功能：每日动态股票池回测（--mode daily）+ 每日增量同步 + PIT 批量取数层

说明：本提交是工作区中此前的未提交工作（在 14ec0c6 之后产生），**非本次会话所写**，
按用户要求整理并推送。已做安全检查（无明文凭据、无大文件、.env/logs/output 仍被忽略），
并完成可执行范围内的测试验证（见「测试」一节）。

## 新增能力

1) `hdiv backtest --mode daily --start <日期>`
   - src/hdiv/backtest/daily.py：两趟式（先逐日选股，再复用既有引擎模拟）
   - 每个交易日按当日可见数据重建股票池（PIT），每个交易日判断买卖点
   - `pool_exit_action`：hold（只减不加、不因掉出池子而清仓）/ sell（掉出即清仓）
   - `profile_on_trade`：买卖决策发生时计算并留痕个股画像，**不区分是否在当日池内**
     （卖出/减仓同样留痕，否则「为什么卖」缺证据）
   - 与 walkforward 的分工：daily 是一条连续路径的推演，不是过拟合检验；
     因此不使用训练段、不冻结分布，阈值口径一律 rolling
   - 拒绝 `--universe-run`（daily 的定义就是逐日重筛，冻结池与之矛盾）

2) PIT 批量取数层 src/hdiv/universe/pit.py
   - PitRepo 继承 Repo，**只重写取数**（按区块批量预载 + 逐日内存切片），
     派生逻辑（最新一期财报合并、单位归一化、支付率口径等）一行不重写
     —— 以保证与逐日单点查询**结果等价**
   - 候选集预剪枝：用「不可能通过」的边界条件提前排除，文档论证为精确等价而非近似
   - src/hdiv/universe/daily.py：每日动态筛选器（仍然调用既有 selector 与四个 Filter）

3) 每日增量同步 `hdiv sync daily`
   - src/hdiv/data/sync/daily.py：只抓「库里还没有的那几天」，
     按「当日股票数 ≥ 当年规模阈值」判定缺口，不重拉历史、不覆盖既有行；
     支持 `--dry-run` 先看待抓清单
   - deploy/daily-sync.sh、deploy/install-sync-schedule.sh、
     deploy/com.hddiv.sync.plist.example（launchd 每天 17:00）
   - 新表 hd_daily_universe（逐日入选成员留痕）+ sql/hd_daily_universe.sql + schema.py
     （该表已存在于库中，`ddl plan` 返回 0 个待执行动作）

4) Web 与文档
   - 前端支持 daily 模式记录下钻（web/app.js、web/app.css、web/index.html、
     web/favicon.svg）
   - README / docs/user-guide.md / docs/implementation-status.md 同步更新：
     三种回测模式的取舍、daily 的成本说明（6.7 年约 1.5 小时）与调优手段

## 测试

tests/ 共 500 项（新增 tests/test_daily.py 43 项、tests/test_sync_daily.py 36 项）。

已验证通过：
- 排除上述两个新文件的 **421 项：全部通过（pytest 退出码 0）**
- 两个新文件的**非 DB 单元测试 60 项：全部通过**

未能在合理时间内跑完：
- 两个新文件中 **19 项 DB 标记的重型测试**。实测瓶颈是一条**无界全表扫描**：
  `SELECT ... FROM hd_cashflow WHERE ann_date <= :asof ORDER BY symbol, end_date, ann_date`
  （31 万行，无 symbol/报告期下限）。全量套件跑到 161 项时已耗时 20 分钟、
  0 失败，按该速率预计需 3 小时以上，因此改为分档验证。
- 旁证：库中存在 3 次成功的 daily 端到端运行（2026-10-05 10:05 / 10:32 / 11:03，
  区间 2024-03-01~03-15），说明该路径可正常完成。

