xiaotutongxue/scripts/e2e-m8-orgfile.py

# -*- coding: utf-8 -*-
"""组织导入 / 租户用量 / 文件视图·直传凭证·批量处理 E2E

（《03-接口设计》§6.4、§6.5、§6.7、§9.5、§9.7、§9.8）

## 这批接口里三处"看着像细节、其实决定对不对"的地方

  ① **组织导入的父子关系用「编码 + 上级编码」表达，不靠行的顺序**。
     客户的表是人整理出来的：顺序会被重排、缩进会被吃掉（Excel 里前导空格尤其容易丢）。
     所以本脚本特意把**子行排在父行之前** —— 那是客户的表里最常见的形态，
     也是"一遍扫描"实现会全部失败的那种。
  ② **租户用量的 `ocrCount` / `smsCount` 恒为 0 并带 notes**。
     两条通道都还没接，编一个数字会让运营拿着假数据去谈续费。
     断言要看的是"**它承认自己没接**"，而不是"字段存在"。
  ③ **上传凭证的 `existed=true` 时不返回 uploadUrl**。
     返回了而前端不用，会让"到底传没传"变成两处判断 —— 迟早不一致。
  ④ **租户用量必须只算"这一所学校"**。平台运营的 `tenantId` 是 0，而租户拦截器
     对它**整张表关掉过滤**（`FillTenantLineHandler.ignoreTable`）—— 于是
     `/tenants/{id}/usage` 会算成**全平台**的学生数与存储量，挂在某一所学校的名下。
     ⚠️ 这个 bug 在"库里只有一所学校有数据"时**与正确值完全相等**，光断言
     "有值、非 null"永远发现不了 —— 所以本脚本要**先造出别的租户的数据**再断言。

## 还有一条不能省的边界

`GET /mp/files/{fileId}` 按 `submitter_id` 授权（学生端）；教师看他班学生的材料
走 `/mp/teacher/files/{id}/preview`（按班级范围）。本脚本要证明**前者的越权会被拒** ——
不证明的话，"任何登录用户能看任何人的身份证照片"这种事只在真实越权时才被发现。

跑法：python scripts/e2e-m8-orgfile.py
"""
import argparse
import datetime
import importlib.util
import json
import os
import sys
import urllib.error
import urllib.parse

sys.stdout.reconfigure(encoding="utf-8")

ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
PASSED = []
FAILED = []
CREATED_TASK_IDS = []
CREATED_ORG_NAMES = []
CREATED_TENANT_IDS = []
USAGE_FIXTURE_HASH_PREFIX = "e2e-usage-scope-"
TMP_ROSTER = os.path.join(ROOT, "scripts", "_e2e-of-roster.xlsx")
TMP_ORG = os.path.join(ROOT, "scripts", "_e2e-of-orgs.xlsx")
TINY_JPEG = os.path.join(ROOT, "scripts", "_e2e-of.jpg")


def _load(name, fn):
    spec = importlib.util.spec_from_file_location(name, os.path.join(ROOT, "scripts", fn))
    mod = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(mod)
    return mod


R = _load("e2e_m2_roster", "e2e-m2-roster.py")
A = _load("e2e-m5-audit", "e2e-m5-audit.py")
BASE = "http://localhost:8080"

STU = 997201
OTHER = 997202


def ok(name):
    PASSED.append(name)
    print("  [OK]   " + name)


def fail(name, detail):
    FAILED.append((name, detail))
    print("  [FAIL] " + name + "\n         " + detail)


def check(name, cond, detail=""):
    if cond:
        ok(name)
    else:
        fail(name, detail or "断言不成立")
    return bool(cond)


def sql(stmt):
    _, out = R.sql(stmt)
    return out


def body_of(method, path, token=None, payload=None):
    _, b = R.call(method, BASE + path, token, payload)
    return b


def mp(uid):
    cfg = R.read_dev_config()
    return R.mint_mp_token(cfg["jwt_secret"], cfg["issuer"], uid, 1, ["STUDENT"], 4)


def upload(path, token, *fields):
    """用 curl 发 multipart（urllib 手搓 multipart 容易出错）。"""
    import subprocess
    args = ["curl", "-s", "-X", "POST", BASE + path, "-H", "Authorization: Bearer " + token]
    for f in fields:
        args += ["-F", f]
    r = subprocess.run(args, capture_output=True, text=True, encoding="utf-8", timeout=120)
    try:
        return json.loads(r.stdout)
    except Exception:
        return {"code": -1, "msg": (r.stdout or r.stderr or "")[:300]}


def main():
    ap = argparse.ArgumentParser(description="组织/租户/文件 E2E")
    ap.parse_args()

    print("=" * 74)
    print(" 组织导入 / 租户用量 / 文件视图·直传·批量处理 E2E")
    print("=" * 74)

