixed spans estimated by hand and then copied from
step to step. That is why it was slow, and why it flickered.
The signal is afterFrames(count, timeout) in ui/render.ts, exposed as
__gm.afterFrames(). It wakes the loop itself, because an editor that
draws nothing would never redeem a promise of a frame, and it gives up
after a timeout instead of hanging the run. settle() in the harness
waits two drawn frames (~32 ms) instead of 150-300 ms and falls back to
sleep(150) when the hook is missing (an old build): silently not
waiting would make every later check a coin toss.
Every action goes through it now - click, doubleClick, wheel, drag,
dragAsset, press, typeText, select, mutate, undo, redo - and with that
about 550 sleep() calls behind those actions became redundant, helpers
that are nothing but a mutate in a coat included. What a frame cannot
answer got its own condition instead: waitUntil(expr), play()/stop()
waiting on __gm.state.mode rather than on 300 ms, saved() for the
autosave (the two hand-written polling loops sampled at 300 ms for
something that arrives after a second), and showTab() as a real tool -
47 places called evaluate("__gm.showTab(...)") raw, and 24 of them
waited for nothing at all and read the old panel.
564 s -> 243 s of step time, the waiting share 88% -> 76%. Three kinds
of wait were kept, and each one bit when it was taken away: loading
bytes (textures, fonts), game time that a repaint cannot stand in for
(time scale, shader and particle clocks, the live values a running
graph carries), and the inertia the scroll step measures on purpose.
Two things the new speed uncovered, both real: 25-animation read the
frame menu unfiltered and had frozen an incomplete list (the CI failure
on c990dbc), and 33-render-on-change measured idleness while the
autosave of the step before was still painting - it waits for
saveStatus to leave dirty/saving now, so it measures that nothing
happens any more instead of that it waited long enough.
The renderer is taken off the ticker; the ticker itself keeps running, so
the clock, a live world and the panels still get their step. Drawing now
happens inside a 200 ms trailing window that every pointer, wheel and key
event, every Widget.invalidate/resize, every Stack.relayout, every
EditorState.emit and both library onChange hooks reopen. The window rather
than a single frame is what makes the ~70 direct redraw() calls that never
pass through invalidate() safe, along with repaints an event only starts:
a microtask, a texture upload, a resolved font.
Editor.movesByItself() keeps asking for every frame while Play or an editor
run is on, a time shader is on screen, or the particle or animation preview
is showing.
The FPS counter now counts drawn frames over the time the renderer was
awake. Idle seconds are not slow seconds, so it answers "how smooth while
it moves" and holds its last answer instead of falling to 0 at rest.
Measured on the built stand, 20 s per variant, CPU of all chromium
processes from /proc: kept awake 75.5 % of a core at 60 fps, render on
change 8.8 % with 0 frames drawn, awake again 83.9 % as an order probe.
No colour check can see a missing repaint: pixelAt extracts from the stage
on demand and is just as green on a screen frozen a minute ago. Step 33
watches the new __gm.renderStats() counter instead.
check() appended the detail by string concatenation, so an object
argument showed up as "[object Object]" in the log (27-console.mjs).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A log layer (src/editor/console/log.ts) keeps a 500-line ring buffer and
wraps console.log/info/debug/warn/error, still calling the original. Notices
and behaviour runtime errors go straight to the layer instead of through
console, so editor tests stay quiet and errors can render red. world.log is
untouched: it is per-run structured state that runtime tests and e2e read
directly, not a text stream.
The panel sits under the visible tab, resizable with Splitter('y'), with a
CONSOLE header and a clear button. Each line carries its wall-clock time in
its own dimmed column, so locate() still finds the message text.
The taller centre area drops the graph fit zoom to 0.60, which broke six e2e
checks that assumed one screen pixel is one graph unit and that a fixed
offset from the canvas centre is empty. They now ask __gm.emptySpot(),
__gm.graph.emptySpot() and __gm.graph.zoom() instead of guessing.
locate/locateAll added the unscaled Pixi text width to the global position,
so click points in the scaled graph canvas landed too far right; they use
getBounds() now.
check.sh: 831 tests. Full e2e: ALL PASSED - 671 checks. Smokes 24/19/20.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
tools/seed-actor.mjs builds the project Actor demo: it uploads
tools/assets/Actor1.png, slices the 576x384 sheet into the 96 cells of a
12x8 grid (48x48 each, no remainder) and wears the character in the top
left corner. Walking is four ping-pong animations of three frames, which
is the original RPG-Maker walk cycle; standing is the middle frame of the
row, not an animation.
