react native app test suite

[awongh]

No description provided.

[guaka]

[codex] E2E/test-suite review note

Current shape looks good to me for PR use:

• Jest stays the broad, fast default for app behavior coverage.
• Android Maestro remains the automatic native smoke gate for relevant nr-app/E2E changes, and now only runs .maestro/main-map-smoke.yaml on PRs.
• iOS Maestro is intentionally opt-in with the ios-e2e label. Add that label when we want to spend macOS runner time validating iOS; remove it for normal PR iteration.
• The latest workflow changes move iOS to macos-26 with explicit Xcode 26.4.1 selection, which avoids the previous Swift 6.2 vs 6.1 toolchain failure and also avoided the long arm64 runner queue.
• Android local prebuild/Gradle is still the slowest step. We added safe native/Gradle caching and kept clean prebuild as the fallback escape hatch via E2E_REUSE_PREBUILD=0.

I also fixed two E2E service readiness problems seen while validating CI:

• tiny-relay now runs Deno with --no-lock, so the read-only repo mount in CI does not try to write deno.lock.
• nr-bridge now has a compose health check, and the workflow readiness check no longer requires / to return 2xx, because nr-bridge only exposes POST API routes.

Local verification before pushing:

• pnpm --filter nr-app test passed: 54 suites, 208 tests.
• Workflow/compose YAML parsed successfully.
• Local E2E Docker network came up healthy for MongoDB, Mailpit, tiny-relay, and nr-bridge.

I am watching the current GitHub run now. Android is automatic; iOS is running only because the ios-e2e label is present.

[guaka]

[codex] E2E follow-up

Latest failed run gave two useful signals:

• Android KVM/emulator setup worked: the emulator reached device, the APK installed, and the old adb offline problem did not recur. The failure was just the workflow command: cd nr-app was run as a separate emulator-runner script line, so the next pnpm run test:maestro:local executed from the repo root and could not find the script. Fixed by running pnpm --dir nr-app run test:maestro:local -- .maestro/main-map-smoke.yaml.
• iOS got through Xcode 26 build/install again, then failed starting services because .devcontainer/nr-bridge/init-mongo.sh was not executable on the macOS runner. Fixed by invoking it via bash.

So the current retry should get Android into the real smoke flow and iOS past Mongo seeding. The ios-e2e label is still present, so iOS will continue to run on this PR; removing that label should skip iOS for normal iteration.

[guaka]

[codex] I implemented a focused CI stability/speed fix on branch nr-app-test-suite and pushed it to PR #227.

What changed:

• iOS Maestro prewarm in is now non-blocking by default (no hard-fail on -type assertion drift). You can make it strict by setting .
• Prewarm visibility check is now optional via ; default is no explicit ready-id assert.
• now trims Maestro artifact upload patterns to only and uploads explicit directories to avoid huge artifact packs.
• iOS job in this workflow is label-gated by and uses for the PR smoke path.

This should reduce false iOS failures at prewarm and keep artifacts smaller for both Android/iOS runs. Current push is waiting on the latest workflow (run 27510780861) with both android-maestro and ios-maestro currently in progress.

[guaka]

[codex] Quick update after latest Maestro failures: both Android/iOS smoke flows still had startup brittleness because the main flow assumed onboarding was always visible (welcome-get-started-button / identity-skip-button) and Android prewarm was gated on a fixed startup state.\n\nI made the flow more robust so onboarding taps now run only when their elements are visible, and the test now waits for screen-map before continuing. I also made Android Maestro prewarm optional in android-e2e-maestro-local.sh and set the Android CI flow to disable prewarm by default (E2E_PREWARM_ENABLED=0, E2E_PREWARM_REQUIRED=0) to avoid a fragile gate.\n\nI also added ios-e2e label to this PR so iOS job runs only under explicit opt-in as requested.