    # ⚠️ 租户相关接口（§6.5 修改 / §6.7 用量）的权限是**运营**（平台级），
    #    学校管理员调它们会得到 40301 —— 那是权限边界在生效，不是缺陷。
    #    所以这里用平台账号。**先清登录失败计数**：`LoginGuard` 连续失败后会要求
    #    图形验证码（`xt:login:fail:{account}`），而验证码是图片、脚本给不出来。
    #    这不是绕过安全，而是"测试夹具要能清掉自己造成的失败计数"。
    for acct in ("platform", "admin"):
        os.system('docker exec xt-redis redis-cli -a redis_dev_pwd --no-auth-warning '
                  'DEL xt:login:fail:%s > NUL 2>&1' % acct)
    # 平台运营账号的密码在 docs/07-开发实施计划.md 里（Platform@123），不是 Admin@123
    _, ops = R.login(BASE, "platform", "Platform@123", None)
    if ops.get("code") != 0:
        print("  x 平台运营账号登录失败（先清 Redis 里的登录失败计数）：" + str(ops))
        return 2
    ops_token = ops["data"]["token"]

    _, adm = R.login(BASE, "admin", "Admin@123", "demo")
    if adm.get("code") != 0:
        print("  x 管理员登录失败：" + str(adm))
        return 2
    at = adm["data"]["token"]

    # ── 1. §6.4 组织导入 ──
    print("\n[1] §6.4 组织批量导入")
    suffix = os.urandom(3).hex()
    root = "E2E学院" + suffix
    mid = "E2E系" + suffix
    leaf = "E2E班" + suffix
    CREATED_ORG_NAMES.extend([root, mid, leaf])
    # ⚠️ 刻意把**子行排在父行之前**：客户的表里最常见，
    #    也是"一遍扫描"实现会全部失败的那种形态
    rows = [
        # 层级 / 名称 / 编码 / 上级编码 / 负责人手机号
        ("3", leaf, "C" + suffix, "B" + suffix, ""),
        ("1", root, "A" + suffix, "", ""),
        ("2", mid, "B" + suffix, "A" + suffix, ""),
    ]
    _build_org_xlsx(rows)
    dry = upload("/api/v1/orgs/import", at,
                 "file=@%s;filename=orgs.xlsx;type=application/vnd.openxmlformats-officedocument.spreadsheetml.sheet" % TMP_ORG,
                 "dryRun=true")
    check("dryRun 可用", dry.get("code") == 0, json.dumps(dry, ensure_ascii=False)[:250])
    dd = dry.get("data") or {}
    check("⭐ dryRun 报告 created=3（子行在前也能算对层级）", dd.get("created") == 3,
          json.dumps(dd, ensure_ascii=False)[:250])
    check("dryRun **没有**落库",
          sql("select count(*) from sys_org where org_name='%s';" % leaf).strip() == "0",
          sql("select count(*) from sys_org where org_name='%s';" % leaf))

    imp = upload("/api/v1/orgs/import", at,
                 "file=@%s;filename=orgs.xlsx;type=application/vnd.openxmlformats-officedocument.spreadsheetml.sheet" % TMP_ORG,
                 "dryRun=false")
    check("正式导入成功", imp.get("code") == 0, json.dumps(imp, ensure_ascii=False)[:250])
    d = imp.get("data") or {}
    check("created=3", d.get("created") == 3, json.dumps(d, ensure_ascii=False))
    check("失败 0 / 报错列表为空", d.get("failed") == 0 and not d.get("errors"),
          json.dumps(d, ensure_ascii=False)[:250])
    check("⭐ 库里真的有这三个组织",
          sql("select count(*) from sys_org where org_name in ('%s','%s','%s');"
              % (root, mid, leaf)).strip() == "3",
          sql("select org_name, org_code from sys_org where org_name like 'E2E%%%s';" % suffix)
          .replace("\n", " | "))
    # 父子关系：子行的 parent 是父行的 id
    check("⭐ 子行的父级接对了（按上级编码，而不是按行序）",
          sql("""select count(*) from sys_org c
                 join sys_org p on p.id = c.parent_id
                 where c.org_name = '%s' and p.org_name = '%s';""" % (mid, root)).strip() == "1",
          sql("""select c.org_name, p.org_name from sys_org c
                 left join sys_org p on p.id = c.parent_id
                 where c.org_name in ('%s','%s');""" % (mid, leaf)).replace("\n", " | "))
    check("⭐ ancestor_path 被补上了（按层级筛选靠它，漏了会静默少数据）",
          sql("select coalesce(ancestor_path,'') from sys_org where org_name='%s';"
              % leaf).strip() not in ("", "None"),
          "leaf.ancestor_path = " + sql("select coalesce(ancestor_path,'(空)') "
                                        "from sys_org where org_name='%s';" % leaf).strip())
    # 重复导入：已存在的按编码命中，不重复建
    imp2 = upload("/api/v1/orgs/import", at, "file=@%s;filename=orgs.xlsx" % TMP_ORG)
    check("⭐ 重复导入按编码命中已存在的（不重复建）",
          (imp2.get("data") or {}).get("updated") == 3
          and sql("select count(*) from sys_org where org_name='%s';" % leaf).strip() == "1",
          json.dumps(imp2.get("data"), ensure_ascii=False)[:250])

