The two "session cookie survives closing and relaunching storebound's own
profile" tests failed with Playwright's headless shell, which ignores
--restore-last-session. They now live in tests/console-persist.test.ts
and launch the same browser kind console fill uses for its profile (the
resolver's "visible" kind: Google Chrome, else the full Chromium; headless
here, so no window). They skip with the reason only when neither is
installed. All other browser tests keep the fast headless shell. The file
is HEAVY in tools/ci-plan.ts, so it never shares a shard with another
browser suite.
`console fill --fresh-profile` failed with "Executable doesn't exist ...
chromium-1187" when only Playwright's headless shell was downloaded,
although Google Chrome was installed (also with STOREBOUND_CHROME set).
The Chrome launch did run, but failed because an earlier storebound run
still held the profile, and the fallback rethrew the FIRST error.
Every launch (renders, captures, console profile and copy modes, the
desktop sidecar) now goes through one resolver that checks the exact
executable before trying it:
- headless: STOREBOUND_CHROME, headless shell, full Chromium, Chrome
- visible / persistent: STOREBOUND_CHROME, Google Chrome, full Chromium
(channel "chromium"); never the headless shell
- copy mode: Google Chrome only
A missing browser is skipped; an installed one that fails reports its own
error; nothing usable names what was looked for plus the doctor /
playwright install hint. A profile whose SingletonLock names a live
process is reported before launching.
Root cause: the per-character font probes were inline spans side by side, so Chrome shaped f + i into the fi ligature across them and the lone i probe reported no glyph of ours. Each probe is now its own block. The Google Fonts text= subsets keep the GSUB ligatures; regression test with a real Montserrat text= subset.
--help / -h prints the command's usage and exits 0 before any config, render or network; unknown flags exit 2 with the usage (cli/data/flags.json from tools/gen-flags.ts). Command tables moved to cli/commands.ts; ads apple reachable again; shots render --scenes rejects unknown ids at once.
The afterAll hook of tests/console-fill.test.ts timed out (30 s) in 2 of 3
runs with 10 CPU hogs. Cause: Google Chrome (the fallback when playwright's
Chromium is not downloaded) hangs in its own shutdown after loading http
pages until its 10 s teardown watchdog fires; under load that passed 30 s.
- tests/helpers/browser.ts: one shared launch per file as a browser server
with a 60 s launch timeout, killed at the end (throwaway profile, nothing
to flush); a launch timeout fails instead of silently skipping; contexts
and engine runs tracked per test and closed in afterEach (capped);
event-driven waits for engine log lines
- console-fill: explicit test timeouts, slow tests marked, waits expected
to succeed get a long cap (they return at once), sign-in / watching
navTimeoutMs 1.5 s -> 5 s, persistent-profile Chromium closed in finally,
temp account dirs removed
- site-overflow, console-session: same plumbing (site-overflow's afterAll
had bun's 5 s default)
- console session: launchTimeoutMs option (tests only)
- aso competitors: top apps per keyword from the iTunes Search API (name, developer,
rating count, release / update dates, localized genre), coverage gaps (words of 2+
developers), new entrants vs the last snapshot (.storebound/aso/competitors/)
- aso suggest: candidates from own text, competitor names, localized category names and
Apple Ads keyword suggestions (POST /v1/suggestions/keywords/query, relative 0-100
popularity, only with credentials); 100-byte keyword field per locale as a diff,
--write marks it a draft that check --aso and aso audit report
- aso audit: check --aso, free keyword room, rank trend arrows, gaps, Play short
description / density, App Store conversion, PPO / Play experiment ideas, next 3 actions
- MCP: read-only aso_competitors, aso_suggest, aso_audit; desktop ASO card (visx rank chart)
- docs: README, aso-basics, store-pitfalls (Search API order, no Play search API), CHANGELOG;
demo-app listing files; tests on recorded Search API and mocked Ads answers
Web capture runs frames in parallel browser contexts (default from the CPU count, lower for GPU-heavy capture),
logs n/total + ETA and goto / waitFor / eval / delay / screenshot per frame, writes .storebound/capture-report.json
with the slowest frames and steps. shots.capture.session 'per-language' boots the app once per language and device
and switches shots via pushState / popstate (navigate 'spa') or goto (navigate 'reload'), falling back to a fresh
page when a precondition fails. .storebound/capture-cache.json skips frames with unchanged inputs; --only-missing,
--force. shots render renders with --concurrency pages and async ffmpeg.
Lilita One lacks most of Latin Extended-A (Turkish s-cedilla, g-breve, dotted I ...), so Turkish titles were drawn
with a silent browser fallback while shots check passed. shots fonts now reads the cmap of every face (pure TS,
TTF/OTF/WOFF/WOFF2), adds a Noto fallback per missing character or fails with language + characters; render
measures which face drew each character (CDP probes) and records it in report.json; check errors on fallbacks the
config did not choose (shots.fonts.fallback) and on system-font glyphs. Default game title font: Paytone One.
Portfolio (every app, both stores, rollout %, crash sparklines), the Growth and
Health charts with a real hover sweep (one captured frame per hovered day, the
pointer follows the recorded positions), the Prices matrix (both stores per
country, gaps marked), then v5's release timeline and 0 blockers. All frames are
real fixture-mode captures (desktop.ts). Tests: scenes 4-5 s, 50-60 s total,
captions >= 3 s, every desktop frame captured by desktop.ts and present.
