Merge pull request 'fix(ci): pr-iphone-deploy no longer self-merges on its own comment' (#13) from fix/ci-lgtm-self-match into main

Reviewed-on: #13

.github/workflows/pr-iphone-deploy.yml

@@ -8,12 +8,11 @@ on:
   pull_request:
     branches: [main]
     types: [opened, synchronize, reopened]
-    # Only run the iOS build+deploy when the PR actually touches the Swift app
-    # or this workflow itself. Docs-only / backend-only / admin-web-only PRs
-    # skip the runner entirely (no point rebuilding an unchanged .app).
-    paths:
-      - 'tabatago-swift/**'
-      - '.github/workflows/pr-iphone-deploy.yml'
+    # NOTE: no trigger-level `paths:` here — Gitea Actions does not reliably
+    # honor `on.pull_request.paths:` filters (the workflow silently fails to
+    # trigger even when changed files match). Path gating is done at the job
+    # level via dorny/paths-filter@v3 in the `changes` job below (same pattern
+    # as ci.yml). Docs/backend-only PRs then skip the macOS runner.
 
 # Every mutation (comments, merge) goes through the Gitea API with PR_API_TOKEN,
 # never the runner's native GITHUB_TOKEN — so a `permissions:` block would be
@@ -26,8 +25,29 @@ concurrency:
 
 jobs:
 
+  # ── Path filter — determines whether the macOS build+deploy is worth running ──
+  # Gitea Actions does not honor trigger-level `on.pull_request.paths:`, so we
+  # gate at the job level with dorny/paths-filter (same pattern as ci.yml).
+  changes:
+    name: Detect Changes
+    runs-on: ubuntu-latest
+    outputs:
+      ios: ${{ steps.filter.outputs.ios }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: dorny/paths-filter@v3
+        id: filter
+        with:
+          filters: |
+            ios:
+              - 'tabatago-swift/**'
+              - '.github/workflows/pr-iphone-deploy.yml'
+
   build-deploy:
     name: Build & Deploy to iPhone (devicectl)
+    needs: changes
+    if: needs.changes.outputs.ios == 'true'
     runs-on: macos
     timeout-minutes: 30
 
@@ -172,15 +192,25 @@ jobs:
         run: |
           PR="${{ github.event.pull_request.number }}"
           REPO="${{ github.repository }}"
+          # The HTML comment is invisible in Gitea's rendered markdown but is
+          # present in the raw body — wait-approval uses it as a sentinel to
+          # skip THIS bot comment when scanning for LGTM/KO (otherwise the
+          # instruction text "Reply LGTM pour merger" would self-trigger a
+          # merge ~30s after deploy). DO NOT remove or reword without also
+          # updating the scanner in the wait-approval job.
           curl -s -X POST \
             -H "Authorization: token ${GT_TOKEN}" \
             -H "Content-Type: application/json" \
-            -d "{\"body\":\"## 📱 Prêt à tester !\\n\\nL'app est déployée sur l'iPhone (devicectl).\\n\\n- Teste les changements\\n- Reply **LGTM** pour merger\\n- Reply **KO** pour bloquer\"}" \
+            -d "{\"body\":\"## 📱 Prêt à tester !\\n\\nL'app est déployée sur l'iPhone (devicectl).\\n\\n<!-- tabatago:ready-to-test -->\\n\\n- Teste les changements\\n- Reply **LGTM** pour merger\\n- Reply **KO** pour bloquer\"}" \
             "${GITEA_URL}/api/v1/repos/${REPO}/issues/${PR}/comments"
 
   wait-approval:
     name: Wait for LGTM comment
     needs: build-deploy
+    # Only run when build-deploy actually deployed. Without this, a skipped
+    # build-deploy (filtered out by `changes`) would still launch this job and
+    # the merge poll would run against a PR that was never deployed to device.
+    if: needs.build-deploy.result == 'success'
     runs-on: macos
     timeout-minutes: 120
 