    # ⚠️ 「认不出表头」这一条必须**放在最后**：它会把坏表头写进同一个临时文件，
    #    而上面的"重复导入"复用那份文件 —— 顺序反了的话，
    #    重复导入实际导的是坏表头那份，报错信息与真正原因完全无关
    # 认不出表头要明确报错
    _build_org_xlsx([("3", "X", "C", "B", "")], header=["甲", "乙", "丙", "丁", "戊"])
    bad = upload("/api/v1/orgs/import", at, "file=@%s;filename=bad.xlsx" % TMP_ORG)
    check("⭐ 认不出表头 → 明确报错并给出应有的列名",
          bad.get("code") != 0 and "组织名称" in (bad.get("msg") or ""),
          "code=%s msg=%s" % (bad.get("code"), bad.get("msg")))

    # ── 2. §6.5 / §6.7 租户修改与用量 ──
    print("\n[2] §6.5 / §6.7 租户修改与用量")
    # 学 校 管 理 员 调 租 户 接 口 应 该 被 拒（权限边界）
    forbidden = body_of("PUT", "/api/v1/tenants/1", at, {"contactName": "E2E不该改成"})
    check("⭐ 学校管理员改租户 → 40301（这是「运营」权限，不是「管理」）",
          forbidden.get("code") == 40301,
          "code=%s msg=%s" % (forbidden.get("code"), forbidden.get("msg")))
    up = body_of("PUT", "/api/v1/tenants/1", ops_token,
                 {"contactName": "张校长（E2E改）"})
    check("修改租户成功", up.get("code") == 0, json.dumps(up, ensure_ascii=False)[:250])
    check("联系人改掉了",
          sql("select contact_name from sys_tenant where id=1;").strip() == "张校长（E2E改）",
          sql("select contact_name from sys_tenant where id=1;").strip())
    check("⭐ tenantCode 没被改动（它出现在域名解析与对接文档里）",
          sql("select tenant_code from sys_tenant where id=1;").strip() == "demo",
          sql("select tenant_code from sys_tenant where id=1;").strip())
    bad = body_of("PUT", "/api/v1/tenants/1", ops_token, {"tenantType": 9})
    check("租户类型越界 → 明确报错", bad.get("code") != 0,
          "code=%s msg=%s" % (bad.get("code"), bad.get("msg")))
    check("不存在的租户 → 明确报错",
          body_of("PUT", "/api/v1/tenants/999999999", ops_token,
                  {"contactName": "x"}).get("code") != 0)

    us = body_of("GET", "/api/v1/tenants/1/usage", ops_token)
    check("用量接口可用", us.get("code") == 0, json.dumps(us, ensure_ascii=False)[:250])
    u = us.get("data") or {}
    check("带学生数与存储", u.get("studentCount") is not None and u.get("storageMb") is not None,
          json.dumps(u, ensure_ascii=False)[:250])
    check("带 quota", isinstance(u.get("quota"), dict) and u["quota"].get("studentLimit"),
          json.dumps(u.get("quota"), ensure_ascii=False))
    check("⭐ ocrCount / smsCount 恒为 0，**并带 notes 说明为什么**"
          "（编一个数字会让运营拿假数据去谈续费）",
          u.get("ocrCount") == 0 and u.get("smsCount") == 0 and len(u.get("notes") or []) >= 2,
          json.dumps({k: u.get(k) for k in ("ocrCount", "smsCount", "notes")},
                     ensure_ascii=False)[:300])
    us2 = body_of("GET", "/api/v1/tenants/1/usage?period=2026Q3", ops_token)
    check("⭐ period 支持季度写法（2026Q3）", us2.get("code") == 0
          and (us2.get("data") or {}).get("period") == "2026Q3",
          json.dumps((us2.get("data") or {}).get("period"), ensure_ascii=False))
    check("period 格式不对 → 明确报错",
          body_of("GET", "/api/v1/tenants/1/usage?period=" + urllib.parse.quote("去年"),
                  ops_token)
          .get("code") != 0)
    b = body_of("GET", "/api/v1/tenants/1/usage", mp(997299))
    check("学生看租户用量 → 403", b.get("code") in (40101, 40301, 403),
          "code=%s" % b.get("code"))

    # ── 2b. 详情脱敏 / 用量口径 / 变更留痕 / 到期日边界 ──
    print("\n[2b] 租户：详情脱敏 / 用量口径 / 变更留痕 / 到期日边界")

    # ① 详情接口必须可用，且联系电话是**脱敏值**。
    #    ⚠️ 修复前这条会 500：`update()` 把手机号**明文**写进了加密列，
    #    而 `AesGcmCipher.decrypt` 对非 Base64 的输入直接抛 —— 库里 id=1 就是那种行。
    det = body_of("GET", "/api/v1/tenants/1", ops_token)
    check("⭐ 租户详情可用（历史明文手机号不能让详情页打不开）",
          det.get("code") == 0, json.dumps(det, ensure_ascii=False)[:250])
    check("详情给出**脱敏**后的联系电话（形如 139****1111）",
          "****" in str((det.get("data") or {}).get("contactPhone")),
          "contactPhone=%r" % ((det.get("data") or {}).get("contactPhone"),))

