ci(release): drop the physical acceptance comment gate

.github/workflows/publish.yml

@@ -21,10 +21,6 @@ on:
         description: 'One successful full CI run ID for this exact SHA'
         required: true
         type: string
-      acceptance_comment:
-        description: 'GitHub issue comment ID containing maintainer physical acceptance JSON'
-        required: true
-        type: string
 
 concurrency:
   group: npm-release
@@ -33,7 +29,6 @@ concurrency:
 permissions:
   contents: read
   actions: read
-  issues: read
 
 jobs:
   release:
@@ -44,13 +39,11 @@ jobs:
     permissions:
       contents: write
       actions: read
-      issues: read
       id-token: write
     env:
       RELEASE_SOURCE: ${{ inputs.source_sha }}
       RELEASE_VERSION: ${{ inputs.version }}
       RELEASE_RUNS: ${{ inputs.ci_runs }}
-      RELEASE_ACCEPTANCE: ${{ inputs.acceptance_comment }}
     steps:
       - uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 # v4.4.0
         with:
@@ -74,13 +67,13 @@ jobs:
       - name: Test readiness gate offline
         run: node --test scripts/release-readiness.test.mjs
 
-      - name: Verify source, actual mobile evidence, package and physical acceptance
+      - name: Verify source, actual mobile evidence and package
         env:
           GH_TOKEN: ${{ github.token }}
         run: >-
           node scripts/release-readiness.mjs
           --source "$RELEASE_SOURCE" --version "$RELEASE_VERSION"
-          --runs "$RELEASE_RUNS" --acceptance "$RELEASE_ACCEPTANCE"
+          --runs "$RELEASE_RUNS"
           --out "$RUNNER_TEMP/verified-release"
 
       - name: Preserve verified release report and tested tarball

docs/releasing.md

@@ -3,13 +3,12 @@
 The prepared `main` revision is the release candidate. `CHANGELOG.md` is the
 release-note source of truth. This procedure does not authorize a version change,
 tag, dispatch, or publication by itself. Historical runs and local or offline
-checks cannot replace a successful run of the prepared source or physical
-acceptance.
+checks cannot replace a successful run of the prepared source.
 
 ## Prepare an unpublished exact version
 
-1. Finish the implementation and its independent review. Final acceptance must
-   use the exact prepared source and version. Release candidates use an explicit
+1. Finish the implementation and its independent review. The full CI run must use
+   the exact prepared source and version. Release candidates use an explicit
    `1.0.0-rc.N`, and the stable release uses `1.0.0`. Never use
    `npm version prerelease` to infer v1.
 2. In a separately reviewed preparation change, use
@@ -25,10 +24,12 @@ acceptance.
    with both mobile suites enabled. The run must completely succeed on attempt 1.
    A failed-jobs rerun is not release evidence. Do not change assertions, retries,
    the 120-second case deadlines, or the 45-minute mobile job deadlines.
+   If the [physical iPhone check](#physical-iphone-check) applies, do it on the
+   same SHA.
 
 RC and stable preparation each change source identity. Evidence for an RC does
-not authorize publishing a subsequently bumped stable package. Repeat full
-acceptance on the stable source. If `main` moves before publication, prepare and
+not authorize publishing a subsequently bumped stable package. Repeat the full
+CI run on the stable source. If `main` moves before publication, prepare and
 test the intended current revision again; never rewrite evidence to match it.
 
