Bound review image retention and separate CI navigation budgets

Reviewed-on: #30

.gitea/workflows/creative-workspace.yml

@@ -11,6 +11,8 @@ on:
       - 'scripts/integration.sh'
       - 'scripts/cleanup-integration.sh'
       - 'scripts/test-integration-cleanup.py'
+      - 'scripts/cleanup-review-images.sh'
+      - 'scripts/test-review-image-cleanup.py'
 concurrency:
   group: creative-workspace-${{ gitea.event.pull_request.number }}
   cancel-in-progress: true
@@ -25,8 +27,9 @@ jobs:
       - name: Check patch formatting
         run: |
           git diff --check HEAD^ HEAD
-          bash -n scripts/integration.sh scripts/cleanup-integration.sh
+          bash -n scripts/integration.sh scripts/cleanup-integration.sh scripts/cleanup-review-images.sh
           python3 scripts/test-integration-cleanup.py
+          python3 scripts/test-review-image-cleanup.py
       - name: Verify Studio types, tests, build and smoke
         run: docker build -f apps/studio/Dockerfile -t "creative-studio-review:${{ gitea.sha }}" .
       - name: Build Wiki integration and runtime images
@@ -45,3 +48,8 @@ jobs:
         env:
           GITEA_RUN_ID: ${{ gitea.run_id }}
         run: bash scripts/cleanup-integration.sh
+      - name: Remove this run disposable review image tags
+        if: always()
+        env:
+          REVIEW_SHA: ${{ gitea.sha }}
+        run: bash scripts/cleanup-review-images.sh

apps/waterfall-wiki/tests/guidance-workspace-ui.ts

@@ -90,6 +90,9 @@ export async function verifyGuidanceWorkspaceUI(base: string, browser: Browser)
     );
     const page = await context.newPage();
     page.setDefaultTimeout(10000);
+    // Navigation includes document and script loading on the shared runner;
+    // keep action/readiness checks at 10s and the suite's overall bound unchanged.
+    page.setDefaultNavigationTimeout(30000);
     page.on("pageerror", (error) => errors.push(error.message));
     page.on("crash", () => errors.push("Browser renderer crashed"));
     const pending = new Set<string>(), network: string[] = [];
@@ -97,12 +100,14 @@ export async function verifyGuidanceWorkspaceUI(base: string, browser: Browser)
     page.on("requestfinished", request => pending.delete(request.url()));
     page.on("requestfailed", request => {pending.delete(request.url());network.push(new URL(request.url()).pathname+": "+request.failure()?.errorText);});
     async function navigate(url: string, reload = false) {
-      const previousSessionReplies = sessionReplies;
+      const previousSessionReplies = sessionReplies, started = Date.now();
       try {
         const navigation = await (reload ? page.reload({waitUntil:"domcontentloaded"}) : page.goto(url,{waitUntil:"domcontentloaded"}));
         assert.ok(navigation?.ok(), "Navigation returns a successful document");
+        const documentMs = Date.now() - started;
         await page.getByText(signedIn ? "editor@example.com" : "登录你的账户",{exact:true}).first().waitFor({state:"attached"});
         assert.ok(sessionReplies > previousSessionReplies, "The new document requests its account fixture and renders the resulting account state");
+        console.log("Workspace UI navigation timing", JSON.stringify({path:new URL(url).pathname,documentMs,accountReadyMs:Date.now()-started}));
       } catch (error) {
         const probeStarted=Date.now();
         const server=await fetch(url,{signal:AbortSignal.timeout(3000)}).then(async response=>{await response.body?.cancel();return {status:response.status,elapsedMs:Date.now()-probeStarted}}).catch(error=>({error:error instanceof Error?error.name:"unknown",elapsedMs:Date.now()-probeStarted}));

scripts/cleanup-review-images.sh

@@ -0,0 +1,18 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+: "${REVIEW_SHA:?The exact review commit SHA is required}"
+[[ "$REVIEW_SHA" =~ ^[0-9a-f]{40}$ ]] || { echo 'Refusing invalid review SHA' >&2; exit 2; }
+# Only this workflow's disposable tags; production tags, volumes and other projects stay out of scope.
+for repository in creative-studio-review creative-wiki-tests creative-wiki-review; do
+ tag="$repository:$REVIEW_SHA"
+ referenced=$(docker ps -aq | xargs -r docker inspect --format '{{.Image}}')
+ if ! image=$(docker image inspect --format '{{.Id}}' "$tag" 2>/dev/null); then
+  echo "No review image: $tag"
+  continue
+ fi
+ if printf '%s\n' "$referenced" | grep -Fxq "$image"; then
+  echo "Keeping container-referenced review image: $tag"
+  continue
+ fi
+ docker image rm "$tag"
+done

