diff --git a/.scratch/acceptance-pool-count-status/work.md b/.scratch/acceptance-pool-count-status/work.md new file mode 100644 index 00000000..473aedea --- /dev/null +++ b/.scratch/acceptance-pool-count-status/work.md @@ -0,0 +1,50 @@ +# #20 验收任务池数量和错误提示(position 593) + +Status: resolved + +授权:本地截图读取、诊断、实现、验证和提交;不推送、不部署。基线 dev / 11cb0929,统一 recent-integration;新鲜度确认包含 origin/dev 和线上版本。共享树原有 AGENTS/docs/管理员池测试及其他任务提交保留。 + +## 截图和实际问题 + +已通过 lark-im 的现有读取权限定位平台调优群 position 593,消息 om_x100b609ee360a8a4c33efa82e7ce19e。截图保存到忽略目录 local-data/diagnostics/issue20/customer-status.png;仅保留该目标消息的上下文。 + +| 红框 | 真实位置和来源 | 判定 | +| --- | --- | --- | +| 上框 | AnnotationOperations 的任务池页签:可领取、我的任务、我已验收;数据来自 GET /api/human-item-counts | 当前筛选 scope 没有成功取得 reviewPoolTotals 时渲染“—”;不是动作判断状态 | +| 下框 | 同一侧栏的 reviewPoolCountError,位于逐段任务列表上方 | 计数请求失败后,经通用错误文案显示“系统暂时无法完成操作”;不是验收决定错误 | + +箭头指向的“全部状态”是已有筛选控件,本截图没有展开菜单,不推断其选项本身有错。截图中的任务卡和验收播放器仍正常,是因为任务列表、详情、决定和计数是独立请求。 + +## 因果证据与历史 + +1. 线上只读核验 ready 的 buildCommit 为 1a6e61909efd2e7d4afcf5d42d2653d7eef28c47。使用既有验收测试账号读取与页面一致的全项目/acceptance+prepackage_acceptance/category=all 计数路由,HTTP 500,响应明确为 `(MaxTimeMSExpired) PlanExecutor error during aggregation :: caused by :: operation exceeded time limit`。证据 live-read.json。API 日志也存在重复计数 500;日志为范围采样,不能把各用户/各路由的采样比例当总体故障率。 +2. human_pool_count_scan.go 在该线上版本与任务基线的 SHA-256 完全相同(afbe3660a4bae652c434b79098d0b0e35cc070637d92a1fdb3c77e8ce237539c)。1dc97107 没有改 AnnotationOperations 或该后端文件,且从未推送/部署;其共用动作左栏在工作台内部,本截图红框位于外层任务池,数据源不同。因此本次客户问题是已有计数超时,不是 592 移植引入。 +3. 原 covered count 把 limit 放在 facet 的匹配分支内,facet 仍要消费剩余组织候选。匹配结果超过一页时,每页都重扫剩余候选,放大聚合工作量。前端失败后不产生有效数量,错误没有局部恢复入口;它会保留至另一次相关加载,不是把质检状态读进验收缓存。 +4. 先建立红反馈:2017 候选原实现扫描 3051 个索引键,超线性预算;缺少局部重试的真实页面用例也变红。后续用 Go overlay 将最终回归直接跑在任务基线后端,12017 候选原实现扫描 78221 个键而失败;当前实现同样候选扫描 12017 个键而通过。overlay 不改 checkout,不创建分支,不改写历史。日志 count-red.log / count-baseline-12k-red.log / mongo-final.log。 + +## 修复与兼容范围 + +- 后端:先限制每个已排序的索引候选页为 1000 条,再进入保留 null/缺失语义的投影/facet。facet 同时返回匹配行和本页最后已扫描 ID/扫描数;按扫描页推进,不能按匹配页推进或因某页无匹配提前结束。继续执行原 API 权限过滤、失败整条请求不返回部分总数、取消和已有最大时限;不增加超时,不增加生产索引或迁移。 +- 前端:数量加载、失败和真实零明确区分;失败显示“任务数量暂时无法读取”及局部“重试”,仍保留未知“—”,不拿当前分页行数补总数。重试只重读数量,不切换任务池、任务或编辑状态,不改审核决定权限。成功清除旧错误;离开计数工作台清除其错误/加载状态;旧筛选响应继续受 AbortController 保护。 +- 返回值完整性按本角色实际页签校验:验收/质检只有 available/mine/reviewed,管理员才要求 all。初版新增校验错误地将 all 视为所有角色必填,被真实 API 浏览器回归抓到;已修正且将新夹具改为真实三字段响应,未提交/上线该中间版本。 +- API、Mongo 记录、切片、动作/意见、验收冻结证据、最终验收供应商匿名及正式决定/记录契约不变;无客户业务数据/配置/索引写入。回退不需要数据迁移,回到旧版会失去本次查询限界和数量恢复体验。 + +## 验证和限制 + +执行 verify:plan;使用既有 scripts/verify-change.py 的 quick/task 执行器,对本任务文件选择任务池计数和审查决定范围,排除他人管理员池测试及未改变的标注准入/编辑流程。每份复用证据记录相关源文件、测试和依赖 SHA-256;未变后端/隔离 Mongo/三播放器证据在本任务内复用,前端角色校验修正后重新执行数量恢复、真实搜索、草稿、决定、权限和构建。最终任务报告 selected-checks-passed,inputsChangedDuringRun=false: + +- local-data/test-results/verify-1794814631078534000/report.json;主日志 local-data/test-results/issue20/task-verified.log。 +- 类型/架构检查、141 项定向前端测试、22 项验证工具测试、Go 格式和 diff 检查通过;最终前端构建通过(既有 chunk 大小提示保留)。 +- 实际页面、本地纯合成 API:验收和质检计数超时、继续处理已有验收任务不受阻、真实三字段计数恢复、清除错误、筛选慢响应/快响应不串用、不完整响应保持未知;无业务写请求,pageErrors 为空。