 ## Dependency advisories
@@ -138,57 +139,43 @@ complete WDIO summary and every passing case title (including duplicate counts)
 A success badge, hand-entered count or summary without case output cannot pass.
 Missing/expired artifacts or logs and changed reporter formats fail closed.
 Recycling, detach and RTL remain mandatory in this same-source 62-case E2E
-inventory on both platforms. They are automated evidence requirements,
-not extra physical smoke confirmations; removing their manual keys does not remove
-any automated cases, assertions or required jobs.
-
-## Explicit physical acceptance
-
-A repository admin or maintainer must record a JSON issue comment after testing
-the exact prepared source/package on a physical iPhone. Supply that comment's
-numeric ID, not an issue ID, to the workflow. Simulator, unit, parser and harness
-passes cannot create or imply this acceptance. Do not paste a passing template
-without performing the checks. The required JSON fields are:
-
-- `schemaVersion`: `1`; `verdict`: `"accepted"`; `platform`: `"physical-ios"`
-- `sourceSha`: full prepared SHA; `version`: exact prepared version
-- `packageIntegrity`: the tested package report's SHA-512 integrity
-- `device`: model (no serial/UDID); `os`: tested iOS version; `testedAt`: ISO timestamp
-- `checks`: an object with each of the following keys explicitly set to `"passed"`
-
-Keep the physical smoke short, with seven grouped confirmations:
-
-- `defaultAppearance`: inspect the default row and action appearance.
-- `physicalSwipesAndActions`: swipe both edges and tap the revealed actions.
-- `autoHaptics`: feel Auto feedback, including repeated threshold crossings within
-  one gesture to confirm feedback is not repeated.
-- `hapticsOn`: feel explicitly enabled feedback.
-- `hapticsOff`: confirm explicitly disabled feedback stays silent.
-- `hapticSuppression`: confirm programmatic changes, cancellation and auto-close
-  do not produce unintended feedback.
-- `reduceMotion`: change Reduce Motion while the app is running and check its
-  live effect on motion and Auto feedback.
-
-The comment author must be a human with current admin/maintain permission, and
-the acceptance comment must be **unedited**. Corrections require a new comment
-and its new numeric ID, even if the original author makes the correction.
-The gate resolves the REST comment's `node_id` through
-[GitHub's global node lookup](https://docs.github.com/en/graphql/guides/using-global-node-ids)
-and matches the returned comment ID, author and exact body. GitHub's
-[IssueComment fields](https://docs.github.com/en/graphql/reference/issues#issuecomment)
-must report `lastEditedAt: null`, `editor: null` and `includesCreatedEdit: false`.
-Missing metadata, GraphQL errors and mismatched snapshots fail closed. REST's
-`user` and `updated_at` alone do not establish who wrote the final body. The test
-time must precede or equal comment creation. Keep unresolved observations out of
-an accepted record; document them as blockers instead.
+inventory on both platforms. They are automated evidence requirements.
+
+## Physical iPhone check
+
+Simulators cannot show haptics. Run this check on a physical iPhone with the
+prepared SHA in these cases. The publish workflow does not check it.
+
+- No earlier check SHA is recorded.
+- Since the last check SHA, one of these changed: `ios/` code, TypeScript code for
+  haptics, appearance or Reduce Motion, or the React Native or Expo version in
+  `example/package.json`.
+
+Steps:
+
+
+1. Build the example app in Release configuration on the iPhone.
+2. Inspect the default row and action appearance.
+3. Swipe both edges and tap the revealed actions.
+4. Feel Auto feedback. Cross the threshold several times in one gesture, and make
+   sure the feedback does not repeat.
+5. Feel feedback with haptics on. Make sure nothing plays with haptics off.
+6. Make sure programmatic changes, cancellation and auto-close play no feedback.
+7. Change Reduce Motion while the app runs. Check its live effect on motion, and
+   make sure Auto feedback does not change.
+
+If a check fails, fix it and prepare the release again. If all checks pass, write
+the tested SHA, the device and the iOS version in a comment on the preparation PR.
+The next check compares against that SHA.
 
 ## Verify, then publish
 
-Configure the GitHub `npm-release` environment with required human reviewers and
-main-only deployment rules, and npm trusted publishing for this workflow. Retain
+Configure the GitHub `npm-release` environment with a main-only deployment rule,
+and npm trusted publishing for this workflow. A required reviewer is optional. It
+adds an approval step before each run. Retain
 the reviewed `main` workflow as the authority. Dispatch **Publish to npm** from
 the prepared `main` revision with `operation=verify`, full `source_sha`, exact
-`version`, `ci_runs=ID`, and `acceptance_comment=ID`. Verification does not tag
+`version`, and `ci_runs=ID`. Verification does not tag
 or publish. It saves a verified report and the tested tarball. The source package,
 lockfile, changelog entry, tarball and eventual `v<version>` tag must agree.
 
@@ -242,7 +229,8 @@ version and its `v<version>` tag are permanent records.
    Release that names the problem and the replacement.
 5. Fix the problem on `main` through the normal review. Then prepare the next exact
    version, for example `1.0.0-rc.<N+1>` after a bad RC. Run the complete procedure
-   on that source: fresh full CI, physical acceptance, verify, then publish. No
+   on that source: fresh full CI, the physical iPhone check if it applies, verify,
+   then publish. No
    evidence of the bad version carries over. The publish of the corrective version
    moves its dist-tag to it.
 6. The readiness gate and the publish workflow accept only `1.0.0-rc.N` and

docs/support.md

@@ -73,8 +73,11 @@ the UIKit view on the simulator.
 