The graph pastes walk- or idle- in front of the direction faced and wears
the result, so eight cases need no branch tree - the List node only holds
numbers. Which way it looks is the key pressed last, corrected whenever
only one axis is still held, or letting go of one key of a diagonal would
walk backwards. Axis input already normalizes its vector.
tools/lib/projects.mjs gains uploadTexture, since data/textures/ is
git-ignored and a seed that ships its picture has to send it. The harness
gains w/a/s/d and keyDown/keyUp halves, because holdKey cannot express two
presses that overlap - which is what a diagonal is.
tools/smoke-actor.mjs plays it in 19 checks and reads the drawn pixels at
the camera edge, where the bounds hold him walking on the spot: three
distinct frames walking, one standing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The graph was only locked by five edit-mode guards in GraphCanvas. The
running game reads the graph by reference each frame, so a new node, edge,
changed value, disconnect or undo takes effect from the next frame. Stop
discards these edits like everything else changed in play. A new On start
node does not fire for a behavior that has already started.
During play every data output shows the value it last produced, in the
highlight colour and refreshed ten times a second. Values are kept per
graph, because node ids are only unique within one; with several
instances the most recent value wins. The Variables node reads the live
variable bag, so only its own outputs are kept, as a snapshot.
Each time an exec edge fires, a bead runs along it in 0.3 s, at most one
every 0.15 s per edge, following the same curve the edge is drawn with.
Beads run on game time and freeze while paused; the game logic is
unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Split GraphCanvas (1076 -> 831 lines) into EdgeLayer and graphMenus;
the port and marker menus were the same list and are now one function.
Value specs of the inspector and the list tab share valueSpecs.ts, and
the remaining copies (compound commands, type glyphs, typed asset
lookup, name rule, play/editor world setup, marquee) are merged.
Four copies had already drifted apart; merging them fixes:
- "Disconnect N links" in the port menu is one undo step, not N
- "+ Add collider" is hidden on a placed container instance, as in
the hierarchy menu, instead of always failing
- a bool property coerced from the text "false" becomes false
- a port command on an unknown node type reports instead of crashing
Comment-only edits leave the bundle byte-identical. README corrected in
35 places (step limit, jump-run clock, suite duration, 409 body, ...).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Steps 18 and 20 both hand-rolled the same object-plus-behavior scaffold, down
to lane offsets that nothing kept distinct. Step 19 claimed in a comment that it
cleaned up after itself but never checked it – the only one of the four that
did not.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A Tetris example project needs to draw 200 board cells and scan rows
for a full line. Neither is expressible without iteration, so the node
system gets a Repeat node plus the missing text side: Substring, Find
text and To number. The interpreter drops its data cache once per
round -- otherwise every round would recompute with the first round's
values.
Object by name takes its name as an input now, not just as a setting;
addressing "c" + index is impossible otherwise. An unconnected input
reads the stored setting, so saved projects are untouched.
The board itself is a 200-character text, "." for empty. That makes
both operations of Tetris a single expression instead of a loop, and
each frame attempts exactly one move -- rotate beats sideways beats
fall -- so the fit check exists once in the whole graph.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Three things the graph view did differently from the scene view, or not
at all:
Panning moves to the middle button, which works from anywhere – a node
under the pointer included, since NodeView now lets that one button
bubble up. Shift-drag on empty canvas pans too. The left button pulls a
rubber band instead and no longer pans.
Nodes can be selected several at a time: by band, by ctrl/shift-click,
moved together, deleted together, each as one undo step. The selection
model gains `nodeIds` alongside the objects' `objectIds`.
Fit really fits. It used to pin the zoom at 1 and merely centre, so a
graph larger than the view always left nodes hanging out, and it sized
every node with one guessed height although real ones vary with their
port count. Both sums now live in editor/graph/framing.ts as pure
functions, which is what made the defect testable at all.
The band selects any node it overlaps rather than one whose position it
contains, unlike the scene view: nodes are large rectangles, and a band
drawn visibly across one has to catch it.
Pixi only fills `detail` on the `click` event it synthesises; on
`pointerdown` it is always 0, so every `e.detail === 2` branch was dead
code. Asset rows, hierarchy folding and select-all in text fields never
responded to a double-click.
A shared click counter in src/ui/clicks.ts replaces the three readings of
`e.detail`. A finished double-click ends the run, so two in a row count
2 and 2 rather than 3 and 4.
The harness could only send `clickCount: 1`, which is why no test caught
this: every "activate" check went through Enter. It can send real
double-clicks now, and five checks drive the three paths separately.
The context menu entry is gone; dragging replaces it entirely. The drop
target is the view under the pointer, not the active tab, so the same
gesture works for scenes and for container templates.
While dragging, an outline shows the footprint the object will have plus
a crosshair on its origin, turning red when the drop would be refused.
Preview and command share one function (placementRejection), so the
outline cannot promise something the drop then rejects.
A left-press used to select the list row immediately. Since the container
tab follows the selection, it switched to the very container being dragged
and every drop would have nested it inside itself. The list now selects on
release without a drag: picking something up is not the same as choosing it.
The harness forced software WebGL (--disable-gpu --use-gl=swiftshader) on every host.