    # ② 改手机号：落库必须是密文，并且能读回脱敏值（写得进 + 解得开）
    upd = body_of("PUT", "/api/v1/tenants/1", ops_token, {"contactPhone": "13712345678"})
    check("改联系电话成功", upd.get("code") == 0, json.dumps(upd, ensure_ascii=False)[:250])
    stored_phone = sql("select contact_phone from sys_tenant where id=1;").strip()
    check("⭐ 联系电话以**密文**落库（明文落库会让详情页从此永远打不开）",
          stored_phone != "13712345678" and len(stored_phone) > 20, "stored=%r" % stored_phone)
    det2 = body_of("GET", "/api/v1/tenants/1", ops_token)
    check("⭐ 新写的号能读回脱敏值（证明密文与解密对得上，不是「写进去就算了」）",
          (det2.get("data") or {}).get("contactPhone") == "137****5678",
          "contactPhone=%r" % ((det2.get("data") or {}).get("contactPhone"),))

    # ③ 到期日筛选：格式不对要**明确报错**，不能静默忽略
    bad_exp = body_of("GET", "/api/v1/tenants?expireBefore=" + urllib.parse.quote("去年"), ops_token)
    check("⭐ expireBefore 写错 → 明确报错（静默忽略会让人以为「我按到期日筛过了」）",
          bad_exp.get("code") != 0, "code=%s msg=%s" % (bad_exp.get("code"), bad_exp.get("msg")))

    # ④ 后面几条用一个**新建的**租户（不动 demo，免得影响别的用例）
    today = datetime.date.today().isoformat()
    _, created = R.call("POST", BASE + "/api/v1/tenants", ops_token, {
        "tenantName": "E2E用量学校 " + suffix,
        "tenantCode": "e2eusage" + suffix,   # ⚠️ 必须以 e2e 开头，dev-clean-e2e.py 按它清
        "tenantType": 1, "contactName": "E2E用量联系人", "contactPhone": "13900139009",
        "expireTime": today, "adminUsername": "admin",
        "adminRealName": "E2E用量管理员", "adminPhone": "13900139010",
    })
    ntid = ((created or {}).get("data") or {}).get("id")
    check("建一所 E2E 租户（后面的断言靠它，收尾会清掉）", isinstance(ntid, int),
          json.dumps(created, ensure_ascii=False)[:250])

    if isinstance(ntid, int):
        CREATED_TENANT_IDS.append(ntid)

        # 开通就要留痕（F-1 点名「租户开通」是优先留痕点，而本模块原先一条都不写）
        row = sql("select action || '|' || tenant_id from sys_operation_log "
                  "where target_type='TENANT' and target_id=%d order by id desc limit 1;" % ntid).strip()
        check("⭐ 开通租户留痕（target_type=TENANT & target_id=租户 id）",
              row.startswith("CREATE_TENANT|"), "最新一行=%r" % row)
        check("⭐ 留痕的 tenant_id 是 **0**（平台身份），不是被操作的那所学校 —— "
              "所以查「这所学校是谁开通的」要用 target，不能用 tenant_id 筛",
              row.split("|")[-1] == "0", "row=%r" % row)

        # 状态流转：2 试用 / 0 停用 / 1 启用，各自留痕（停用与启用不能并成一个动作）
        for st, act in ((2, "UPDATE_TENANT"), (0, "DISABLE_TENANT"), (1, "ENABLE_TENANT")):
            r = body_of("PUT", "/api/v1/tenants/%d" % ntid, ops_token, {"status": st})
            check("status=%d 能改（⚠️ status=2「试用」原先被拒：只放行 0/1，而前端有「试用」）" % st,
                  r.get("code") == 0, json.dumps(r, ensure_ascii=False)[:200])
            got = sql("select action from sys_operation_log where target_type='TENANT' "
                      "and target_id=%d order by id desc limit 1;" % ntid).strip()
            check("status→%d 的留痕动作是 %s" % (st, act), got == act, "实际=%r" % got)

        r4 = body_of("PUT", "/api/v1/tenants/%d" % ntid, ops_token, {"tenantType": 4})
        check("⭐ tenantType=4（企业）能改（原先只放行 1~3，而前端下拉有 4 项）",
              r4.get("code") == 0, json.dumps(r4, ensure_ascii=False)[:200])