## 已知待改进

- 上述 `hd_cashflow`（及同类「按 ann_date 上界取全历史」）的查询缺
  symbol / 报告期下限，是 daily 模式的主要性能瓶颈，建议下一轮优化。

tests/test_sync_daily.py

@@ -0,0 +1,449 @@
+"""每日增量同步（``hdiv sync daily``）测试。
+
+要锁定的三条性质：
+
+1. **缺口口径与断点续传一致** —— 判定「这天已同步」用的是 ``price.fetched_days``
+   的按年份规模阈值，而不是「当天有没有行」。否则只填了几百只的半成品日会被
+   当成已完成，形成难以察觉的数据空洞（历史上真的踩过这个坑）。
+
+2. **只抓缺口，不重拉** —— 计划里出现的交易日必须**恰好**是缺失的那些；
+   已完整的日、已最新的指数、已有数据的股票都不能出现在待抓清单里。
+
+3. **财报水位按披露截止日推算** —— 年报/一季报 4-30、半年报 8-31、三季报 10-31。
+   水位算错会让财报季的队列要么永远排不空、要么漏掉整季新披露。
+"""
+
+from __future__ import annotations
+
+import os
+import plistlib
+from datetime import date
+
+import pytest
+
+from hdiv.core.paths import project_root
+from hdiv.data.sync import daily as daily_sync
+
+# ---------------------------------------------------------------------------
+# 财报报告期水位
+# ---------------------------------------------------------------------------
+
+
+class TestFinancialWatermark:
+    @pytest.mark.parametrize(
+        "asof,expected",
+        [
+            # 三季报 10-31 截止 → 11-01 起水位是当年三季报
+            (date(2026, 11, 1), date(2026, 9, 30)),
+            (date(2026, 12, 31), date(2026, 9, 30)),
+            # 半年报 8-31 截止 → 9-01 起水位是当年半年报
+            (date(2026, 9, 1), date(2026, 6, 30)),
+            (date(2026, 10, 31), date(2026, 6, 30)),
+            # 一季报 4-30 截止 → 5-01 起水位是当年一季报
+            (date(2026, 5, 1), date(2026, 3, 31)),
+            (date(2026, 8, 31), date(2026, 3, 31)),
+            # 1~4 月是上一年年报季
+            (date(2026, 4, 30), date(2025, 12, 31)),
+            (date(2026, 1, 1), date(2025, 12, 31)),
+        ],
+    )
+    def test_watermark(self, asof: date, expected: date) -> None:
+        assert daily_sync.financial_watermark(asof) == expected
+
+    def test_watermark_boundaries_are_monotonic(self) -> None:
+        """水位只能随时间前进，不能回退 —— 否则已补的股票会反复进队列。"""
+        days = [date(2025, m, d) for m in range(1, 13) for d in (1, 15, 28)]
+        days += [date(2026, m, d) for m in range(1, 13) for d in (1, 15, 28)]
+        wm = [daily_sync.financial_watermark(d) for d in sorted(days)]
+        assert wm == sorted(wm)
+
+
+# ---------------------------------------------------------------------------
+# 目标选择
+# ---------------------------------------------------------------------------
+
+
+class TestResolveOnly:
+    def test_default_is_all_groups(self) -> None:
+        assert daily_sync._resolve_only(None) == set(daily_sync.GROUPS)
+        assert daily_sync._resolve_only([]) == set(daily_sync.GROUPS)
+
+    def test_group_names(self) -> None:
+        assert daily_sync._resolve_only(["price"]) == {"price"}
+        assert daily_sync._resolve_only(["price", "dividend"]) == {"price", "dividend"}
+
+    def test_single_target_maps_to_its_group(self) -> None:
+        assert daily_sync._resolve_only(["daily_basic"]) == {"price"}
+        assert daily_sync._resolve_only(["suspend"]) == {"trading"}
+        assert daily_sync._resolve_only(["limit"]) == {"trading"}
+        assert daily_sync._resolve_only(["index_weight"]) == {"index"}
+
+    def test_unknown_target_is_rejected(self) -> None:
+        with pytest.raises(ValueError, match="未知的同步目标"):
+            daily_sync._resolve_only(["nope"])
+
+    def test_blank_entries_are_ignored(self) -> None:
+        assert daily_sync._resolve_only(["", "  ", "index"]) == {"index"}
+
+
+# ---------------------------------------------------------------------------
+# 缺口判定（隔离数据库）
+# ---------------------------------------------------------------------------
+
+
+class TestMissingDays:
+    def test_only_incomplete_days_are_returned(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        """4 个交易日里只有 2 天完整 → 待抓恰好是另外 2 天。"""
+        days = [date(2026, 9, 1), date(2026, 9, 2), date(2026, 9, 3), date(2026, 9, 4)]
+        monkeypatch.setattr(daily_sync, "_trading_days", lambda s, e, cfg: days)
+        monkeypatch.setattr(
+            daily_sync.price_sync, "fetched_days", lambda *a, **k: {days[0], days[2]}
+        )
+        out = daily_sync.missing_days("stock_daily", days[0], days[-1], cfg=None)
+        assert out == [days[1], days[3]]
+
+    def test_min_symbols_is_forwarded(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        """停牌表必须按「有行即算」传 min_symbols=1，否则每天都判为未完成。"""
+        seen: dict = {}
+
+        def fake_fetched(table, start, end, cfg, *, min_symbols=None):  # noqa: ANN001
+            seen["min_symbols"] = min_symbols
+            return set()
+
+        monkeypatch.setattr(daily_sync, "_trading_days", lambda s, e, cfg: [date(2026, 9, 1)])
+        monkeypatch.setattr(daily_sync.price_sync, "fetched_days", fake_fetched)
+        daily_sync.missing_days("hd_suspend", date(2026, 9, 1), date(2026, 9, 1), cfg=None, min_symbols=1)
+        assert seen["min_symbols"] == 1
+
+    def test_no_trading_days_means_no_gap(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        monkeypatch.setattr(daily_sync, "_trading_days", lambda s, e, cfg: [])
+        assert daily_sync.missing_days("stock_daily", date(2026, 10, 1), date(2026, 10, 5), cfg=None) == []
+
+
+# ---------------------------------------------------------------------------
+# 分红按日期补
+# ---------------------------------------------------------------------------
+
+
+class TestDividendGaps:
+    def test_future_ex_date_does_not_push_window_forward(
+        self, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        """已公告但尚未除权的记录会把 MAX(ex_date) 推到未来。
+
+        若直接拿它当锚点，窗口会落在未来、一个交易日都查不到 ——
+        新公告的分红就永远补不进来。锚点必须被夹到 asof。
+        """
+        import pandas as pd
+
+        asof = date(2026, 10, 5)
+        calls: list[tuple] = []
+
+        def fake_read_sql(sql, params=None, *, cfg=None):  # noqa: ANN001
+            if "MAX(ann_date)" in sql:
+                return pd.DataFrame({"m": [date(2026, 10, 23)]})  # 未来除权日
+            calls.append((params["s"], params["e"]))
+            return pd.DataFrame({"d": []})
+
+        monkeypatch.setattr(daily_sync.db, "read_sql", fake_read_sql)
+        monkeypatch.setattr(
+            daily_sync,
+            "_trading_days",
+            lambda s, e, cfg: [date(2026, 9, 28), date(2026, 9, 29), date(2026, 9, 30)],
+        )
+        days, anchor = daily_sync.dividend_gaps(date(2026, 8, 21), asof, asof, cfg=None)
+        assert anchor == asof
+        assert days == [date(2026, 9, 28), date(2026, 9, 29), date(2026, 9, 30)]
+        start, _ = calls[0]
+        assert start == date(2026, 9, 28), "窗口起点应是 asof 往前 DIVIDEND_OVERLAP_DAYS 天"
+
+    def test_days_already_queried_are_not_repeated(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        import pandas as pd
+
+        asof = date(2026, 10, 5)
+        seen_day = date(2026, 9, 30)
+
+        def fake_read_sql(sql, params=None, *, cfg=None):  # noqa: ANN001
+            if "MAX(ann_date)" in sql:
+                return pd.DataFrame({"m": [seen_day]})
+            return pd.DataFrame({"d": [seen_day]})
+
+        monkeypatch.setattr(daily_sync.db, "read_sql", fake_read_sql)
+        monkeypatch.setattr(
+            daily_sync, "_trading_days", lambda s, e, cfg: [seen_day, date(2026, 10, 1)]
+        )
+        days, _ = daily_sync.dividend_gaps(date(2026, 8, 21), asof, asof, cfg=None)