日志 ui-green.log,截图 browser/{acceptance,quality}-count-{failed,recovered}.png,人工查看验收两种状态。 +- 隔离本地 Mongo:12017 候选、末页仅3个匹配、前面整页无匹配,计数正确并保持线性扫描;2108 条含 null/缺失/继承返工等旧形状的等价回归通过,documents fetched=0。日志 mongo-final.log。数据库名称为本次唯一合成库,测试后清理,不碰共享快照中的任务认领状态。 +- 实际 API + 隔离 Mongo:管理员和验收角色各 12000 条计数与分页基线一致,包含正常/搜索/无匹配;取消、失败无部分总数、并发身份隔离、供应商授权撤销、QC项目范围和分页/搜索等价通过。相关日志在任务报告及 verify-1794814291784921000/09.log;并发身份隔离额外定向证据为 count-concurrent.log。 +- 真实任务池搜索/翻页/历史回查/过时响应保护通过(search-browser-final.log);草稿持久化、决定失败恢复及只读/权限保护通过。 +- native / movi / movi-fractional 实际验收整页各15次切片变化、只读/决定禁用、批注保存、刷新/恢复、短动作和双时间等通过。证据 players/report.json,执行日志在 verify-1794814291784921000/16.log。播放器实现本次未改。 + +完整 Go 门禁不能宣称通过:扩大检查触及的 TestPoolPerformanceMongoBacklog 调用另一条标注随机领取计数,返回0而预期12000;使用 npm run test:go:gate 对任务基线 overlay 和当前实现分别运行,均报同一确切用例失败。两次门禁还报告已有 local-data/statistics-baseline.go 等遗留脚本编译错误。没有改 known-failures.json、没有删除/重置他人文件,没有把这些失败算进成功;门禁日志 backlog-baseline.log / backlog-current.log 及同名日志目录。最初任务级报告 verify-1794814291784921000 因角色四字段校验失败也保留,最终修正后通过,不覆盖为成功。 + +扫描量、耗时和响应字节仅为本地合成规模证据,不证明生产并发或当前线上超时已消失。本次故障探测时线上为 1a6e6190;本任务未部署,不能据本地通过宣称客户故障已经恢复。需另行完成授权发布及业务验收。 + +本地提交仅包含本任务六个文件;使用 lint-staged --no-stash 和等价检查避免临时 stash 共享工作。提交后 agent_status --no-fetch 汇报分支状态。#18/#19/592 的既有提交及其他任务提交均保留,本修复追加到待推送 dev 批次。 + +收尾状态核对:git ls-remote 读取 origin/dev 为 11cb09296b398770a630c45448dfd9bf667cc0b7,已包含此前 #18/#19/592 和其他任务提交;本任务没有执行 push/fetch 改写远端操作。#20 作为单独新增本地提交保留,未推送;不把此前提交仍称为全部未推送。 diff --git a/internal/store/human_pool_count_bounded_test.go b/internal/store/human_pool_count_bounded_test.go new file mode 100644 index 00000000..20308eeb --- /dev/null +++ b/internal/store/human_pool_count_bounded_test.go @@ -0,0 +1,139 @@ +package store + +import ( + "context" + "fmt" + "os" + "sync" + "testing" + "time" + + "droboticflow/internal/domain" + "go.mongodb.org/mongo-driver/bson" + "go.mongodb.org/mongo-driver/event" + "go.mongodb.org/mongo-driver/mongo" + "go.mongodb.org/mongo-driver/mongo/options" +) + +// A covered count must not scan the remaining organization again for every +// result page. This is an isolated local database, never a customer database. +func TestHumanPoolCountsBoundedMongoScan(t *testing.T) { + if os.Getenv("POOL_COUNT_LOCAL_MONGO") != "1" { + t.Skip("isolated local Mongo opt-in") + } + ctx, cancel := context.WithTimeout(context.Background(), time.Minute) + defer cancel() + var mu sync.Mutex + var commands []bson.M + client, err := mongo.Connect(ctx, options.Client().ApplyURI("mongodb://127.0.0.1:27018/?directConnection=true").SetServerSelectionTimeout(3*time.Second).SetMonitor(&event.CommandMonitor{Started: func(_ context.Context, e *event.CommandStartedEvent) { + if e.CommandName != "aggregate" { + return + } + var cmd bson.M + if bson.Unmarshal(e.Command, &cmd) != nil { + return + } + if cmd["aggregate"] != "humanItems" { + return + } + mu.Lock() + commands = append(commands, cmd) + mu.Unlock() + }})) + if err != nil { + t.Fatal(err) + } + db := client.Database(fmt.Sprintf("pool_count_bounded_%d", time.Now().UnixNano())) + t.Cleanup(func() { + work, stop := context.WithTimeout(context.Background(), 10*time.Second) + defer