fix(reliability): SR-187 — ban datetime.utcnow(), enforce UTC-aware datetimes

scripts/check_banned_patterns.py

@@ -99,6 +99,12 @@
      "skylight-cli click --element-index is unstable"),
     (re.compile(r'subprocess\.Popen.*Chrome.*(?!remote-allow-origins=\\"\*\\")'),
      'Chrome MUST be launched with --remote-allow-origins="*" (with quotes!)'),
+    # SR-187: datetime.utcnow() is deprecated in Python 3.12 (DeprecationWarning)
+    # and slated for removal in 3.14. It returns a NAIVE datetime — comparing
+    # naive against UTC-aware DB timestamps silently mis-orders. Use
+    # `datetime.now(timezone.utc)` instead. See issue #186 for migration log.
+    (re.compile(r'\bdatetime\.utcnow\s*\('),
+     'datetime.utcnow() returns NAIVE dt; use datetime.now(timezone.utc) (SR-187)'),
 ]
 
 ROOT = Path(__file__).resolve().parent.parent

scripts/tests/test_check_banned_patterns.py

@@ -159,6 +159,89 @@ def test_real_webauto_call_IS_flagged(self) -> None:
             self.assertEqual(cbp.scan_file(f), [])
 
 
+class UtcnowBanTests(unittest.TestCase):
+    """SR-187 (issue #186): `datetime.utcnow()` is banned.
+
+    - POSITIVE: a real call must be flagged (so CI blocks regressions).
+    - NEGATIVE: the same token in a docstring/comment must NOT be flagged
+      (this file itself documents `datetime.utcnow()` in comments).
+    """
+
+    def _write(self, tmp_path: Path, name: str, content: str) -> Path:
+        p = tmp_path / name
+        p.write_text(content)
+        return p
+
+    def test_real_utcnow_call_IS_flagged(self) -> None:
+        import tempfile
+        with tempfile.TemporaryDirectory() as td:
+            f = self._write(Path(td), "bad_dt.py", (
+                'from datetime import datetime\n'
+                'now = datetime.utcnow()\n'
+            ))
+            hits = cbp.scan_file(f)
+            self.assertEqual(len(hits), 1,
+                f"datetime.utcnow() at column-zero must be flagged; got {hits}")
+            line_no, reason, _snippet = hits[0]
+            self.assertEqual(line_no, 2)
+            self.assertIn("SR-187", reason)
+
+    def test_utcnow_chained_call_IS_flagged(self) -> None:
+        # The real production sites used `datetime.utcnow().isoformat()`.
+        # The `\b…(` anchor must still match because `\(` matches the
+        # opening paren of `utcnow()`, not the chained `.isoformat`.
+        import tempfile
+        with tempfile.TemporaryDirectory() as td:
+            f = self._write(Path(td), "bad_dt_chain.py", (
+                'from datetime import datetime\n'
+                'ts = datetime.utcnow().isoformat() + "Z"\n'
+            ))
+            hits = cbp.scan_file(f)
+            self.assertEqual(len(hits), 1,
+                f"chained datetime.utcnow().isoformat() must be flagged; got {hits}")
+
+    def test_utcnow_in_docstring_is_NOT_flagged(self) -> None:
+        # The pattern lives in EXECUTABLE code; docstrings are masked. This
+        # mirrors the SR-60 contract for every other banned pattern.
+        import tempfile
+        with tempfile.TemporaryDirectory() as td:
+            f = self._write(Path(td), "doc_dt.py", (
+                '"""Migration note: replaced datetime.utcnow() with\n'
+                'datetime.now(timezone.utc) per SR-187.\n'
+                '"""\n'
+                'value = 42\n'
+            ))
+            self.assertEqual(cbp.scan_file(f), [],
+                "datetime.utcnow() in docstring must NOT be flagged")
+
+    def test_utcnow_in_comment_is_NOT_flagged(self) -> None:
+        import tempfile
+        with tempfile.TemporaryDirectory() as td:
+            f = self._write(Path(td), "comment_dt.py", (
+                '# SR-187: replaced datetime.utcnow() with timezone.utc form\n'
+                'from datetime import datetime, timezone\n'
+                'now = datetime.now(timezone.utc)\n'
+            ))
+            self.assertEqual(cbp.scan_file(f), [],
+                "datetime.utcnow() in comment must NOT be flagged")
+
+    def test_substring_match_is_NOT_flagged(self) -> None:
+        # `\b` boundary: `not_datetime.utcnow(` should NOT match because
+        # the rule targets `datetime.utcnow(` after a word boundary.
+        # `my_datetime.utcnow(` is a (theoretical) custom class; we
+        # accept the false-negative there. What we MUST not do is fire
+        # on an unrelated `xutcnow(` token.
+        import tempfile
+        with tempfile.TemporaryDirectory() as td:
+            f = self._write(Path(td), "noisy.py", (
+                'def xutcnow():\n'
+                '    return 1\n'
+                'xutcnow()\n'
+            ))
+            self.assertEqual(cbp.scan_file(f), [],
+                "bare `xutcnow()` must NOT be flagged")
+
+
 class SelfScanTests(unittest.TestCase):
     """The script's OWN documentation includes BANNED tokens (this is the
     whole point of SR-60). Scanning the live `check_banned_patterns.py`