+        assert days == [date(2026, 10, 1)]
+
+
+# ---------------------------------------------------------------------------
+# 指数行情 / 成分权重窗口
+# ---------------------------------------------------------------------------
+
+
+class TestIndexGaps:
+    def test_per_index_gap_starts_after_its_last_day(
+        self, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        import pandas as pd
+
+        days = [date(2026, 9, 29), date(2026, 9, 30)]
+        monkeypatch.setattr(daily_sync, "_trading_days", lambda s, e, cfg: days)
+        monkeypatch.setattr(
+            daily_sync.index_sync,
+            "default_indices",
+            lambda cfg: [{"code": "000300.SH", "name": "沪深300"}, {"code": "399006.SZ"}],
+        )
+        monkeypatch.setattr(
+            daily_sync.db,
+            "read_sql",
+            lambda *a, **k: pd.DataFrame(
+                {"c": ["000300.SH"], "m": [date(2026, 9, 29)]}
+            ),
+        )
+        out = daily_sync.index_daily_gaps(date(2026, 9, 1), date(2026, 9, 30), cfg=None)
+        assert out == {"000300.SH": [date(2026, 9, 30)], "399006.SZ": days}
+
+    def test_up_to_date_table_has_no_window(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        import pandas as pd
+
+        monkeypatch.setattr(
+            daily_sync, "_trading_days", lambda s, e, cfg: [date(2026, 9, 29), date(2026, 9, 30)]
+        )
+        monkeypatch.setattr(
+            daily_sync.db, "read_sql", lambda *a, **k: pd.DataFrame({"m": [date(2026, 9, 30)]})
+        )
+        assert daily_sync._index_weight_window(date(2026, 8, 21), date(2026, 9, 30), None) is None
+
+    def test_window_end_is_clamped_to_last_trading_day(
+        self, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        """右端点收到最近交易日：否则长假里每天都会重查一段空区间。"""
+        import pandas as pd
+
+        monkeypatch.setattr(
+            daily_sync, "_trading_days", lambda s, e, cfg: [date(2026, 9, 29), date(2026, 9, 30)]
+        )
+        monkeypatch.setattr(daily_sync.db, "read_sql", lambda *a, **k: pd.DataFrame({"m": [None]}))
+        got = daily_sync._index_weight_window(date(2026, 8, 21), date(2026, 10, 5), None)
+        assert got == (date(2026, 8, 21), date(2026, 9, 30))
+
+    def test_holiday_only_window_is_skipped(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        """整段区间都是假期 → 没有可补的权重日，一次调用都不该发。"""
+        monkeypatch.setattr(daily_sync, "_trading_days", lambda s, e, cfg: [])
+        assert daily_sync._index_weight_window(date(2026, 10, 1), date(2026, 10, 5), None) is None
+
+    def test_incremental_window_starts_after_last_weight(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        import pandas as pd
+
+        monkeypatch.setattr(daily_sync, "_trading_days", lambda s, e, cfg: [date(2026, 9, 30)])
+        monkeypatch.setattr(
+            daily_sync.db, "read_sql", lambda *a, **k: pd.DataFrame({"m": [date(2026, 8, 31)]})
+        )
+        got = daily_sync._index_weight_window(date(2026, 8, 21), date(2026, 10, 5), None)
+        assert got == (date(2026, 9, 1), date(2026, 9, 30))
+
+
+# ---------------------------------------------------------------------------
+# 计划对象与摘要
+# ---------------------------------------------------------------------------
+
+
+class TestPlanObject:
+    def test_empty_plan_detects_nothing_to_do(self) -> None:
+        p = daily_sync.DailyPlan(asof=date(2026, 10, 5), window_start=date(2026, 8, 21))
+        assert p.empty
+        assert p.total_days == 0
+
+    def test_any_gap_makes_plan_non_empty(self) -> None:
