Implemented ui and unit test cases along with iOS Test workflow

.github/scripts/publish-reports.sh

@@ -0,0 +1,69 @@
+#!/usr/bin/env bash
+# Attaches CI report files to the "ios-ci-reports" pre-release and links them from the job's
+# summary page, then deletes that release's assets older than RETENTION_DAYS.
+#
+# Used instead of actions/upload-artifact: from this self-hosted runner its very first request
+# (CreateArtifact, made before any bytes are sent) times out on every attempt, so no artifact is
+# ever created. Release assets go through uploads.github.com instead, don't count against the
+# artifact storage quota, and don't grow the git repo (the release only holds files). Same
+# approach as the Android repo's .github/scripts/publish-reports.sh.
+#
+# Tag is "ios-ci-reports", not "ci-reports": an earlier workflow version pushed a *branch*
+# called ci-reports to this repo, and a tag with the same name would be ambiguous.
+#
+# Needs the gh CLI on the runner and GH_TOKEN with contents: write (to create the release/tag
+# and upload/delete its assets).
+#
+# Usage: publish-reports.sh <file> [<file> ...]   (missing files are skipped with a warning)
+set -euo pipefail
+
+tag="ios-ci-reports"
+retention_days=14
+
+command -v gh >/dev/null || { echo "::error::gh CLI not found on this runner (brew install gh)"; exit 1; }
+
+files=()
+for f in "$@"; do
+  if [ -f "$f" ]; then files+=("$f"); else echo "::warning::$f not found - skipping"; fi
+done
+if [ "${#files[@]}" -eq 0 ]; then
+  echo "::warning::no report files to publish"
+  exit 0
+fi
+
+# Two runs (e.g. a push and the Friday schedule) can race to create the release the first time -
+# if the create fails, the other run has won the race, as long as the release exists now
+if ! gh release view "$tag" >/dev/null 2>&1; then
+  gh release create "$tag" --prerelease --target "$GITHUB_SHA" --title "iOS CI reports" \
+    --notes "Test report files from iOS CI runs, linked from each run's summary page. Files older than $retention_days days are deleted automatically." \
+    || gh release view "$tag" >/dev/null
+fi
+
+base_url="$GITHUB_SERVER_URL/$GITHUB_REPOSITORY/releases/download/$tag"
+{
+  echo "## Test reports"
+  echo "Attached to the [$tag]($GITHUB_SERVER_URL/$GITHUB_REPOSITORY/releases/tag/$tag) pre-release, kept for $retention_days days:"
+} >> "$GITHUB_STEP_SUMMARY"
+
+for file in "${files[@]}"; do
+  # run id + attempt in the name, since all runs share this one release (a re-run gets its own
+  # file instead of replacing the reports of the attempt before it)
+  name="${GITHUB_RUN_ID}-${GITHUB_RUN_ATTEMPT}-$(basename "$file")"
+  staged="$RUNNER_TEMP/$name"
+  cp "$file" "$staged"
+  gh release upload "$tag" "$staged" --clobber
+  rm -f "$staged"
+  echo "- [$name]($base_url/$name)" >> "$GITHUB_STEP_SUMMARY"
+  echo "published: $base_url/$name"
+done
+
+# Delete assets older than the retention window. BSD date (macOS runner) first, GNU date as a
+# fallback. Paginated by release id: the release object itself only embeds a limited asset list.
+cutoff=$(date -u -v-"${retention_days}"d +%Y-%m-%dT%H:%M:%SZ 2>/dev/null \
+  || date -u -d "$retention_days days ago" +%Y-%m-%dT%H:%M:%SZ)
+release_id=$(gh api "repos/$GITHUB_REPOSITORY/releases/tags/$tag" --jq .id)
+gh api --paginate "repos/$GITHUB_REPOSITORY/releases/$release_id/assets?per_page=100" \
+  --jq ".[] | select(.created_at < \"$cutoff\") | .id" |
+  while read -r id; do
+    gh api -X DELETE "repos/$GITHUB_REPOSITORY/releases/assets/$id" || echo "::warning::couldn't delete old asset $id"
+  done

GoInfoGame.xctestplan

@@ -9,12 +9,13 @@
     }
   ],
   "defaultOptions" : {
-    "codeCoverage" : false,
     "targetForVariableExpansion" : {
       "containerPath" : "container:GoInfoGame.xcodeproj",
       "identifier" : "FAD5C4EE2AFCBE700040C61A",
       "name" : "GoInfoGame"
-    }
+    },
+    "uiTestingScreenshotsLifetime" : "deleteOnSuccess",
+    "userAttachmentLifetime" : "deleteOnSuccess"
   },
   "testTargets" : [
     {
@@ -23,6 +24,13 @@
         "identifier" : "FAD5C5042AFCBE720040C61A",
         "name" : "GoInfoGameTests"
       }
+    },
+    {
+      "target" : {
+        "containerPath" : "container:GoInfoGame.xcodeproj",
+        "identifier" : "FAD5C50E2AFCBE720040C61A",
+        "name" : "GoInfoGameUITests"
+      }
     }
   ],
   "version" : 1

GoInfoGame/GoInfoGame/AppDelegate.swift

@@ -7,6 +7,7 @@
 
 import SwiftUI
 import FirebaseCore
+import RealmSwift
 
 @main
 class AppDelegate: UIResponder, UIApplicationDelegate {
@@ -15,7 +16,37 @@ class AppDelegate: UIResponder, UIApplicationDelegate {
 
     func application(_ application: UIApplication, didFinishLaunchingWithOptions launchOptions: [UIApplication.LaunchOptionsKey: Any]?) -> Bool {
         // Override point for customization after application launch.
-
+
+        // Absolute first thing, before anything else on this or any other thread has a
+        // chance to run: force Realm's schema to be computed once, synchronously, here.
+        // RLMSchema.sharedSchema (a process-wide singleton, cached for the rest of the
+        // process once computed) discovers Realm's model classes by walking every
+        // Objective-C class currently loaded via RLMRegisterClassLocalNames. If that walk
+        // races with something else registering a class on another thread - which is
+        // routine seconds later, once Firebase/StoreKit/etc. are initializing and
+        // SceneDelegate.sceneDidBecomeActive fires NotesSubmissionManager and
+        // FeatureSubmissionManager's resumePendingUploads(), each opening its own
+        // Task.detached that can independently trigger this same first-ever Realm access -
+        // class_getSuperclass can read a pointer the runtime is mutating out from under it
+        // and crash with EXC_BAD_ACCESS (seen as a crash inside
+        // RLMRegisterClassLocalNames/RLMIsObjectSubclass). Doing it here, alone, before any
+        // of that concurrent activity starts, means every later access - including the ones
+        // from those detached tasks - just reuses the cached schema and never repeats the
+        // race. Intermittent, because it depends on how the OS scheduler happens to
+        // interleave threads on a given launch; far more likely to surface under a UI test
+        // suite, which launches a fresh process (and so re-triggers this one-time
+        // computation) dozens of times in a row.
+        _ = try? Realm(configuration: RealmConfig.configuration)
+
+        // Before anything else: a UI test run needs its stubs registered ahead of the
+        // first request, and ApiManager.shared is created lazily by whoever asks first.
+        // Reset first: SceneDelegate picks its root view from `loggedIn`, so stale state
+        // has to be gone before the scene connects.
+        #if DEBUG
+        UITestStubs.resetStateIfNeeded()
+        UITestStubs.installIfNeeded()
+        #endif
+
         //clear DB to avoid overlap of data after workspace selection
        // DatabaseConnector.shared.clearDB()
         validateAccessToken()