survey-cli/commands/answer_survey.py

@@ -49,9 +49,10 @@
 import json
 import os
 import subprocess
+import time
 import hashlib
-from datetime import datetime
-from typing import Dict
+from datetime import datetime, timezone
+from typing import Dict, Any, List, Optional
 
 # ═════════════════════════════════════════════════════════════════════════════
 # IMPORTS — vorhandene tools nutzen
@@ -60,7 +61,8 @@
 import sys
 sys.path.insert(0, sys_path)
 
-from tools.tool_snapshot import EXTRACTOR_JS
+from tools.tool_snapshot import EXTRACTOR_JS, snapshot, find_submit, find_unfilled
+from survey.completion_detector import CompletionDetector
 from survey.session_validator import validate_session
 
 CHROME_PORT = 9999
@@ -216,8 +218,7 @@ async def solve_captcha(ws) -> str:
     Model: meta/llama-3.2-90b-vision-instruct (NVIDIA NIM)
     API: https://integrate.api.nvidia.com/v1/chat/completions
     """
-    import base64
-    import urllib.request
+    import os, base64, urllib.request
 
     # Get page text to detect captcha type
     page_text = await cdp_execute_js(ws, 100, "document.body.innerText.substring(0, 500)")
@@ -372,6 +373,7 @@ async def solve_drag_drop(ws) -> str:
     Uses CDP Input.dispatchMouseEvent (Approach B) — REAL browser-level mouse events
     trigger Angular CDK's pointer event handlers. Verified working 2026-05-10.
     """
+    import time
 
     # Extract target number
     number = await cdp_execute_js(ws, 200, """
@@ -476,7 +478,7 @@ async def solve_drag_drop(ws) -> str:
 """)
             await asyncio.sleep(3)
             return f"DRAG_SOLVED:{number}->btn_enabled"
-        return "DRAG_FAILED:btn_disabled"
+        return f"DRAG_FAILED:btn_disabled"
     except:
         return f"DRAG_VERIFY_ERROR:{verify}"
 
@@ -529,7 +531,7 @@ def detect_page_type(snap: Dict) -> str:
         "unknown"    → Nicht erkannt
     """
     body = (snap.get("bodyText", "") or "").lower()
-    (snap.get("url", "") or "").lower()
+    url = (snap.get("url", "") or "").lower()
 
     # Screen-Out Keywords
     for kw in ["umfrage passt nicht", "leider", "nicht geeignet", "vorzeitig beendet",
@@ -763,14 +765,14 @@ async def answer_survey(survey_ws_url: str, max_pages: int = MAX_PAGES) -> Dict:
         }
     """
     print(f"\n{'='*60}")
-    print("  ANSWER_SURVEY — Manual Testing Mode")
+    print(f"  ANSWER_SURVEY — Manual Testing Mode")
     print(f"{'='*60}")
     print(f"  Survey Tab WS: {survey_ws_url[:60]}...")
 
     # Pre-Flight: Session validieren
     if not validate_session(CHROME_PORT):
         return {"status": "error", "reason": "Session invalid — cookies expired?"}
-    print("  [PREFLIGHT] Session: OK")
+    print(f"  [PREFLIGHT] Session: OK")
 
     # Import websockets here
     import websockets
@@ -830,7 +832,7 @@ async def answer_survey(survey_ws_url: str, max_pages: int = MAX_PAGES) -> Dict:
                 break
 
             # 4. Submit (Next / Continue / Submit)
-            print("  [SUBMIT] Clicking submit button...")
+            print(f"  [SUBMIT] Clicking submit button...")
             submit_result = await click_submit(ws)
             print(f"    [SUBMIT] {submit_result}")
             results["actions"].append(f"SUBMIT:{submit_result}")
@@ -865,7 +867,9 @@ def _update_registry(command_id: str, success: bool, result: Dict):
     except:
         registry = {"version": "1.0.0", "commands": []}
 
-    now = datetime.utcnow().isoformat() + "Z"
+    # SR-187: UTC-aware (naive utcnow() is deprecated in Py 3.12, gone in 3.14).
+    # Keep historical "Z" suffix for command_registry.json wire-format stability.
+    now = datetime.now(timezone.utc).isoformat().replace("+00:00", "Z")
 