+        p = daily_sync.DailyPlan(
+            asof=date(2026, 10, 5),
+            window_start=date(2026, 8, 21),
+            day_gaps={"daily": [date(2026, 9, 30)]},
+        )
+        assert not p.empty
+        assert p.total_days == 1
+
+    def test_plan_does_not_share_default_dicts(self) -> None:
+        """两个计划的默认容器必须互相独立（dataclass 默认值的经典坑）。"""
+        a = daily_sync.DailyPlan(asof=date(2026, 10, 5), window_start=date(2026, 8, 21))
+        b = daily_sync.DailyPlan(asof=date(2026, 10, 5), window_start=date(2026, 8, 21))
+        a.day_gaps["daily"] = [date(2026, 9, 30)]
+        a.notes.append("x")
+        assert b.day_gaps == {}
+        assert b.notes == []
+
+
+class TestFormatSummary:
+    def test_success(self) -> None:
+        assert "成功" in daily_sync.format_summary({"targets": {}, "errors": [], "ok": True})
+
+    def test_failure_lists_reasons(self) -> None:
+        s = daily_sync.format_summary({"errors": ["daily: SyncError: 限频"], "ok": False})
+        assert "失败 1 项" in s
+        assert "限频" in s
+
+    def test_dry_run_reports_plan(self) -> None:
+        s = daily_sync.format_summary(
+            {
+                "dry_run": True,
+                "planned": {"day_gaps": {"daily": 3}, "index_days": {"000300.SH": 2},
+                            "dividend_days": 1},
+            }
+        )
+        assert "计划" in s
+        assert "5" in s
+
+
+# ---------------------------------------------------------------------------
+# 部署契约：每天 17:00
+# ---------------------------------------------------------------------------
+
+
+class TestScheduleContract:
+    """「每天下午 5 点定时抓取」是本次的硬需求，必须有可执行的证据。
+
+    定时任务最容易在交付后悄悄失效：plist 写成 KeepAlive 会把一次性任务变成
+    常驻进程、小时写错会让它在收盘前跑到空数据。这两点用测试钉死。
+    """
+
+    def test_plist_triggers_at_17_00(self) -> None:
+        path = project_root() / "deploy" / "com.hddiv.sync.plist.example"
+        assert path.is_file(), "缺少 launchd 模板"
+        data = plistlib.loads(path.read_bytes())
+        assert data["Label"] == "com.hddiv.sync"
+        assert data["StartCalendarInterval"] == {"Hour": 17, "Minute": 0}, (
+            "触发时间必须是 17:00（收盘 15:00 后，当日数据已出）"
+        )
+        assert data["KeepAlive"] is False, "定时任务不能 KeepAlive —— 否则会被无限拉起"
+        assert data["RunAtLoad"] is False, "RunAtLoad 会让每次登录都多抓一次"
+
+    def test_runner_script_is_wired(self) -> None:
+        root = project_root()
+        runner = root / "deploy" / "daily-sync.sh"
+        assert runner.is_file(), "缺少执行包装脚本"
+        assert os.access(runner, os.X_OK), "daily-sync.sh 必须可执行"
+        text = runner.read_text(encoding="utf-8")
+        assert "sync daily" in text, "包装脚本必须调用 hdiv sync daily"
+        assert "LOCK_DIR" in text, "包装脚本必须有单实例锁（避免重叠运行互相抢限频额度）"
+        installer = root / "deploy" / "install-sync-schedule.sh"
+        assert installer.is_file(), "缺少安装脚本"
+        itext = installer.read_text(encoding="utf-8")
+        assert "com.hddiv.sync.plist.example" in itext
+        assert "StartCalendarInterval" in (
+            root / "deploy" / "com.hddiv.sync.plist.example"
+        ).read_text(encoding="utf-8")
+
+
+# ---------------------------------------------------------------------------
+# 真实数据库：计划与执行的一致性
+# ---------------------------------------------------------------------------
+
+
+@pytest.mark.db
+def test_plan_matches_database_state() -> None:
+    """计划里的缺口必须与「库里实际缺的日子」逐日一致。
+
+    这是本模块唯一不能靠 mock 保证的性质：``fetched_days`` 的阈值口径、
+    ``trading_calendar`` 的交易日、各表的日期列含义都来自真实库。
+    """
+    from hdiv.core.config import load_config
+    from hdiv.data import db
