[nulltask]

Summary

• AUTO checkbox (155bd8d): be-music-preview-select (list header) and be-music-preview-start (under the play button) gain autoPlay and emit auto-play-change. The embed starts it from ?mode= and switches between the AUTO preview and the viewer's play without touching the URL; under the play button it loads the play again in the other mode.
• Lanes down to the seek bar (b9bc65c): on a phone held upright the lanes stopped well above the controls; held sideways (SP or DP) the judgement line sat under the seek bar; a DP chart letterboxed on a phone held upright had the controls over its lanes. previewJudgementInset keeps clear only the controls' 54 CSS px as tall as they get on a 320 px phone plus 16 px, in both orientations, and the preview HUD puts the controls along the bottom of the screen rather than the stage.
• Load before downloading on mobile (c6437bc): a phone or tablet (or a browser asking to save data) shows a Load button with the archive's file name and size (from a HEAD request) and downloads only once it is pressed. /__url-load-proxy answers HEAD with the upstream's Content-Length (unless compressed).
• Docs (EN/JA), changesets for @be-music/app-play, @be-music/player-ui and @be-music/worker-play, and the bench case for the new export.

Tests

• pnpm run check
• vp test run packages/player-ui packages/app-play/src/embed packages/worker-play (new: embedModeOfAutoPlay, embedWaitsToDownload, readDownloadSize / describeDownload, previewJudgementInset, the proxy's HEAD size)
• New embed E2E cases (AUTO checkbox, Load on a touch context) run in CI only.
• Checked by hand in the dev server: the checkbox in the list header (right edge on the rows') and under the play button, DP lanes in 375×700 portrait and 700×340 landscape, and the Load button at 375×812 (HEAD only until pressed).

🤖 Generated with Claude Code

[github-actions]

Exports Benchmark

🔴 3 regressions · 0 improvements · 19 within noise · 888 compared

_{188f6456 → c6437bcb · 3 runs base / 3 runs head · a regression or improvement is a change past ±8% whose runs do not overlap · times are per call (median)}

Regressions (3)

API	Base	Head	Change	Run spread (base / head)
player-ui.resolvePreviewPortrait	13.3 ns	44.7 ns	236.3% slower	±0.3% / ±4.8%
player-ui.resolvePreviewInfoBox	11.4 ns	14.9 ns	31.1% slower	±2.1% / ±0.5%
player-engine.course.stitchNonstopCourse	281 µs	366 µs	30.1% slower	±0.4% / ±16.0%

Changed packages

app-play, player-ui, worker-play: 9 cases compared, median 0.1% slower.

All 9 cases

API	Base	Head	Change	Run spread (base / head)
player-ui.defineBeMusicElements	603 ns	604 ns	0.1% slower	±3.0% / ±8.1%
player-ui.formatPlaybackTime	43.1 ns	42.9 ns	0.5% faster	±0.6% / ±0.2%
player-ui.paintPlayfieldParts	58.9 µs	59.3 µs	0.7% slower	±0.6% / ±0.3%
player-ui.previewContentBottom	6.91 ns	6.94 ns	0.4% slower	±3.0% / ±3.2%
player-ui.resolvePlayfieldParts	2.20 µs	2.18 µs	0.7% faster	±0.5% / ±1.3%
player-ui.resolvePreviewInfoBox	11.4 ns	14.9 ns	31.1% slower 🔴	±2.1% / ±0.5%
player-ui.resolvePreviewPortrait	13.3 ns	44.7 ns	236.3% slower 🔴	±0.3% / ±4.8%
player-ui.resolveSeekTarget	6.28 ns	6.06 ns	3.5% faster	±19.8% / ±2.6%
player-ui.setBeMusicTheme	19.3 ns	19.2 ns	0.8% faster	±1.3% / ±0.3%

New exports (1)

API	Head
player-ui.previewJudgementInset	7.39 ns

Within noise (19): past the threshold, but the runs overlap

API	Base	Head	Change	Run spread (base / head)
skin-beatoraja.scene.createBeatorajaJudgeStateGraph	4.66 µs	2.29 µs	50.8% faster	±48.3% / ±34.0%
player-node-audio.audio-sink.createBrowserAudioSink	4.75 µs	2.91 µs	38.7% faster	±28.9% / ±19.6%
player-browser.setMasterVolume	325 µs	249 µs	23.4% faster	±44.2% / ±37.8%
chart-parser.resolveBMSControlFlowWithDraws	7.16 µs	5.57 µs	22.2% faster	±12.9% / ±17.7%
chart-editor.deleteNote	5.24 µs	6.94 µs	32.6% slower	±7.6% / ±20.2%
chart-editor.addNote	6.62 µs	5.52 µs	16.7% faster	±16.1% / ±18.5%
shared-utils.workerize.invokeWorkerizedFunction	752 ns	632 ns	16.0% faster	±13.3% / ±15.1%
chart-model.createSpeedTimeline	32.1 ns	39.0 ns	21.3% slower	±7.1% / ±12.7%
player-engine.course.prepareNonstopCourse	120 µs	145 µs	21.0% slower	±6.6% / ±15.4%
chart-model.getPlayModeLayouts	12.7 ns	11.0 ns	13.0% faster	±31.5% / ±13.3%
chart-writer.stringifyBMS	49.2 µs	42.8 µs	13.0% faster	±9.3% / ±8.4%
player-engine.replay.resolveRulesetResultLamp	40.7 ns	46.4 ns	14.1% slower	±6.4% / ±12.5%