@@ -202,6 +232,19 @@ jobs:
           # Re-fetch ALL comments each cycle (not since_id) so we also catch
           # edits — e.g. a reviewer changing "KO" → "LGTM". Negligible cost for
           # PRs with <50 comments.
+          #
+          # Parsing is done with python3 (preinstalled on the macOS runner and
+          # already used by scripts/ci-status.py) rather than grep over raw JSON:
+          #   - the bot's "Ready to test" comment body literally contains the
+          #     words LGTM/KO as instructions to the reviewer, so any naive
+          #     body grep would self-match and auto-merge ~30s after deploy.
+          #     We skip it via the <!-- tabatago:ready-to-test --> sentinel
+          #     embedded by the build-deploy job;
+          #   - json.loads avoids false matches when the literal "body" key or
+          #     trigger words appear inside another string field.
+          # KO is checked BEFORE LGTM so a PR with both signals blocks (matches
+          # the existing "KO blocks" intent — a reviewer who flip-flops shouldn't
+          # merge just because an older LGTM is still in the history).
 
           while [ $TRIES -lt $MAX ]; do
             sleep 30
@@ -210,23 +253,51 @@ jobs:
             COMMENTS=$(curl -s -H "Authorization: token ${GT_TOKEN}" \
               "${API}/issues/${PR}/comments?limit=50&page=1")
 
-            # Match LGTM/KO anywhere in the body (with a word boundary so "KOM"
-            # or "LGTMX" don't trigger). Case-insensitive.
-            if echo "$COMMENTS" | grep -qiE '"body":[[:space:]]*"[^"]*\bLGTM\b'; then
-              echo "✅ LGTM reçu ! Squash-merge..."
-              curl -s -X POST \
-                -H "Authorization: token ${GT_TOKEN}" \
-                -H "Content-Type: application/json" \
-                -d "{\"Do\":\"squash\"}" \
-                "${API}/pulls/${PR}/merge"
-              echo "✅ Mergée (squash)."
-              exit 0
-            fi
+            DECISION=$(python3 - "$COMMENTS" <<'PY'
+import json, re, sys
+MARKER = "<!-- tabatago:ready-to-test -->"
+try:
+    comments = json.loads(sys.argv[1] or "[]")
+except Exception:
+    comments = []
+lgtm = re.compile(r"\bLGTM\b", re.IGNORECASE)
+ko    = re.compile(r"\bKO\b",   re.IGNORECASE)
+hit_lgtm = hit_ko = False
+for c in comments:
+    body = c.get("body") or ""
+    if MARKER in body:
+        # Skip the bot's own "Ready to test" comment — its body mentions
+        # LGTM/KO as instructions and would otherwise self-trigger.
+        continue
+    if ko.search(body):
+        hit_ko = True
+    if lgtm.search(body):
+        hit_lgtm = True
+if hit_ko:
+    print("KO")
+elif hit_lgtm:
+    print("LGTM")
+else:
+    print("PENDING")
+PY
+)
 
-            if echo "$COMMENTS" | grep -qiE '"body":[[:space:]]*"[^"]*\bKO\b'; then
-              echo "❌ KO reçu. Bloquée."
-              exit 1
-            fi
+            case "$DECISION" in
+              LGTM)
+                echo "✅ LGTM reçu ! Squash-merge..."
+                curl -s -X POST \
+                  -H "Authorization: token ${GT_TOKEN}" \
+                  -H "Content-Type: application/json" \
+                  -d "{\"Do\":\"squash\"}" \
+                  "${API}/pulls/${PR}/merge"
+                echo "✅ Mergée (squash)."
+                exit 0
+                ;;
+              KO)
+                echo "❌ KO reçu. Bloquée."
+                exit 1
+                ;;
+            esac
 
             if [ $((TRIES % 4)) -eq 0 ]; then
               echo "   ⏳ ... (${TRIES}/240, $(date +%H:%M))"