        # 到期日**含当天**：把到期日设成今天，`expireBefore=今天` 必须能查到它
        body_of("PUT", "/api/v1/tenants/%d" % ntid, ops_token, {"expireTime": today})
        lst = body_of("GET", "/api/v1/tenants?expireBefore=%s&pageSize=200" % today, ops_token)
        hit = [t.get("id") for t in ((lst.get("data") or {}).get("list") or [])]
        check("⭐ expireBefore=今天 能查到「今天到期」的租户"
              "（原先用 atStartOfDay() 比，差一天，查不到）",
              ntid in hit, "今天=%s 命中 id=%s" % (today, hit[:10]))

        # ⭐ 用量口径 —— 本段最重要的一条。
        #    平台身份会让租户拦截器**整张表关掉过滤**，所以 `/tenants/{id}/usage`
        #    曾经返回的是**全平台**的学生数与存储量。库里只有一所学校有数据时
        #    它与正确值完全相等，所以这里必须先往**另一个租户**塞一行名册。
        R.sql("insert into collect_roster (tenant_id, task_id, real_name, id_card_no, id_card_hash) "
              "values (%d, -1, 'E2E用量范围', 'x', '%s%s');"
              % (ntid, USAGE_FIXTURE_HASH_PREFIX, suffix))
        own = int(sql("select count(distinct id_card_hash) from collect_roster "
                      "where tenant_id=1 and deleted=0;").strip() or 0)
        allc = int(sql("select count(distinct id_card_hash) from collect_roster "
                       "where deleted=0;").strip() or 0)
        u1 = (body_of("GET", "/api/v1/tenants/1/usage", ops_token).get("data") or {})
        check("⭐ 平台看到的「1 号学校学生数」= 库里 tenant_id=1 的去重人数，**不是全平台**",
              u1.get("studentCount") == own,
              "接口=%s　库里 tenant_id=1 去重=%s　全平台去重=%s（接口等于后者就说明过滤被算丢了）"
              % (u1.get("studentCount"), own, allc))
        un = (body_of("GET", "/api/v1/tenants/%d/usage" % ntid, ops_token).get("data") or {})
        check("⭐ 新租户看到的是**它自己**那一行（顺带证明不是恒 0）",
              un.get("studentCount") == 1, "新租户 studentCount=%s" % un.get("studentCount"))
        check("全平台确实不止 1 个（否则上面那条断言是空转的）", allc > own,
              "全平台=%s 一号学校=%s" % (allc, own))

    # ── 3. 文件夹具 ──
    print("\n[3] 摆文件夹具（一个任务 + 一份草稿 + 一个材料文件）")
    _, b = R.call("POST", BASE + "/api/v1/tasks", at, {
        "taskName": "E2E文件 " + suffix, "batchNo": "202609", "majorName": "数控加工",
        "collectMode": 1, "certType": 1, "startTime": "2026-01-01", "endTime": "2026-12-31"
    })
    task_id = b["data"]["id"]
    CREATED_TASK_IDS.append(task_id)
    _, b = R.call("GET", BASE + "/api/v1/orgs/tree", at)
    classes = []

    def walk(nodes):
        for n in nodes or []:
            if n.get("orgType") == 3:
                classes.append(n)
            walk(n.get("children"))

    walk(b.get("data") or [])
    org = classes[0]
    R.call("PUT", BASE + "/api/v1/tasks/%d/orgs" % task_id, at,
           {"orgs": [{"orgId": org["id"], "expectedCount": 1}]})
    R.call("POST", BASE + "/api/v1/tasks/%d/materials/apply-template" % task_id, at,
           {"templateCode": "TPL_SCHOOL_ROLL"})
    _build_roster([("文件甲", R.with_checksum("36102920080808091"))])
    body = R.upload_roster(BASE, at, task_id, TMP_ROSTER,
                           extra={"dryRun": "false", "mode": "APPEND", "orgId": org["id"]})
    check("夹具：名册就绪", body.get("code") == 0, json.dumps(body, ensure_ascii=False)[:180])
    R.call("POST", BASE + "/api/v1/tasks/%d/publish" % task_id, at, {"force": True})

    os.system('docker exec xt-redis redis-cli -a redis_dev_pwd --no-auth-warning '
              'DEL xt:mp:gate:user:%d > NUL 2>&1' % STU)
    tok = mp(STU)
    _, b = R.call("POST", BASE + "/api/v1/mp/roster/match", tok,
                  {"taskId": task_id, "realName": "文件甲",
                   "idCardNo": R.with_checksum("36102920080808091")})
    if b.get("code") != 0:
        print("  x 门禁核验失败：" + json.dumps(b, ensure_ascii=False)[:200])
        return 2
    _, b = R.call("POST", BASE + "/api/v1/mp/submissions", tok, {"taskId": task_id})
    sid = b["data"]["id"]
    check("夹具：学生建了草稿", b.get("code") == 0, json.dumps(b, ensure_ascii=False)[:150])

