* test: migrate test app to expo dev client
* docs: align test app device targeting
* docs: clarify dev client setup tradeoffs
* docs: remove stale sdk reference
* feat: expose web network dump through agent-browser
* fix: reduce web network mapper complexity
* fix: gate web network headers by include mode
* test: assert compact web network summary
* refactor: simplify web network dump mapping
* refactor: trim web network coverage
* chore: remove test-only dead exports
Remove 13 production exports that were consumed only by their own unit
tests, along with those tests. Found via a fallow production-vs-default
unused-export differential (exports unused by production but kept "live"
by a test), then each candidate was reachability-traced and the verdict
adversarially re-verified to confirm zero production use (internal,
re-export, registry, or dynamic dispatch).
Removed exports + their dedicated tests:
- backend.ts: hasBackendCapability, hasBackendEscapeHatch (+ orphaned
BACKEND_CAPABILITY_ESCAPE_HATCH_METHODS) — cascade from runtime change
- runtime.ts: assertBackendCapabilityAllowed
- command-catalog/capabilities: (kept — see below)
- commands/cli-grammar/common.ts: commandNameSet (no test; pure dead)
- compat/maestro/support-matrix.ts: MAESTRO_COMPAT_UNSUPPORTED_CAPABILITIES
- core/dispatch-resolve.ts: resolveIosDevice
- daemon-client.ts: openApp (standalone fn; Backend.openApp untouched)
- daemon/daemon-command-registry.ts: listDaemonHandlerCommands
- daemon/post-gesture-stabilization.ts: capturePostGestureStabilizedSnapshot
- platforms/android/scroll-hints.ts: annotateAndroidScrollableContentHints
- platforms/ios/runner-session.ts: stopRunnerSession
- platforms/ios/runner-xctestrun-products.ts: xctestrunReferencesExistingProducts
- replay/script.ts: parseReplayScript
- utils/cli-option-schema.ts: getOptionSpecForToken
- utils/finders.ts: findNodeByLocator
Tests whose real subject is live code (handler routing, snapshot status,
writeReplayScript, maestro doc-sync) were rewritten to drop the dead
symbol while keeping coverage, not deleted.
typecheck, lint, and the unit suite (2568 tests) pass; fallow
production unused-exports drops 231 -> 218 with no new dead code.
* test: restore replay script parser coverage
* test: re-point replay parser cases to parseReplayScriptDetailed
Address review on #836: parseReplayScript was a thin wrapper, but the
parser it drove (parseReplayScriptDetailed) is live via
compat/replay-input.ts and compat/maestro/export-flow.ts. Restore the
record/screenshot/snapshot/gesture/type parsing assertions deleted with
the wrapper, re-pointed to parseReplayScriptDetailed(...).actions, so
.ad parser regressions (--hide-touches/--fullscreen/--force-full,
gesture positionals, --delay-ms) stay covered.
* Revert "test: re-point replay parser cases to parseReplayScriptDetailed"
This reverts commit 20a01e8641.
* feat(recording): make iOS export quality configurable
Wire the existing recording-export-quality enum through the record command
down to the Swift export preset. Adds a `--export-quality <medium|high>`
option for iOS recordings that controls the AVAssetExportSession preset used
when a recording is re-encoded.
`medium` stays the default and selects AVAssetExportPresetMediumQuality, which
preserves the fast simulator-friendly export. `high` opts into
AVAssetExportPresetHighestQuality for evidence-grade output. This is separate
from the existing integer `--quality <5-10>` capture flag that scales render
resolution.
Closes#568
* fix(recording): apply export quality to touch-overlay export path
The --export-quality flag was only wired into the resize export path. The
touch-overlay re-encode (finalizeRecordingOverlay -> overlayRecordingTouches ->
recording-overlay.swift) ignored it and always picked AVAssetExportPresetMediumQuality,
so record stop with --export-quality high had no effect when the stop path
re-encodes only to burn in touch overlays.
Thread the recording's exportQuality through finalizeRecordingOverlay and
overlayRecordingTouches, pass it as --export-quality to recording-overlay.swift,
and resolve the preset there via the same exportPresetName() helper used by
recording-resize.swift. Medium stays the default when the arg is absent, so
behavior is unchanged for callers that do not set it.
* feat: align recording quality and size flags
---------
Co-authored-by: Michał Pierzchała <thymikee@gmail.com>
Michał Pierzchała