 - iOS versions below the simulator runtime above, and Android API levels other than 34.
 - Physical Android devices for 1.0. Earlier 0.x benchmarks used a Pixel 8 Pro.
-- Haptics and Reduce Motion on a physical iPhone. The maintainer checks them by
-  hand on each release candidate. This page will name the device for 1.0.0.
+- Automated tests of how haptics feel and how Reduce Motion behaves on a device.
+  The maintainer checks them by hand on a physical iPhone. The check runs when no
+  earlier check is recorded, and after a change to the iOS code, the haptics,
+  appearance or Reduce Motion code, or the example app's React Native or Expo
+  version.
 
 If you run the package on another configuration, an issue with the result helps,
 whether it works or not.

e2e/README.md

@@ -2,7 +2,7 @@
 
 The mobile suites exercise the demo through WebdriverIO and Appium. The command
 ownership regressions exercise the installed test framework and HTTP transport
-offline; they do not replace Android, iOS, or physical-device acceptance.
+offline; they do not replace Android, iOS, or the physical iPhone check.
 
 ## Offline harness checks
 

scripts/release-readiness.mjs

@@ -18,10 +18,6 @@ export const REQUIRED_JOBS = Object.freeze([
   'iOS / performance', 'iOS / rtl', 'iOS / swipe-back', 'iOS / swipeable', 'iOS / two-sided',
   'e2e-ios'
 ])
-export const PHYSICAL_CHECKS = Object.freeze([
-  'defaultAppearance', 'physicalSwipesAndActions', 'autoHaptics', 'hapticsOn',
-  'hapticsOff', 'hapticSuppression', 'reduceMotion'
-])
 const packageName = 'react-native-swipeable-actions'
 const shaPattern = /^[a-f0-9]{40}$/
 const idPattern = /^[1-9][0-9]*$/
@@ -118,61 +114,6 @@ export function validateAndroidLog(log, manifest) {
   assert.equal(seen.size, 5)
 }
 
-// REST's user identifies the original author, not whoever last edited the body.
-// Resolve its global node ID and require GitHub's edit-specific metadata.
-export async function fetchCommentEditMetadata(nodeId, token, request = fetch) {
-  assert.ok(typeof nodeId === 'string' && nodeId.length > 0, 'Missing comment node ID')
-  const response = await request('https://api.github.com/graphql', {
-    method: 'POST',
-    headers: { Authorization: 'Bearer ' + token, 'Content-Type': 'application/json' },
-    body: JSON.stringify({
-      query: `query AcceptanceComment($id: ID!) {
-        node(id: $id) {
-          __typename
-          ... on IssueComment {
-            id body lastEditedAt includesCreatedEdit
-            author { login }
-            editor { login }
-          }
-        }
-      }`,
-      variables: { id: nodeId }
-    }),
-    signal: AbortSignal.timeout(120000)
-  })
-  assert.ok(response.ok, 'GitHub comment edit metadata request failed: ' + response.status)
-  const result = await response.json()
-  assert.ok(!result.errors?.length, 'GitHub comment edit metadata returned GraphQL errors')
-  assert.ok(result.data?.node, 'GitHub comment edit metadata unavailable')
-  return result.data.node
-}
-
-export function validatePhysicalAcceptance(comment, permission, { sourceSha, version, packageIntegrity }) {
-  assert.equal(comment.user?.type, 'User', 'Physical acceptance must be a human statement')
-  assert.ok(['admin', 'maintain'].includes(permission), 'Physical acceptance must be from a maintainer')
-  const metadata = comment.editMetadata
-  assert.equal(metadata?.__typename, 'IssueComment', 'Authoritative comment edit metadata required')
-  assert.ok(typeof comment.node_id === 'string' && comment.node_id.length > 0, 'Missing comment node ID')
-  assert.equal(metadata.id, comment.node_id, 'Edit metadata belongs to another comment')
-  assert.equal(metadata.body, comment.body, 'Comment changed while fetching acceptance; retry with a new statement')
-  assert.equal(metadata.author?.login, comment.user.login, 'Comment author differs between API responses')
-  assert.equal(metadata.lastEditedAt, null, 'Physical acceptance must be unedited; post a new maintainer statement')
-  assert.equal(metadata.editor, null, 'Physical acceptance has an editor; post a new maintainer statement')
-  assert.equal(metadata.includesCreatedEdit, false, 'Physical acceptance has edit history; post a new maintainer statement')
-  const report = JSON.parse(comment.body)
-  assert.equal(report.schemaVersion, 1)
-  assert.equal(report.verdict, 'accepted', 'Physical acceptance is incomplete')
-  assert.equal(report.sourceSha, sourceSha, 'Physical acceptance is for another source')
-  assert.equal(report.version, version, 'Physical acceptance is for another version')
-  assert.equal(report.packageIntegrity, packageIntegrity, 'Physical acceptance is for another package')
-  assert.equal(report.platform, 'physical-ios', 'Simulator evidence cannot establish haptic acceptance')
-  for (const field of ['device', 'os']) assert.ok(typeof report[field] === 'string' && report[field].trim())
-  const testedAt = Date.parse(report.testedAt)
-  assert.ok(Number.isFinite(testedAt) && testedAt <= Date.parse(comment.created_at), 'Invalid physical test time')
-  for (const check of PHYSICAL_CHECKS) assert.equal(report.checks?.[check], 'passed', 'Physical check not accepted: ' + check)
-  return { commentId: comment.id, reviewer: comment.user.login, testedAt: report.testedAt }
-}
-
 export function validateReadiness(input) {
   const release = validateSource(input)
   assert.equal(input.runs.length, 1, 'Exactly one fresh, fully green full CI run required')
@@ -194,15 +135,14 @@ export function validateReadiness(input) {
   }
   validateAndroidLog(androidLog, manifest)
   const packageIntegrity = validatePackage(report, bytes, input.version, input.sourceSha, packedPackage)
-  const acceptance = validatePhysicalAcceptance(input.comment, input.permission, { ...input, packageIntegrity })
   return { schemaVersion: 1, sourceSha: input.sourceSha, version: input.version, ...release,
-    packageIntegrity, ciRuns: [run.id], acceptance }
+    packageIntegrity, ciRuns: [run.id] }
 }
 