AGENTS.md

@@ -167,7 +167,7 @@ Complications : `TabataGoComplication`.
 6. **xcodegen generate** → `xcodebuild -resolvePackageDependencies` → **build** (`-scheme TabataGo`, Debug, auto-provisioning, team `2MJF39L8VY`). Le scheme build les 4 targets : `TabataGo` (app iOS), `TabataGoWidget` (widget iOS, vrai target — ne pas confondre avec `TabataGoWatchWidget` qui est watchOS), `TabataGoWatch` (app watchOS), `TabataGoWatchWidget` (complication watchOS).
 7. **Deploy iPhone** UDID `00008120-000925CE3672201E` : `xcrun devicectl device install app` uniquement. `devicectl` gère nativement la découverte WiFi (réseau) et filaire (USB-C/Thunderbolt) — pas de fallback `ios-deploy`.
 8. **Post comment** "Prêt à tester" sur la PR.
-9. **Job `wait-approval`** : poll (30s, max 240 = 2h). **Re-fetche TOUS les comments** chaque cycle (pas `since_id`) pour attraper aussi les edits (un reviewer passant de "KO" à "LGTM"). Match LGTM/KO n'importe où dans le body (regex `\bLGTM\b` / `\bKO\b`, case-insensitive) — "Tested, LGTM!" compte ; "KOM" ne compte pas. `LGTM` → **squash-merge** (`{"Do":"squash"}`). `KO` → blocage. Timeout → fail.
+9. **Job `wait-approval`** : poll (30s, max 240 = 2h). **Re-fetche TOUS les comments** chaque cycle (pas `since_id`) pour attraper aussi les edits (un reviewer passant de "KO" à "LGTM"). Le parsing se fait en **python3** (`json.loads` + regex, pas de `grep` sur JSON brut). **Le commentaire bot "Prêt à tester" porte un marker sentinel `<!-- tabatago:ready-to-test -->`** et est **ignoré** par le scanner — sinon son propre body (qui mentionne LGTM/KO comme instructions au reviewer) déclencherait un auto-merge ~30s après le deploy (bug historique). KO est checké **avant** LGTM (un PR avec les deux signaux bloque). Match LGTM/KO n'importe où dans le body (regex `\bLGTM\b` / `\bKO\b`, case-insensitive) — "Tested, LGTM!" compte ; "KOM" ne compte pas. `LGTM` → **squash-merge** (`{"Do":"squash"}`). `KO` → blocage. Timeout → fail.
 
 ### Secrets
 
@@ -187,6 +187,7 @@ Complications : `TabataGoComplication`.
 - `-skipPackagePluginValidation -allowProvisioningUpdates`.
 - **Concurrency** : le bloc `concurrency:` est volontaire — ne pas le retirer, sinon les pushes successifs empilent des pipelines et tentent des double-merges.
 - **LGTM/KO regex** : matche n'importe où dans le body avec `\b...\b` (word boundary). Ne pas revenir à un `grep '"body": *"LGTM"'` ancré — il raterait "Tested, LGTM!".
+- **Self-match sentinel** : le commentaire bot "Prêt à tester" contient littéralement `Reply **LGTM** pour merger` et `Reply **KO**` — sans précaution, le scanner matche son **propre** commentaire et auto-merge ~30s après le deploy. Le body porte donc un marker `<!-- tabatago:ready-to-test -->` que `wait-approval` ignore. **Tout commentaire bot posté par ce workflow doit porter ce marker** ; ne pas le retirer ni poster d'autre commentaire contenant LGTM/KO sans marker.
 - **Squash-merge** (`{"Do":"squash"}`) — pas merge commit ni rebase. Garde l'historique `main` linéaire.
 - **Comment edits** : `wait-approval` re-fetche tous les comments chaque cycle (pas `since_id`) pour attraper les edits.
 
@@ -249,6 +250,8 @@ Node.js (`server.js`, `package.json`, `Dockerfile`). Télécharge l'audio de pla
 | Compter sur un bloc `permissions:` natif Gitea Actions | Tout passe par l'API Gitea avec `PR_API_TOKEN` ; le bloc natif serait no-op |
 | Retirer le bloc `concurrency:` | Permet double-merge et pipelines empilés |
 | Ancre le grep LGTM au début du body (`'"body": *"LGTM"'`) | Regex `\bLGTM\b` n'importe où dans le body (attrape "Tested, LGTM!") |
+| Poster un commentaire bot dont le body contient "LGTM"/"KO" sans le marker `<!-- tabatago:ready-to-test -->` | Le scanner bot-scan ignore les comments portant le marker — tout commentaire bot du workflow doit l'inclure (sinon auto-merge en ~30s) |
+| Scanner les comments en `grep` sur le JSON brut | `python3` + `json.loads` (gère unicode/quotes, pas de false match sur la clé `"body"`) |
 | `{"Do":"merge"}` (merge commit) | `{"Do":"squash"}` pour un `main` linéaire |
 | Supprimer le scheme explicite dans `project.yml` | xcodegen 2.45.4 n'en crée pas — scheme doit lister `TabataGo`/`TabataGoWidget`/`TabataGoWatch`/`TabataGoWatchWidget` |
 | Compter sur `admin-web/` pour la app iOS | Dashboard admin séparé, communique via Supabase uniquement |