Merge pull request #416 from tmchow/worktree-inherited-discovering-pebble

fix(ci): run full pytest suite, repair 13 rotted tests

.github/workflows/validate.yml

@@ -10,7 +10,7 @@ permissions:
   contents: read
 
 jobs:
-  plugin-contract:
+  tests:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
@@ -22,5 +22,5 @@ jobs:
       - name: Set up Python
         run: uv python install 3.12
 
-      - name: Run plugin contract tests
+      - name: Run test suite
-        run: uv run pytest tests/test_plugin_contract.py tests/test_version_consistency.py
+        run: uv run pytest

skills/last30days/scripts/store.py

@@ -14,7 +14,7 @@ import argparse
 import json
 import sqlite3
 import sys
-from datetime import datetime, timedelta
+from datetime import datetime, timedelta, timezone
 from pathlib import Path
 from typing import Any, Dict, List, Optional
 
@@ -519,7 +519,7 @@ def get_daily_cost(date: Optional[str] = None) -> float:
     conn = _connect()
     try:
         if not date:
-            date = datetime.now().strftime("%Y-%m-%d")
+            date = datetime.now(timezone.utc).strftime("%Y-%m-%d")
         row = conn.execute(
             """SELECT COALESCE(SUM(token_cost), 0) as total
                FROM research_runs
@@ -575,7 +575,7 @@ def get_stats() -> Dict[str, Any]:
         topic_count = conn.execute("SELECT COUNT(*) FROM topics WHERE enabled = 1").fetchone()[0]
         finding_count = conn.execute("SELECT COUNT(*) FROM findings").fetchone()[0]
 
-        week_ago = (datetime.now() - timedelta(days=7)).strftime("%Y-%m-%d")
+        week_ago = (datetime.now(timezone.utc) - timedelta(days=7)).strftime("%Y-%m-%d")
         runs_7d = conn.execute(
             "SELECT COUNT(*) FROM research_runs WHERE run_date >= ?", (week_ago,)
         ).fetchone()[0]
@@ -621,7 +621,7 @@ def get_trending(days: int = 7) -> List[Dict[str, Any]]:
     """Get topics ranked by recent finding activity."""
     conn = _connect()
     try:
-        since = (datetime.now() - timedelta(days=days)).strftime("%Y-%m-%d")
+        since = (datetime.now(timezone.utc) - timedelta(days=days)).strftime("%Y-%m-%d")
         rows = conn.execute(
             """SELECT t.name, t.id,
                       COUNT(f.id) as new_findings,
@@ -676,24 +676,28 @@ def findings_from_report(
 
     Uses ranked candidates (post-rerank) when available for quality scores and explanations.
     Supplements with raw items from items_by_source for HN/PM that didn't rank highly
-    but are valuable for watchlist persistence.
+    but are valuable for watchlist persistence. When ranked_candidates is empty
+    (degraded path — rerank failed or was skipped), falls back to supplementing
+    all sources from items_by_source so findings aren't silently dropped.
     """
     findings = []
     seen_urls = set()
 
-    # Phase 1: Process ranked candidates (high-quality data with explanations and corroboration)
     for candidate in report.ranked_candidates:
-        finding = finding_from_candidate(candidate)
+        findings.append(finding_from_candidate(candidate))
-        findings.append(finding)
         seen_urls.add(candidate.url)
 
-    # Phase 2: Add HN/PM items not already captured in ranked candidates
+    supplement_sources = (
-    for source_name in ["hackernews", "polymarket"]:
+        list(report.items_by_source)
+        if not report.ranked_candidates
+        else ["hackernews", "polymarket"]
+    )
+    for source_name in supplement_sources:
         if source_name not in report.items_by_source:
             continue
         for item in report.items_by_source[source_name]:
             if item.url in seen_urls:
-                continue  # Already captured with rich data
+                continue
             findings.append({
                 "source": source_name,
                 "source_url": item.url,
@@ -706,7 +710,6 @@ def findings_from_report(
             })
             seen_urls.add(item.url)
 
-    # Apply global limit after collecting all findings (fix: was per-source, now global)
     return findings[:limit] if limit is not None else findings
 
 
@@ -722,9 +725,10 @@ def _cli_query(args):
 
     since = None
     if args.since:
-        # Parse duration like "7d", "30d"
+        # Parse duration like "7d", "30d". Use UTC to match SQLite's
+        # datetime('now') which writes first_seen in UTC.
         days = int(args.since.rstrip("d"))
-        since = (datetime.now() - timedelta(days=days)).strftime("%Y-%m-%d")
+        since = (datetime.now(timezone.utc) - timedelta(days=days)).strftime("%Y-%m-%d")
 
     findings = get_new_findings(topic["id"], since)
     print(json.dumps({"topic": topic["name"], "findings": findings, "count": len(findings)}, default=str))

tests/test_footer_nudge_suppression.py

@@ -6,6 +6,7 @@ from __future__ import annotations
 import os
 import subprocess
 import sys
+import tempfile
 import unittest
 from pathlib import Path
 