…and 7 more.

I/O-bound, not judged (19)

These cases touch the filesystem, so their time follows the OS cache and the disk more than the code.

API	Base	Head	Change	Run spread (base / head)
chart-audio-renderer.renderChartFile	6.38 ms	6.61 ms	3.6% slower	±1.7% / ±35.5%
chart-audio-renderer.renderJSON	3.82 ms	3.94 ms	3.2% slower	±3.5% / ±4.2%
chart-audio-renderer.renderSingleSample	8.11 µs	8.05 µs	0.8% faster	±0.4% / ±2.3%
chart-audio-renderer.writeAudioFile	163 µs	202 µs	23.5% slower	±17.6% / ±34.4%
chart-editor.exportChart	210 µs	219 µs	4.4% slower	±2.1% / ±5.9%
chart-editor.importChart	124 µs	134 µs	8.0% slower	±3.9% / ±6.8%
chart-editor.loadJSONFile	141 µs	142 µs	0.4% slower	±4.1% / ±2.8%
chart-editor.saveJSONFile	212 µs	234 µs	10.2% slower	±15.6% / ±5.8%
chart-parser.parseChartFile	116 µs	137 µs	17.4% slower	±11.9% / ±2.3%
player-browser.readFilesIntoBytesMap	4.66 µs	4.64 µs	0.5% faster	±3.2% / ±2.2%
player-terminal.createBgaANSIRenderer	3.26 µs	3.20 µs	1.6% faster	±3.1% / ±2.4%
player-terminal.loadStageFileANSIImage	101 ns	100 ns	0.9% faster	±1.0% / ±0.5%
player-terminal.loadStageFileANSILines	163 ns	164 ns	0.6% slower	±1.2% / ±0.5%
player-terminal.loadTerminalANSIImage	1.03 ms	1.06 ms	3.0% slower	±4.4% / ±2.3%
shared-utils.createFileLogger	196 µs	195 µs	0.5% faster	±2.4% / ±1.7%
shared-utils.readFilesIntoEntryMap	2.32 µs	2.30 µs	0.7% faster	±2.5% / ±0.5%
shared-utils.resolveFirstExistingPath	90.9 µs	92.2 µs	1.4% slower	±4.4% / ±3.6%
shared-utils.workerize.workerize	10.8 ms	10.8 ms	±0.0%	±0.6% / ±0.3%
skin-beatoraja.readFilesIntoBytesMap	19.9 µs	19.7 µs	0.9% faster	±2.4% / ±5.0%

Excluded: too fast to time (74)

A timed sample lasted 1 µs or less on at least one side (usually a base commit from before calls were batched), so its time is timer noise and no change is shown.

API	Base	Head
chart-model.createMeasureBoundariesBeats	72.0 ns	72.0 ns
chart-model.mirrorBeatAtTempoReversal	6.15 ns	6.21 ns
chart-model.parseBmsARGB	272 ns	264 ns
chart-model.parseBMSExBMP	608 ns	608 ns
chart-model.resolveChartMainBPM	107 ns	107 ns
player-browser.applyBeatorajaStretchRect	28.7 ns	28.4 ns
player-browser.checkBrowserCompat	159 ns	160 ns
player-browser.createRandomMixEntry	13.7 ns	13.5 ns
player-browser.destinationToSpriteProps	85.6 ns	85.6 ns
player-browser.GameplayRecorder	150 ns	78.0 ns
player-browser.SceneController	248 ns	256 ns
player-browser.setBrowserWorkerCount	38.4 ns	38.4 ns

…and 62 more.

Run details

Measure	Value
Median change (all compared cases)	-0.03%
Mean change	-0.05%
Overall verdict (median vs. threshold)	unchanged
Compared / within noise / I/O-bound / unchanged	888 / 19 / 19 / 847
Benchmarked / skipped (head)	963 / 38
Time / warmup per case	50 ms / 25 ms
Node / platform	v26.12.1 / linux