Repository navigation
Conversation
# Conflicts: # apps/tester-federation-v2/package.json # pnpm-lock.yaml
…ures-to-mf2-testing-app
…es-to-mf2-testing-app
|
|
@MrMuzyk is attempting to deploy a commit to the Callstack Team on Vercel. A member of the Team first needs to authorize it. |
|
Still in draft. Waits for #1514 |
Apex Sentinel · ✅ Approve2 finding(s): 2 minor. 🟡 Minor (2)
Not in the reviewers' prompt (1 file)
3 findings below 💬 Talking to SentinelIn a finding's thread, reply with:
On this conversation, comment |
There was a problem hiding this comment.
Apex Sentinel: 2 new findings on 46a6035. The verdict and the full list are in the summary comment.
| <Screen testID="symbolication-screen"> | ||
| <ActionButton | ||
| testID="symbolication-probe-button" | ||
| label={frame ? 'Symbolicated' : 'Symbolicate a stack trace'} | ||
| disabled={frame !== undefined} | ||
| onPress={probe} | ||
| /> | ||
| {frame && ( | ||
| <Report | ||
| testID="symbolication-frame-report" | ||
| name="frame" | ||
| value={frame} | ||
| /> | ||
| )} |
There was a problem hiding this comment.
🟡 Minor · Failed symbolication leaves the button reading "Symbolicated" and disabled
When symbolicateStackTrace rejects, the screen reports frame: failed: <message> but the button still reads "Symbolicated" and is disabled — no retry, and the state claims success. frame is a single string holding either a symbolicated frame or an error message, and app/symbolication.tsx derives both the label and disabled from frame !== undefined.
| <Screen testID="symbolication-screen"> | |
| <ActionButton | |
| testID="symbolication-probe-button" | |
| label={frame ? 'Symbolicated' : 'Symbolicate a stack trace'} | |
| disabled={frame !== undefined} | |
| onPress={probe} | |
| /> | |
| {frame && ( | |
| <Report | |
| testID="symbolication-frame-report" | |
| name="frame" | |
| value={frame} | |
| /> | |
| )} | |
| <ActionButton | |
| testID="symbolication-probe-button" | |
| label={frame ? (frame.startsWith('failed:') ? 'Retry' : 'Symbolicated') : 'Symbolicate a stack trace'} | |
| disabled={frame !== undefined && !frame.startsWith('failed:')} | |
| onPress={probe} | |
| /> |
Suggested change
Have the hook return a status (idle | symbolicated | failed) alongside the frame — or separate frame and error — and drive the label and disabled from that, so a failed probe stays tappable and never says "Symbolicated".
Details
persona: maintainability, javascript, senior-developer
Useful? Reply @apex-sentinel-staging like, or @apex-sentinel-staging dislike <why> if it is not.
Reply @apex-sentinel-staging note ... to add context, or @apex-sentinel-staging pivot ... if it is wrong — I will re-check the cited code.
| run: | | ||
| curl -sf --max-time 300 -o /dev/null 'http://localhost:8081/index.bundle?platform=ios&dev=true&lazy=true&minify=false&app=com.callstack.repackexpo&modulesOnly=false&runModule=true' | ||
| curl -sf --max-time 300 -o /dev/null 'http://localhost:8082/ios/mf-manifest.json' | ||
| curl -sf --max-time 300 -o /dev/null 'http://localhost:8083/ios/mf-manifest.json' | ||
| echo 'All three served; compilers are warm.' | ||
|
|
||
| - name: Check shared versions are real semver | ||
| run: | | ||
| for p in 8081 8082; do | ||
| M=$(curl -sf --retry 3 --retry-delay 2 "http://localhost:$p/ios/mf-manifest.json") || { echo "::error::port $p: mf-manifest.json not served"; exit 1; } | ||
| [ "$(jq '.shared | length' <<<"$M")" -gt 0 ] || { echo "::error::port $p: manifest has no shared entries"; exit 1; } | ||
| BAD=$(jq -r '.shared[] | select((.version // "") | test("^[0-9]+\\.[0-9]+\\.[0-9]+") | not) | "\(.name): \(.version)"' <<<"$M") | ||
| [ -z "$BAD" ] || { echo "::error::port $p has non-semver shared versions: $BAD"; exit 1; } | ||
| done | ||
|
|
||
| # SIM_RUNTIME_PREFIX is tied to the Xcode picked in DEVELOPER_DIR | ||
| - name: Boot simulator | ||
| env: | ||
| SIM_RUNTIME_PREFIX: com.apple.CoreSimulator.SimRuntime.iOS-26- |
There was a problem hiding this comment.