Measured here on an idle editor page over 20 s: that costs 373% of a core at 8 fps,
while the Intel HD 630 via ANGLE/Vulkan costs 69% at 56 fps. The full suite now runs
in ~124 s. Note the app is not rendering 60 fps as previously assumed - in software
it only reaches 8, and a single frame costs more than any throttle interval would save,
which is why capping the framerate had measured no effect.
--use-angle=gl and --use-gl=egl look cheaper still but get no WebGL context at all,
so the app never renders. They are not usable and are named in the comment as such.
There is deliberately no software fallback: a host without a usable GPU should fail
loudly rather than drag the machine to a standstill. None of the 267 checks reads the
rendered image anyway - they read window.__gm.locate() and the debug hook, and the
screenshots assert nothing - but Pixi needs WebGL2 to run at all.
The browser launch moves out of harness.mjs into tools/lib/browser.mjs, which also
supports GM_REMOTE=user@host via an SSH-forwarded CDP port. That path is unfinished:
the tunnel degraded until the suite crawled at ~25 s per check, and forcing software
WebGL on a desktop with a real card brought it to its knees. README says so plainly
rather than advertising it.
Also: the jump-run HUD check asserted a fixed "Coins 1/10", which fails on a faster
host where the same walk sweeps up two coins. It now asserts the HUD against the count
the run actually reached - the real guarantee is that HUD and state agree.
Verified: npm run check green (404 tests), 267 e2e ALL PASSED and 24 jump-run checks
ALL PASSED against the GPU flags. The fallback removal leaves the flag set on this host
identical (same flags, different order), so it was not re-run.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
typeText fired Input.insertText per character and never looked back, so a
partially typed value was indistinguishable from a wrong one: a run committed
x=3 for "300" and the failing check blamed the value rather than the three lost
characters.
Trying to reproduce that failed over 21 rounds across four variants (no pause,
400ms between characters, with the preceding scrub, and both commit paths), and
the code rules out the obvious suspects: typing lands in a hidden input, each
character is only a preview, objectSignature omits position so no rebuild is
triggered, and NumberField.setNumber guards write-back with isEditing. Cause
unknown, so this makes the symptom name itself instead of hardening against a
mechanism that may not exist.
Typing stays per character - that exercises the live-preview and search-filter
paths - but now asserts the text is in the focused input, reporting focus and
actual contents when it is not.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A run whose chromium died stopped mid-step and exited with code 0, silently
reporting success for checks it never ran: CDP.send() left its promise pending
when the socket closed, and an unsettled promise does not keep node alive.
Waiting on the server had the mirror problem – a bare fetch keeps its socket
open, so a stalled response hung the run forever.
Every wait can now time out, the socket rejects what it can no longer answer,
and reaching process exit without a summary reports RUN DIED with exit 1.
Chromium's stderr and exit signal are kept, so "killed by SIGKILL" is visible
rather than guessed at.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Two regressions from the template-resolution rework, plus the harness gap
that hid them:
- `instanceTargetOf` decided from the shape of the id. Unpacking keeps the
derived ids, so unpacked content still counted as template content. It now
asks whether the object is spelled out in the holder.
- The parent picker captured its holder when the row was built. Since
`viewOf()` returns a derived copy, the menu listed the objects of that
moment (measured: 40 in the model, 17 in the menu). It resolves on open.
- `evaluate()` swallowed page exceptions, so a mutate that died half-way let
every later check pass or fail for the wrong reason.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A scene now stores only the container object; its content is resolved from
the template on every read, with deviations kept as sparse instanceEdits on
the container object. This drops the propagation bookkeeping entirely --
templateSync.ts is gone and the three structural gaps (add, remove, reparent)
cannot diverge any more.
Values stay editable on an instance, structure does not; Unpack writes the
content into the scene, keeping ids. Play mode bakes containers into the
throwaway copy before entities are built. No migration: formatVersion is 2
and the seeders write the new format.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Tooling:
- Prettier as the single formatting authority (110 columns, single quotes,
measured against the existing style); Markdown excluded because table-cell
padding turns the README glossary into 1000-character lines.
- Type-aware ESLint ending in eslint-config-prettier, so a rule always means
"this is a bug", never "this looks different".
- .editorconfig and an `npm run check` that runs typecheck, lint, format and
tests in one go.
Findings fixed on the first run: three more duplicate imports, four
String(unknown) calls (parameter types narrowed instead of casts), two dead
initialisers, two redundant casts, a console.log for a real error. Two
warnings were false here and are disabled with a reason in place.
Tests 132 -> 269:
- node-catalog.test.ts covers every node with a run(), with a guard that fails
when a new node has no row.
- behavior-events.test.ts covers the ten event nodes through a running World,
checking timing rather than existence.
- Input's state machine (press/release/releaseAll) is reachable without DOM
events now; the DOM handlers are thin adapters.
Compare fell back to == without a configured operator although the schema
default is <; aligned.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
adminandClaude Opus 5.5