+    from hdiv.data.sync import price as price_sync
+
+    try:
+        db.load_dotenv_once()
+        cfg = load_config("datasource")
+        db.row_count("stock_daily", cfg)
+    except Exception as exc:  # pragma: no cover
+        pytest.skip(f"数据库不可用：{exc}")
+
+    p = daily_sync.plan(cfg, asof=date(2026, 9, 30), lookback_days=45, include_financial=False)
+    # 窗口内每一天：要么在待抓清单里，要么在「已完整」集合里，二者互斥且完备
+    window_days = price_sync.open_days(date(2026, 8, 16), date(2026, 9, 30), cfg)
+    done = price_sync.fetched_days("stock_daily", date(2026, 8, 16), date(2026, 9, 30), cfg)
+    planned = set(p.day_gaps["daily"])
+    for d in window_days:
+        assert (d in planned) != (d in done), f"{d} 的缺口判定与 fetched_days 不一致"
+    assert planned == {d for d in window_days if d not in done}
+
+    # 计划不包含窗口之外的日期（增量只补近期空洞）
+    assert all(d >= p.window_start for days in p.day_gaps.values() for d in days)
+
+
+@pytest.mark.db
+def test_financial_queue_excludes_long_delisted() -> None:
+    """已退市多年的标的不能再进每日队列。
+
+    回归：实测「报告期滞后」的股票里 200+ 只早已退市（最后一份财报停在退市前），
+    不过滤的话每天都要为它们发 800 次调用、拉回同一批旧数据 —— 实测一次全量
+    重拉耗时约 2 分钟、返回 6.5 万行，全部无功而返。
+    """
+    from hdiv.core.config import load_config
+    from hdiv.data import db
+
+    asof = date(2026, 9, 30)
+    try:
+        db.load_dotenv_once()
+        cfg = load_config("datasource")
+        db.row_count("stock", cfg)
+    except Exception as exc:  # pragma: no cover
+        pytest.skip(f"数据库不可用：{exc}")
+
+    p = daily_sync.plan(cfg, asof=asof)
+    assert len(p.financial_symbols) < 50, (
+        f"每日财报队列 {len(p.financial_symbols)} 只，退市过滤可能失效"
+    )
+    cutoff = daily_sync._active_since(asof)
+    dead = db.read_sql(
+        "SELECT symbol FROM stock WHERE delist_date IS NOT NULL AND delist_date <= :c",
+        {"c": cutoff},
+        cfg=cfg,
+    )
+    dead_set = set(dead["symbol"].astype(str)) if not dead.empty else set()
+    assert not (set(p.financial_symbols) & dead_set), "长期退市标的混进了每日财报队列"
+
+
+@pytest.mark.db
+def test_json_mode_writes_no_progress_text(capsys: pytest.CaptureFixture[str]) -> None:
+    """``--json`` 的 stdout 必须是**纯 JSON**。
+
+    回归：进度与计划文字混进 stdout 会让 ``| jq`` 之类的消费者直接解析失败，
+    而这类问题在交互式运行时完全看不出来（人眼只看到 JSON 在前面或后面）。
+    """
+    from hdiv.core.config import load_config
+    from hdiv.data import db
+
+    try:
+        db.load_dotenv_once()
+        cfg = load_config("datasource")
+        db.row_count("stock", cfg)
+    except Exception as exc:  # pragma: no cover
+        pytest.skip(f"数据库不可用：{exc}")
+
+    summary = daily_sync.run(
+        cfg, asof=date(2026, 9, 30), dry_run=True, include_financial=False, verbose=False
+    )
+    captured = capsys.readouterr()
+    assert captured.out == "", f"verbose=False 仍有 stdout 输出：{captured.out[:200]}"
+    assert summary["dry_run"] is True
+
+
+@pytest.mark.db
+def test_dry_run_writes_nothing() -> None:
+    """``--dry-run`` 不得写库，也不得调用 Tushare。"""
+    from hdiv.core.config import load_config
+    from hdiv.data import db
+
+    try:
+        db.load_dotenv_once()
+        cfg = load_config("datasource")
+        logs_before = db.read_sql("SELECT MAX(id) AS m FROM hd_sync_log", cfg=cfg)["m"].iloc[0]
+    except Exception as exc:  # pragma: no cover
+        pytest.skip(f"数据库不可用：{exc}")
+
+    summary = daily_sync.run(cfg, asof=date(2026, 9, 30), dry_run=True, include_financial=False)
+    assert summary["dry_run"] is True
+    assert summary["errors"] == []
+    after_logs = db.read_sql("SELECT MAX(id) AS m FROM hd_sync_log", cfg=cfg)["m"].iloc[0]
+    assert after_logs == logs_before, "dry-run 不应产生新的 hd_sync_log 记录"