🟡 Minor · e2e-expo iOS job boots a simulator before prebuild; simctl create can die on a stale e2e device
xcrun simctl create e2e exits 1 with "device already exists" on any re-run or stateful macOS runner where the e2e device persists in ~/Library/Developer/CoreSimulator/Devices from a prior boot. simctl create is not idempotent, so the step kills the job before any test runs.
Suggested change
Add xcrun simctl delete e2e >/dev/null 2>&1 || true before xcrun simctl create e2e to remove any pre-existing device. Also resolve the runtime via simctl list runtimes -j | jq filtering on $SIM_RUNTIME_PREFIX and assert it is non-null before passing it to create.
Details
persona: senior-developer
Useful? Reply @apex-sentinel-staging like, or @apex-sentinel-staging dislike <why> if it is not.
Reply @apex-sentinel-staging note ... to add context, or @apex-sentinel-staging pivot ... if it is wrong — I will re-check the cited code.
Summary
Covers
apps/tester-expowith e2e tests, so an Expo change no longer means booting the Expo apps by hand. Until now PRs only typechecked it. A second lane,e2e-expo.yml, runs it on the same tooling and rules as the MF2 lane: aFEATURES.tsregistry driving Home, one screen and one spec per feature, every spec owning one distinct way to fail. Nine specs: Expo native module, image, density image, font, local async chunk, Expo-built remote (ExpoWidget), plain Re.Pack remote (OrdinaryWidget), Fast Refresh, symbolication.tester-expo-widget(8082) and MF2'sstart:expo-host-widget(8083). The workflow starts and warms all three;adb reversegets a third port.mainand dispatch — same as the MF2 lane.expo prebuildgenerates the native folders first, since they are not in git.fast-refresheditscomponents/FastRefreshLeaf.tsxon disk and checks that the new text shows up while the counter keeps its value (no full reload). The file is restored infinally.symbolicationsends its own error through RN'sparseErrorStack+symbolicateStackTraceand checks that the first frame maps back toSymbolicationProbe.tsx:<line>; the line is read from the file, never hardcoded.babel.config.cjssetsdisableDeepImportWarnings: true. RN 0.88 turns the first warning into a toast that steals e2e taps, and babel-preset-expo's deep-import lint fires on the linkedpackages/repackand on the probe's RN imports. No LogBox filter, so a new warning still shows.e2e.ymltriggers change from apathsallow-list topaths-ignore: website/**, **/*.md, same as the new workflow. Root files likepnpm-workspace.yamlchange what the apps run, and the allow-list skipped them.agent_context/e2e-oct2026/design.mdis split into shared rules, the MF2 lane and the Expo lane; the Expo part has the coverage table, the break-once record and the traps.Test plan
Demo video:
tester-expo.mp4
CI: https://github.com/MrMuzyk/repack/actions/runs/41818544687 — Android + iOS, artifacts carry junit, screenshots, logcat and dev-server logs on failure.
Locally (boot an emulator/simulator and install the app first —
pnpm repack:android/pnpm repack:iosinapps/tester-expo; Node ≥ 24.8):The Debug build loads JS from all three servers, so they must stay up. Warm them on a fresh start (the
index.bundleandmf-manifest.jsonURLs the workflow uses), otherwise the cold compile eats the boot anchor. After apnpm installthat changes the lockfile, restart them — a running server keeps the old package paths.9/9 green locally on Android (API 35 emulator) and iOS (iPhone 17 Pro simulator). Each spec was broken once on purpose and went red at its own assert — the break-once table in
design.mdlists what was broken.