stop() + if err := db.Drop(work); err != nil { + t.Error(err) + } + client.Disconnect(work) + }) + ds := NewMongoStore(db) + if err := ds.EnsureIndexes(ctx); err != nil { + t.Fatal(err) + } + if _, err := db.Collection("humanItems").Indexes().CreateMany(ctx, PlatformReadIndexes()["humanItems"]); err != nil { + t.Fatal(err) + } + const n = 12017 + rows := make([]any, 0, n) + for i := 0; i < n; i++ { + actor := "other" + if i >= n-3 { + actor = "needle" + } + rows = append(rows, bson.M{"id": fmt.Sprintf("item-%06d", i), "batchId": "batch", "subject": bson.M{"itemKey": fmt.Sprintf("source-%d", i)}, "organizationId": "org", "pluginId": "annotation-qa", "projectId": "p", "currentStageId": "acceptance", "status": "claimed", "assignedUserId": actor}) + } + if _, err := db.Collection("humanItems").InsertMany(ctx, rows); err != nil { + t.Fatal(err) + } + for _, tc := range []struct { + name, actor string + want int + }{{"all", "", n}, {"sparse-last-page", "needle", 3}} { + t.Run(tc.name, func(t *testing.T) { + mu.Lock() + commands = nil + mu.Unlock() + count := 0 + err := ds.VisitHumanPoolCountCandidates(ctx, "org", domain.HumanItemQuery{PluginID: "annotation-qa", StageIDs: []string{"acceptance"}, ActorUserID: tc.actor, MinePoolOnly: true}, func(items []domain.HumanItemRecord) error { count += len(items); return nil }) + if err != nil { + t.Fatal(err) + } + if count != tc.want { + t.Fatalf("count=%d want=%d", count, tc.want) + } + mu.Lock() + captured := append([]bson.M(nil), commands...) + mu.Unlock() + if len(captured) == 0 { + t.Fatal("covered count was not exercised") + } + var examined int64 + for _, cmd := range captured { + clean := bson.D{} + for _, key := range []string{"aggregate", "pipeline", "cursor", "hint", "maxTimeMS"} { + if value, ok := cmd[key]; ok { + clean = append(clean, bson.E{Key: key, Value: value}) + } + } + var result bson.M + if err := db.RunCommand(ctx, bson.D{{Key: "explain", Value: clean}, {Key: "verbosity", Value: "executionStats"}}).Decode(&result); err != nil { + t.Fatal(err) + } + var walk func(any) + walk = func(v any) { + switch x := v.(type) { + case bson.M: + if stats, ok := x["executionStats"].(bson.M); ok { + switch keys := stats["totalKeysExamined"].(type) { + case int32: + examined += int64(keys) + case int64: + examined += keys + } + } + for key, value := range x { + if key != "executionStats" { + walk(value) + } + } + case bson.A: + for _, value := range x { + walk(value) + } + } + } + walk(result) + } + t.Logf("count=%d aggregate_calls=%d examined_keys=%d", count, len(captured), examined) + if examined == 0 || examined > n+n/10 { + t.Fatalf("count rescanned the organization: keys=%d, linear budget=%d", examined, n+n/10) + } + }) + } +} diff --git a/internal/store/human_pool_count_scan.go b/internal/store/human_pool_count_scan.go index 20bfa0db..6e2b1a56 100644 --- a/internal/store/human_pool_count_scan.go +++ b/internal/store/human_pool_count_scan.go @@ -99,13 +99,17 @@ func (s *MongoStore) visitCoveredHumanPoolCounts(ctx context.Context, org string if lastID != "" { anchor["id"] = bson.M{"$gt": lastID} } + // $facet