     # Find or create command entry
     found = False
@@ -906,7 +910,7 @@ def _update_registry(command_id: str, success: bool, result: Dict):
 
 # ═════════════════════════════════════════════════════════════════════════════
 # CLI — MANUELLE TESTING ONLY!
-# ═════════════════════════════════════════════════════════════════════════════
+# ═════════════════════════════════════════════���═══════════════════════════════
 
 if __name__ == "__main__":
     if len(sys.argv) < 2:
@@ -953,4 +957,4 @@ def _update_registry(command_id: str, success: bool, result: Dict):
     elif result["status"] == "max_pages":
         print("  → Max Seiten erreicht (30). Timeout.")
     else:
-        print(f"  → Unbekannter Status: {result.get('status')}")
+        print(f"  → Unbekannter Status: {result.get('status')}")

survey-cli/survey/captcha/fallback_chain.py

@@ -50,7 +50,7 @@
 import logging
 import time
 from dataclasses import dataclass, field
-from datetime import datetime
+from datetime import datetime, timezone
 from pathlib import Path
 from typing import Any, Callable, Optional
 
@@ -186,15 +186,20 @@ def _log_human_handoff(
         Pfad zur Log-Datei
     """
     logs_dir = _ensure_logs_dir()
-    date_str = datetime.utcnow().strftime("%Y%m%d")
+    # SR-187: UTC-aware datetimes (naive utcnow() is deprecated in Py 3.12,
+    # removed in 3.14; comparing naive against tz-aware silently mis-orders).
+    now_utc = datetime.now(timezone.utc)
+    date_str = now_utc.strftime("%Y%m%d")
     log_path = logs_dir / f"captcha-failures-{date_str}.jsonl"
 
     # Capture screenshot
     screenshot_b64 = _capture_screenshot_b64(cdp)
 
     # Build log entry
     entry = {
-        "timestamp": datetime.utcnow().isoformat() + "Z",
+        # SR-187: isoformat() on tz-aware dt emits "+00:00"; we keep the
+        # historical "Z" suffix for jsonl-consumer compatibility.
+        "timestamp": now_utc.isoformat().replace("+00:00", "Z"),
         "detected_type": detection.captcha_type,
         "page_url": page_url,
         "frame_id": detection.frame_id,

survey-cli/survey/daemon/answer_engine.py

@@ -19,7 +19,7 @@
 import re
 import sqlite3
 from dataclasses import dataclass, field
-from datetime import datetime
+from datetime import datetime, timezone
 from pathlib import Path
 from typing import Any
 
@@ -1002,7 +1002,8 @@ def _store_answer(self, question: Question, answer: Answer) -> None:
                 question.text[:500],
                 json.dumps(answer.value),
                 self._hash_persona(),
-                datetime.utcnow().isoformat(),
+                # SR-187: UTC-aware (naive utcnow() is deprecated in Py 3.12).
+                datetime.now(timezone.utc).isoformat(),
             ),
         )
         conn.commit()

survey-cli/survey/daemon/survey_agent_graph.py

@@ -16,7 +16,7 @@
 import json
 import logging
 import sqlite3
-from datetime import datetime
+from datetime import datetime, timezone
 from enum import Enum
 from pathlib import Path
 from typing import TypedDict
@@ -199,7 +199,8 @@ async def _navigate(self, state: AgentState) -> AgentState:
         """Navigate to survey URL."""
         logger.info(f"Navigating to: {state['survey_url']}")
         state["status"] = SurveyStatus.NAVIGATING.value
-        state["started_at"] = datetime.utcnow().isoformat()
+        # SR-187: UTC-aware (naive utcnow() is deprecated in Py 3.12).
+        state["started_at"] = datetime.now(timezone.utc).isoformat()
 
         if not self._browser:
             self._browser = BrowserDriver(headless=self.headless)
@@ -266,7 +267,8 @@ async def _check_status(self, state: AgentState) -> AgentState:
         for pattern in complete_patterns:
             if pattern in html:
                 state["status"] = SurveyStatus.COMPLETED.value
-                state["completed_at"] = datetime.utcnow().isoformat()
+                # SR-187: UTC-aware (naive utcnow() is deprecated in Py 3.12).
+                state["completed_at"] = datetime.now(timezone.utc).isoformat()
                 return state
 
         # Check for captcha
@@ -582,7 +584,8 @@ async def _complete(self, state: AgentState) -> AgentState:
         """Handle successful completion."""
         logger.info("Survey completed successfully")
         state["status"] = SurveyStatus.COMPLETED.value
-        state["completed_at"] = datetime.utcnow().isoformat()
+        # SR-187: UTC-aware (naive utcnow() is deprecated in Py 3.12).
+        state["completed_at"] = datetime.now(timezone.utc).isoformat()
 
         self._save_state(state)