scripts/integration.sh

@@ -17,7 +17,15 @@ compose(){ docker compose -p "$COMPOSE_PROJECT_NAME" --env-file "$TEST_ENV" "${C
 cleanup(){
  code=$?
  trap - EXIT INT TERM
- if [ "$code" -ne 0 ]; then compose logs --tail 160 waterfall-wiki waterfall-wiki-worker "${STUDIO_SERVICES[@]}" || true; fi
+ if [ "$code" -ne 0 ]; then
+  echo 'Runner pressure at integration failure (before disposable stack cleanup)'
+  uptime || true
+  free -m || true
+  df -h / || true
+  for resource in cpu memory io; do cat "/proc/pressure/$resource" || true; done
+  timeout 10s docker stats --no-stream --format '{{.Name}} cpu={{.CPUPerc}} memory={{.MemUsage}} io={{.BlockIO}} pids={{.PIDs}}' || true
+  compose logs --tail 160 waterfall-wiki waterfall-wiki-worker "${STUDIO_SERVICES[@]}" || true
+ fi
  cleanup_code=0
  bash scripts/cleanup-integration.sh || cleanup_code=$?
  rm -f "$TEST_ENV"

scripts/test-review-image-cleanup.py

@@ -0,0 +1,73 @@
+#!/usr/bin/env python3
+"""Check review cleanup scope without Docker or build output."""
+import json
+import os
+from pathlib import Path
+import subprocess
+import tempfile
+import unittest
+
+SCRIPT = Path(__file__).with_name('cleanup-review-images.sh')
+SHA = 'a' * 40
+FAKE_DOCKER = '''#!/usr/bin/env python3
+import json, os, sys
+args=sys.argv[1:]
+with open(os.environ['DOCKER_CALLS'], 'a') as output:
+    output.write(json.dumps(args)+'\\n')
+if args[0]=='ps':
+    if os.environ.get('FAIL_INVENTORY'): sys.exit(1)
+    if os.environ.get('REFERENCED'): print('container-id')
+elif args[0]=='inspect':
+    print('sha256:review-image')
+elif args[:2]==['image','inspect']:
+    if os.environ.get('MISSING_IMAGE'): sys.exit(1)
+    print('sha256:review-image')
+'''
+
+
+class ReviewCleanupTests(unittest.TestCase):
+    def run_cleanup(self, sha=SHA, **extra):
+        with tempfile.TemporaryDirectory() as directory:
+            root = Path(directory)
+            docker = root / 'docker'
+            docker.write_text(FAKE_DOCKER)
+            docker.chmod(0o700)
+            calls = root / 'calls'
+            result = subprocess.run(['bash', str(SCRIPT)], capture_output=True, env={
+                **os.environ, 'PATH': directory + os.pathsep + os.environ['PATH'],
+                'DOCKER_CALLS': str(calls), 'REVIEW_SHA': sha, **extra})
+            return result.returncode, [json.loads(line) for line in calls.read_text().splitlines()] if calls.exists() else []
+
+    def test_removes_only_exact_review_tags_without_force(self):
+        code, calls = self.run_cleanup()
+        self.assertEqual(code, 0)
+        self.assertEqual([args for args in calls if args[:2] == ['image', 'rm']], [
+            ['image', 'rm', name + ':' + SHA] for name in
+            ['creative-studio-review', 'creative-wiki-tests', 'creative-wiki-review']])
+        self.assertFalse(any('-f' in args or 'prune' in args or 'volume' in args for args in calls))
+
+    def test_invalid_sha_never_contacts_docker(self):
+        for value in ['', 'main', 'a' * 39, 'a' * 41, 'A' * 40, '*' * 40]:
+            with self.subTest(value=value):
+                code, calls = self.run_cleanup(value)
+                self.assertNotEqual(code, 0)
+                self.assertEqual(calls, [])
+
+    def test_container_referenced_images_are_kept(self):
+        code, calls = self.run_cleanup(REFERENCED='1')
+        self.assertEqual(code, 0)
+        self.assertFalse(any(args[:2] == ['image', 'rm'] for args in calls))
+
+    def test_missing_tags_are_a_successful_noop(self):
+        code, calls = self.run_cleanup(MISSING_IMAGE='1')
+        self.assertEqual(code, 0)
+        self.assertFalse(any(args[:2] == ['image', 'rm'] for args in calls))
+
+    def test_inventory_failure_prevents_removal(self):
+        code, calls = self.run_cleanup(FAIL_INVENTORY='1')
+        self.assertNotEqual(code, 0)
+        self.assertEqual(len(calls), 1)
+
+
+if __name__ == '__main__':
+    unittest.main()