consumes its whole input even when its rows branch has a + // limit. Bound the scanned index page before it, not only the matches. pipeline := mongo.Pipeline{ {{Key: "$match", Value: anchor}}, {{Key: "$sort", Value: bson.D{{Key: "id", Value: 1}}}}, + {{Key: "$limit", Value: humanPoolCountBatchSize}}, {{Key: "$project", Value: projection}}, - {{Key: "$facet", Value: bson.M{"rows": bson.A{bson.M{"$match": filter}, bson.M{"$limit": humanPoolCountBatchSize}}}}}, - {{Key: "$unwind", Value: "$rows"}}, - {{Key: "$replaceRoot", Value: bson.M{"newRoot": "$rows"}}}, + {{Key: "$facet", Value: bson.M{ + "rows": bson.A{bson.M{"$match": filter}}, + "scan": bson.A{bson.M{"$group": bson.M{"_id": nil, "lastID": bson.M{"$last": "$id"}, "count": bson.M{"$sum": 1}}}}, + }}}, } cur, err := s.humanItems.Aggregate(ctx, pipeline, options.Aggregate().SetHint(humanPoolCountIndex).SetMaxTime(8*time.Second).SetBatchSize(humanPoolCountBatchSize)) if err != nil { @@ -114,22 +118,33 @@ func (s *MongoStore) visitCoveredHumanPoolCounts(ctx context.Context, org string } return true, err } - var rows []domain.HumanItemRecord - err = cur.All(ctx, &rows) + var pages []struct { + Rows []domain.HumanItemRecord `bson:"rows"` + Scan []struct { + LastID string `bson:"lastID"` + Count int `bson:"count"` + } `bson:"scan"` + } + err = cur.All(ctx, &pages) if err != nil { return true, err } - if len(rows) == 0 { + if len(pages) == 0 || len(pages[0].Scan) == 0 { return true, nil } - next := rows[len(rows)-1].ID + page := pages[0] + next := page.Scan[0].LastID if next == "" || next <= lastID { return true, fmt.Errorf("pool count cursor did not advance") } - if err := visit(rows); err != nil { - return true, err + if len(page.Rows) > 0 { + if err := visit(page.Rows); err != nil { + return true, err + } } - if len(rows) < humanPoolCountBatchSize { + // Sparse matches must not terminate the scan or reuse the last + // matching ID: entire index pages may have no visible candidates. + if page.Scan[0].Count < humanPoolCountBatchSize { return true, nil } lastID = next diff --git a/scripts/verification-scopes.json b/scripts/verification-scopes.json index b3d79017..d85fdec8 100644 --- a/scripts/verification-scopes.json +++ b/scripts/verification-scopes.json @@ -340,7 +340,7 @@ "^internal/(api|store)/(human_pool|admin_annotation_pool|human_loop)", "^internal/(api/(human_display|human_list_reads)|store/human_run_display)", "^docs/operations/acceptance-search-indexes-v2\\.json$", - "^tests/browser/(review-pool-search|admin-annotation-pool)", + "^tests/browser/(review-pool-search|review-pool-count-status|admin-annotation-pool)", "^src/lib/review-claim-candidates\\.js$" ], "frontend": [ @@ -360,6 +360,24 @@ "./internal/store": ["^TestHumanPoolCountCandidateScan", "^TestAdminSourceExceptionsMemory"] }, "task": [ + { + "argv": ["node", "tests/browser/review-pool-count-status.mjs"], + "tests": true + }, + { + "argv": [ + "go", + "test", + "./internal/store", + "-run", + "^TestHumanPoolCountsBoundedMongoScan$|^TestPlatformPoolCountCoveredScanParityAndBounds$", + "-count=1", + "-timeout=3m", + "-v" + ], + "env": { "POOL_COUNT_LOCAL_MONGO": "1", "POOL_LOCAL_MONGO_URI": "mongodb://127.0.0.1:27018" }, + "tests": true + }, { "argv": [ "go", diff --git a/src/pages/AnnotationOperations.tsx b/src/pages/AnnotationOperations.tsx index 1579557c..78b82ab8 100644 --- a/src/pages/AnnotationOperations.tsx +++ b/src/pages/AnnotationOperations.tsx @@ -757,6 +757,8 @@ export default function AnnotationOperations({ workPage = false }: { workPage?: counts: Record; } | null>(null); const [reviewPoolCountError, setReviewPoolCountError] = useState(""); + const [reviewPoolCountRevision, setReviewPoolCountRevision] = useState(0); + const [reviewPoolCountLoading, setReviewPoolCountLoading] = useState(false); const [acceptancePoolDetail, setAcceptancePoolDetail] = useState<{ scope: string; item: HumanItem } | null>(null); const [acceptancePoolDetailError, setAcceptancePoolDetailError] = useState(""); const [acceptanceDetailRetry, setAcceptanceDetailRetry] = useState(0); @@ -1077,9 +1079,14 @@ export default function AnnotationOperations({ workPage = false }: { workPage?: ...(pagedReviewPool ? [adminSearch, taskStatusFilter, requestedCaptureDirectory] : []), ]); useEffect(() => { - if (!pagedReviewPool) return; + if (!pagedReviewPool) { + setReviewPoolCountError(""); + setReviewPoolCountLoading(false); + return; + } const controller = new AbortController(); setReviewPoolCountError(""); + setReviewPoolCountLoading(true); void humanLoopApi .poolCounts( { @@ -1098,17 +1105,26 @@ export default function AnnotationOperations({ workPage = false }: { workPage?: { signal: controller.signal }, ) .then((result) => { - if (!controller.signal.aborted) setReviewPoolTotals({ scope: projectDataScope, counts: result.counts }); + if (controller.signal.aborted) return; + if ( + !["available", "mine", "reviewed", ...(isManager ? ["all"] : [])].every( + (key) => Number.isInteger(result.counts?.[key]) && result.counts[key] >= 0, + ) + ) + throw new Error("Incomplete task counts"); + setReviewPoolTotals({ scope: projectDataScope, counts: result.counts }); }) - .catch((cause) => { + .catch(() => { if (!controller.signal.aborted) { setReviewPoolTotals(null); - if (cause?.status !== 404 && cause?.status !== 405) - setReviewPoolCountError(getActionErrorMessage(cause, "加载任务数量失败")); + setReviewPoolCountError("任务数量暂时无法读取"); } + }) + .finally(() => { + if (!controller.signal.aborted) setReviewPoolCountLoading(false); }); return () => controller.abort(); - }, [pagedReviewPool, projectDataScope, humanPools]); + }, [pagedReviewPool, projectDataScope, humanPools, reviewPoolCountRevision]); const reviewPoolViewRef = useRef(taskPoolView); reviewPoolViewRef.current = taskPoolView; const reviewActionScopeRef = useRef(projectDataScope); @@ -7293,9 +7309,27 @@ export default function AnnotationOperations({ workPage = false }: { workPage?:
逐段任务
+ {pagedReviewPool && reviewPoolCountLoading && ( +
+ 任务数量加载中… +
+ )} {reviewPoolCountError && ( -
- {reviewPoolCountError} +
+ {reviewPoolCountError} +
)}