    _make_jpeg()
    up_res = upload("/api/v1/mp/files/upload", tok,
                    "submissionId=%d" % sid, "materialCode=IDCARD",
                    "file=@%s;filename=乱名字.jpg;type=image/jpeg" % TINY_JPEG)
    check("夹具：上传一份材料", up_res.get("code") == 0,
          json.dumps(up_res, ensure_ascii=False)[:250])
    file_id = (up_res.get("data") or {}).get("id") or (up_res.get("data") or {}).get("fileId")
    check("夹具：拿到 fileId", isinstance(file_id, int), str(file_id))

    # ── 4. §9.8 预览 ──
    print("\n[4] §9.8 预览（签名 URL + 元信息）")
    v = body_of("GET", "/api/v1/mp/files/%d" % file_id, tok)
    check("预览可用", v.get("code") == 0, json.dumps(v, ensure_ascii=False)[:250])
    vd = v.get("data") or {}
    check("带 previewUrl / standardName / 尺寸",
          vd.get("previewUrl") and vd.get("standardName") and vd.get("width"),
          json.dumps(vd, ensure_ascii=False)[:300])
    check("带 expireSeconds（签名 URL 有有效期，不能长期保存）",
          vd.get("expireSeconds") and vd.get("expireSeconds") > 0,
          str(vd.get("expireSeconds")))
    check("⭐ 带 actualKind（请求 original 时原始件可能不存在，会退到规范件）",
          vd.get("actualKind") in ("ORIGINAL", "STANDARD"), str(vd.get("actualKind")))
    vo = body_of("GET", "/api/v1/mp/files/%d?prefer=original" % file_id, tok)
    check("prefer=original 也能用（原始件默认不保留时会退到规范件并标注）",
          vo.get("code") == 0 and (vo.get("data") or {}).get("actualKind") in
          ("ORIGINAL", "STANDARD"),
          json.dumps(vo.get("data"), ensure_ascii=False)[:250])
    check("⭐ **别人的文件** → 拒绝（不区分「不存在」与「不是你的」，避免枚举）",
          body_of("GET", "/api/v1/mp/files/%d" % file_id, mp(OTHER)).get("code") != 0,
          "code=%s" % body_of("GET", "/api/v1/mp/files/%d" % file_id, mp(OTHER)).get("code"))
    check("不存在的文件 → 明确报错",
          body_of("GET", "/api/v1/mp/files/999999999", tok).get("code") != 0)
    check("未登录 → 40101", body_of("GET", "/api/v1/mp/files/%d" % file_id).get("code") == 40101)

    # ── 5. §9.5 上传凭证 ──
    print("\n[5] §9.5 上传凭证（预检 + 秒传 + 压缩提示）")
    tk = body_of("POST", "/api/v1/mp/files/upload-token", tok, {
        "taskId": task_id, "materialCode": "PHOTO1", "fileName": "相册照片.jpg",
        "fileSize": 300000, "contentType": "image/jpeg"
    })
    check("凭证接口可用", tk.get("code") == 0, json.dumps(tk, ensure_ascii=False)[:250])
    td = tk.get("data") or {}
    check("ok=true 且没命中秒传", td.get("ok") is True and td.get("existed") is False,
          json.dumps(td, ensure_ascii=False)[:250])
    check("⭐ 带压缩提示（让学生在自己手机上压，省的是他的流量与等待）",
          (td.get("clientCompressHint") or {}).get("enabled") is True,
          json.dumps(td.get("clientCompressHint"), ensure_ascii=False))
    check("带 objectKey / uploadToken / expireSeconds=900",
          td.get("objectKey") and td.get("uploadToken") and td.get("expireSeconds") == 900,
          json.dumps({k: td.get(k) for k in ("objectKey", "expireSeconds")}, ensure_ascii=False))
    check("⭐ 一期 direct=false（七牛直传要先把「上传完成回调」打通，见 F-71）",
          td.get("direct") is False and len(td.get("notes") or []) >= 1,
          json.dumps({k: td.get(k) for k in ("direct", "notes")}, ensure_ascii=False)[:250])
    # 体积超限：**不传字节就能拦**
    big = body_of("POST", "/api/v1/mp/files/upload-token", tok, {
        "taskId": task_id, "materialCode": "PHOTO1", "fileName": "巨图.jpg",
        "fileSize": 8 * 1024 * 1024, "contentType": "image/jpeg"
    })
    bd = big.get("data") or {}
    check("⭐ 体积超限 → ok=false + **能照着做的原因**（说清上限与当前值）",
          bd.get("ok") is False and "KB" in (bd.get("rejectReason") or ""),
          json.dumps(bd, ensure_ascii=False)[:250])
    check("超限时压缩提示里 needed=true",
          (bd.get("clientCompressHint") or {}).get("needed") is True,
          json.dumps(bd.get("clientCompressHint"), ensure_ascii=False))
    # 份数超限（IDCARD 已经传了 1 份，模板 maxCount=2 → 再传 1 份可以，第 3 份不行）
    for i in range(1):
        upload("/api/v1/mp/files/upload", tok, "submissionId=%d" % sid,
               "materialCode=IDCARD",
               "file=@%s;filename=第二次.jpg;type=image/jpeg" % TINY_JPEG)
    full = body_of("POST", "/api/v1/mp/files/upload-token", tok, {
        "taskId": task_id, "materialCode": "IDCARD", "fileName": "第三张.jpg",
        "fileSize": 100000, "contentType": "image/jpeg"
    })
    fd = full.get("data") or {}
    check("⭐ 份数超限 → ok=false 并说清「最多几份、你已经传了几份、要换请先删」",
          fd.get("ok") is False and "最多" in (fd.get("rejectReason") or ""),
          json.dumps(fd, ensure_ascii=False)[:250])
    # 秒传：拿已经传过那份的 fileHash
    real_hash = sql("select file_hash from collect_file where id=%d;" % file_id).strip()
    fast = body_of("POST", "/api/v1/mp/files/upload-token", tok, {
        "taskId": task_id, "materialCode": "IDCARD", "fileName": "同一张.jpg",
        "fileSize": 100, "contentType": "image/jpeg", "fileHash": real_hash
    })
    fd2 = fast.get("data") or {}
    check("⭐ 秒传命中（同一个人在同一个槽位传同样的文件）→ existed=true",
          fd2.get("existed") is True and fd2.get("fileId") == file_id,
          json.dumps(fd2, ensure_ascii=False)[:250])
    check("⭐ 秒传时**不给 uploadUrl**（返回了而前端不用，会让「到底传没传」变成两处判断）",
          fd2.get("uploadUrl") is None, str(fd2.get("uploadUrl")))
    bad = body_of("POST", "/api/v1/mp/files/upload-token", tok, {
        "taskId": task_id, "materialCode": "NO_SUCH_MATERIAL", "fileName": "x.jpg",
        "fileSize": 100, "contentType": "image/jpeg"
    })
    check("材料码不存在 → 明确报错", bad.get("code") != 0,
          "code=%s msg=%s" % (bad.get("code"), bad.get("msg")))