 async function main(args) {
-  assert.equal(args.length, 10, 'Expected --source SHA --version VERSION --runs ID --acceptance COMMENT_ID --out DIRECTORY')
-  const options = Object.fromEntries(Array.from({ length: 5 }, (_, i) => [args[i * 2], args[i * 2 + 1]]))
-  assert.deepEqual(Object.keys(options).sort(), ['--acceptance', '--out', '--runs', '--source', '--version'])
+  assert.equal(args.length, 8, 'Expected --source SHA --version VERSION --runs ID --out DIRECTORY')
+  const options = Object.fromEntries(Array.from({ length: 4 }, (_, i) => [args[i * 2], args[i * 2 + 1]]))
+  assert.deepEqual(Object.keys(options).sort(), ['--out', '--runs', '--source', '--version'])
   const sourceSha = options['--source']
   const version = options['--version']
   const repository = process.env.GITHUB_REPOSITORY
@@ -219,7 +159,7 @@ async function main(args) {
   validateSource(input)
   const runIds = options['--runs'].split(',')
   assert.equal(runIds.length, 1, 'Exactly one full CI run ID required')
-  for (const id of [...runIds, options['--acceptance']]) assert.match(id, idPattern)
+  for (const id of runIds) assert.match(id, idPattern)
   const base = 'https://api.github.com/repos/' + repository
   const request = async (route) => {
     const response = await fetch(base + route, {
@@ -293,10 +233,6 @@ async function main(args) {
       const packedPackage = JSON.parse(execFileSync('tar', ['-xOf', tarball, 'package/package.json'], { encoding: 'utf8', maxBuffer: 1024 * 1024 }))
       input.runs.push({ run, jobs, manifest, expectedManifest, bundles, androidLog, report, bytes, packedPackage })
     }
-    input.comment = await json('/issues/comments/' + options['--acceptance'])
-    input.comment.editMetadata = await fetchCommentEditMetadata(input.comment.node_id, process.env.GH_TOKEN)
-    const collaborator = await json('/collaborators/' + encodeURIComponent(input.comment.user.login) + '/permission')
-    input.permission = collaborator.permission === 'admin' ? 'admin' : collaborator.role_name
     const result = validateReadiness(input)
     await mkdir(options['--out'], { recursive: true })
     await writeFile(path.join(options['--out'], 'package.tgz'), input.runs[0].bytes, { flag: 'wx' })