@@ -27,13 +28,31 @@ class FooterNudgeSuppressionTests(unittest.TestCase):
             "--emit=md",
             *argv,
         ]
-        env = {**os.environ, "LAST30DAYS_SKIP_PREFLIGHT": "1"}
+        env = {
+            **os.environ,
+            "LAST30DAYS_SKIP_PREFLIGHT": "1",
+            # Skip ~/.config/last30days/.env so a contributor's saved
+            # BRAVE/EXA/SERPER/PARALLEL key doesn't make grounding "available"
+            # and suppress the promo we're checking for.
+            "LAST30DAYS_CONFIG_DIR": "",
+            # Pin X as available so _missing_sources_for_promo selects "web"
+            # (otherwise the "x" promo wins and the BRAVE_API_KEY string never
+            # appears).
+            "XAI_API_KEY": "test-stub",
+        }
         # Strip any grounded-web keys the host might have so the promo path
-        # triggers deterministically in mock + no-backend.
+        # triggers deterministically in mock + no-backend. Also strip X cookie
+        # credentials so XAI_API_KEY is the unambiguous X backend.
         for key in ("BRAVE_API_KEY", "EXA_API_KEY", "SERPER_API_KEY",
-                    "PARALLEL_API_KEY", "OPENROUTER_API_KEY"):
+                    "PARALLEL_API_KEY", "OPENROUTER_API_KEY",
+                    "AUTH_TOKEN", "CT0", "LAST30DAYS_X_BACKEND"):
             env.pop(key, None)
-        return subprocess.run(cmd, capture_output=True, text=True, env=env)
+        # Run from a tmpdir so _find_project_env() can't walk up into any
+        # .claude/last30days.env above the repo on the contributor's machine.
+        with tempfile.TemporaryDirectory() as tmp:
+            return subprocess.run(
+                cmd, capture_output=True, text=True, env=env, cwd=tmp,
+            )
 
     def test_bare_run_emits_web_promo(self):
         result = self._run(topic="OpenAI")

tests/test_setup_openclaw.py

@@ -213,8 +213,10 @@ class TestPollDeviceAuth:
     @patch("lib.setup_wizard.urlopen")
     def test_timeout_returns_none(self, mock_urlopen, mock_time):
         """Returns None when timeout is exceeded."""
-        # Simulate time passing beyond deadline
+        # poll_device_auth calls time.time() for deadline init, last_reminder init,
-        mock_time.time = MagicMock(side_effect=[0, 301])
+        # then once per while-loop iteration. Three values are enough for one check
+        # that exceeds the deadline.
+        mock_time.time = MagicMock(side_effect=[0, 0, 301])
         mock_time.sleep = MagicMock()
 
         result = setup_wizard.poll_device_auth("dc-123", interval=5, timeout=300)
@@ -224,7 +226,9 @@ class TestPollDeviceAuth:
     @patch("lib.setup_wizard.urlopen")
     def test_expired_token_returns_none(self, mock_urlopen, mock_time):
         """Returns None on expired_token error."""
-        mock_time.time = MagicMock(side_effect=[0, 0])
+        # Loop terminates via urlopen response, not the clock — pin time to 0
+        # so the deadline check stays a non-event regardless of call count.
+        mock_time.time = MagicMock(return_value=0)
         mock_time.sleep = MagicMock()
 
         expired_resp = MagicMock()
@@ -243,7 +247,7 @@ class TestPollDeviceAuth:
         """HTTP 400 during polling continues (authorization pending)."""
         from urllib.error import HTTPError
 
-        mock_time.time = MagicMock(side_effect=[0, 0, 0])
+        mock_time.time = MagicMock(return_value=0)
         mock_time.sleep = MagicMock()
 
         success_resp = MagicMock()

tests/test_store.py

@@ -3,7 +3,7 @@
 import json
 import sqlite3
 import tempfile
-from datetime import datetime, timedelta
+from datetime import datetime, timedelta, timezone
 from pathlib import Path
 
 import pytest
@@ -503,14 +503,14 @@ def test_get_new_findings_filters_by_date(temp_db, sample_report):
     findings = store.findings_from_report(sample_report)
     store.store_findings(run_id, topic["id"], findings)
     
-    # Get findings since tomorrow (should be empty)
+    # Use UTC because store writes first_seen via SQLite's datetime('now') (UTC).
-    tomorrow = (datetime.now() + timedelta(days=1)).strftime("%Y-%m-%d")
+    # Local-time math here would flake near midnight UTC.
+    tomorrow = (datetime.now(timezone.utc) + timedelta(days=1)).strftime("%Y-%m-%d")
     new_findings = store.get_new_findings(topic["id"], since=tomorrow)
 
     assert len(new_findings) == 0
 
-    # Get findings since yesterday (should have all)
+    yesterday = (datetime.now(timezone.utc) - timedelta(days=1)).strftime("%Y-%m-%d")
-    yesterday = (datetime.now() - timedelta(days=1)).strftime("%Y-%m-%d")
     new_findings = store.get_new_findings(topic["id"], since=yesterday)
 
     assert len(new_findings) == 4