    # ── 6. §9.7 批量处理 ──
    print("\n[6] §9.7 批量处理（多材料一次提交）")
    bp = body_of("POST", "/api/v1/mp/files/batch-process", tok, {
        "submissionId": sid, "fileIds": [file_id], "async": True
    })
    check("批量处理可用", bp.get("code") == 0, json.dumps(bp, ensure_ascii=False)[:250])
    bpr = bp.get("data") or {}
    check("带 batchToken / total / results / summary",
          bpr.get("batchToken") and bpr.get("total") == 1 and bpr.get("summary"),
          json.dumps(bpr, ensure_ascii=False)[:250])
    item = (bpr.get("results") or [{}])[0]
    check("结果项带 fileId 与 status", item.get("fileId") == file_id
          and item.get("status") is not None, json.dumps(item, ensure_ascii=False))
    check("⭐ 一期 conflict 恒为 false、detectMaterialCode 恒为空，**并说明为什么**"
          "（不是「识别后没问题」，而是「还没识别」）",
          item.get("conflict") is False and item.get("detectMaterialCode") is None
          and len(bpr.get("notes") or []) >= 1,
          json.dumps({k: item.get(k) for k in ("conflict", "detectMaterialCode")},
                     ensure_ascii=False) + " notes=" + str(bpr.get("notes"))[:120])
    over = body_of("POST", "/api/v1/mp/files/batch-process", tok, {
        "submissionId": sid, "fileIds": list(range(1, 25))
    })
    check("⭐ 超过 20 个 → 40011（与批量审批同一个上限口径）", over.get("code") == 40011,
          "code=%s msg=%s" % (over.get("code"), over.get("msg")))
    foreign = body_of("POST", "/api/v1/mp/files/batch-process", tok, {
        "submissionId": sid, "fileIds": [999999999]
    })
    check("不属于本填报单的文件 → 记成失败项而不是整批失败",
          (foreign.get("data") or {}).get("summary", {}).get("failed") == 1,
          json.dumps(foreign.get("data"), ensure_ascii=False)[:250])
    noids = body_of("POST", "/api/v1/mp/files/batch-process", tok, {"submissionId": sid})
    check("不传 fileIds → 明确报错", noids.get("code") != 0,
          "code=%s msg=%s" % (noids.get("code"), noids.get("msg")))

    cleanup()
    print("\n" + "=" * 74)
    print("通过 %d　失败 %d　共 %d" % (len(PASSED), len(FAILED), len(PASSED) + len(FAILED)))
    print("=" * 74)
    for n, det in FAILED:
        print("  x " + n + "\n      " + det)
    return 1 if FAILED else 0


def _build_org_xlsx(rows, header=None):
    from openpyxl import Workbook
    wb = Workbook()
    ws = wb.active
    ws.title = "组织"
    head = header or ["层级", "组织名称", "组织编码", "上级编码", "负责人手机号"]
    for c, t in enumerate(head, start=1):
        ws.cell(row=1, column=c, value=t)
    for i, r in enumerate(rows):
        for c, v in enumerate(r, start=1):
            ws.cell(row=i + 2, column=c, value=v)
    wb.save(TMP_ORG)


def _build_roster(rows):
    from openpyxl import Workbook
    wb = Workbook()
    ws = wb.active
    ws.title = "文件班"
    header = ["序号", "姓名", "身份证号", "学历", "联系电话", "工作单位",
              "报考职业", "报考工种", "报考等级", "是否补考",
              "补考理论", "补考实操", "补考综合", "是否在校生",
              "在读学校", "所学专业", "入学时间"]
    for c, t in enumerate(header, start=1):
        ws.cell(row=1, column=c, value=t)
    for i, (name, idc) in enumerate(rows):
        vals = [i + 1, name, idc, "中专", "1380013880%d" % i, "无",
                "车工", "数控车工", "四级", "否", "", "", "", "是",
                "示范职业学校", "数控加工", "2024-09-01"]
        for c, v in enumerate(vals, start=1):
            if v != "":
                ws.cell(row=i + 2, column=c, value=v)
    wb.save(TMP_ROSTER)


def _make_jpeg():
    import base64
    b64 = ("/9j/4AAQSkZJRgABAQEAYABgAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0a"
           "HBwgJC4nICIsIxwcKDcpLDAxNDQ0Hyc5PTgyPC4zNDL/wAALCAABAAEBAREA/8QAFAABAAAAAAAA"
           "AAAAAAAAAAAACf/EABQQAQAAAAAAAAAAAAAAAAAAAAD/2gAIAQEAAD8AKp//2Q==")
    with open(TINY_JPEG, "wb") as f:
        f.write(base64.b64decode(b64))


_TENANT_TABLES = None


def _tables_with_tenant_id():
    """
    带 `tenant_id` 的表 —— **动态枚举**，不手抄表名。

    手抄的清单在"新增了一张业务表"那天就开始漏，而漏掉的那张表恰好很难被注意到
    （租户没了、行还在）。`dev-clean-e2e.py` 是同一口径。
    """
    global _TENANT_TABLES
    if _TENANT_TABLES is None:
        out = sql("SELECT table_name FROM information_schema.columns "
                  "WHERE table_schema='public' AND column_name='tenant_id'")
        _TENANT_TABLES = [t.strip() for t in out.splitlines() if t.strip()]
    return _TENANT_TABLES


def cleanup():
    for tid in CREATED_TASK_IDS:
        R.sql("delete from collect_validation_issue where task_id=%d;" % tid)
        R.sql("delete from collect_audit_record where task_id=%d;" % tid)
        R.sql("delete from collect_notify_record where task_id=%d;" % tid)
        R.sql("delete from collect_file where task_id=%d;" % tid)
        R.sql("delete from collect_submission where task_id=%d;" % tid)
        R.sql("delete from collect_roster where task_id=%d;" % tid)
        R.sql("delete from collect_task_org where task_id=%d;" % tid)
        R.sql("delete from collect_material_def where task_id=%d;" % tid)
        R.sql("delete from collect_task where id=%d;" % tid)
    for name in CREATED_ORG_NAMES:
        R.sql("delete from sys_org where org_name = '%s';" % name)

    # 用量夹具（本脚本往 E2E 租户里插的那一行名册）
    R.sql("delete from collect_roster where id_card_hash like '%s%%';" % USAGE_FIXTURE_HASH_PREFIX)

    # 本脚本建的 E2E 租户：连带它的用户 / 角色 / 组织一起删。
    # ⚠️ 日志行要单独删 —— 它们的 `tenant_id` 是 **0**（平台身份），
    #    用 tenant_id 清是清不掉的，得按 target 清。
    for tid in CREATED_TENANT_IDS:
        R.sql("delete from sys_operation_log where target_type='TENANT' and target_id=%d;" % tid)
        for table in _tables_with_tenant_id():
            R.sql("delete from %s where tenant_id=%d;" % (table, tid))
        R.sql("delete from sys_tenant where id=%d;" % tid)
    del CREATED_TENANT_IDS[:]

    # 联系人 / 联系电话复原：本条用例改过它们（手机号改完是**密文**，
    # 所以复原成改动前那个明文值 —— 顺带保留"库里有一行历史明文"这个现状，
    # 那正是详情接口的容错分支要处理的输入）
    R.sql("update sys_tenant set contact_name='张校长', contact_phone='13800138000' where id=1;")
    R.sql("delete from collect_message where user_id in (997201, 997202, 997299);")
    for p in (TMP_ROSTER, TMP_ORG, TINY_JPEG):
        if os.path.exists(p):
            os.remove(p)
    print("  已清理：任务 / 材料文件 / 测试组织 / E2E 租户与用量夹具 / "
          "租户联系人复原 / 临时文件")


if __name__ == "__main__":
    raise SystemExit(main())