diff --git a/tests/browser/review-pool-count-status.mjs b/tests/browser/review-pool-count-status.mjs new file mode 100644 index 00000000..64b17ae8 --- /dev/null +++ b/tests/browser/review-pool-count-status.mjs @@ -0,0 +1,159 @@ +// The actual application, isolated read-only API responses. No production endpoints. +import assert from "node:assert/strict"; +import fs from "node:fs/promises"; +import { createServer } from "vite"; +import { chromium } from "playwright"; +import { browserExecutableOptions } from "../../scripts/lib/browser-executable.mjs"; +import { closeBrowserFixture } from "../../scripts/lib/browser-test-lifecycle.mjs"; +const out = "local-data/test-results/issue20/browser"; +await fs.mkdir(out, { recursive: true }); +let browser, server; +try { + server = await createServer({ + configFile: "tests/browser/vite.platform-feedback.config.ts", + server: { host: "127.0.0.1", port: 0 }, + }); + await server.listen(); + browser = await chromium.launch({ ...browserExecutableOptions(), headless: true, args: ["--no-proxy-server"] }); + for (const [role, mode] of [ + ["acceptance_reviewer", "acceptance"], + ["quality_reviewer", "quality"], + ]) { + const page = await browser.newPage({ viewport: { width: 1500, height: 1000 } }); + page.setDefaultTimeout(10000); + let fail = true, + reads = 0; + const errors = []; + page.on("pageerror", (e) => errors.push(e.message)); + const slowSeen = Promise.withResolvers(), + releaseSlow = Promise.withResolvers(); + await page.route("**/worker-api/**", (route) => route.fulfill({ json: { ready: true, status: "ok" } })); + await page.route("**/api/**", async (route) => { + const req = route.request(), + u = new URL(req.url()), + p = u.pathname; + if (!p.startsWith("/api/")) return route.continue(); + assert.equal(req.method(), "GET", "pool status verification must not write"); + const json = (data) => route.fulfill({ json: data }); + if (p === "/api/auth/me") + return json({ + user: { id: "reviewer", name: "Synthetic", role, status: "active", organizationId: "org" }, + organization: { id: "org", status: "active" }, + }); + if (p === "/api/plugins") + return json({ items: [{ enabled: true, manifest: { id: "annotation-qa", name: "标注", version: "1" } }] }); + if (p === "/api/projects") + return json({ items: [{ id: "p", name: "Synthetic project", organizationId: "org", status: "active" }] }); + if (p.endsWith("/members")) return json({ items: [{ userId: "reviewer", projectId: "p", roles: [role] }] }); + if (p === "/api/human-item-counts") { + reads++; + const search = u.searchParams.get("search"); + if (search === "slow") { + slowSeen.resolve(); + await releaseSlow.promise; + return json({ counts: { available: 99, mine: 99, reviewed: 99 }, filtersApplied: true }).catch(() => {}); + } + if (search === "fast") return json({ counts: { available: 3, mine: 4, reviewed: 5 }, filtersApplied: true }); + if (search === "incomplete") return json({ counts: { available: 3 }, filtersApplied: true }); + return fail + ? route.fulfill({ + status: 500, + json: { + error: "(MaxTimeMSExpired) PlanExecutor error during aggregation: operation exceeded time limit", + }, + }) + : json({ counts: { available: 7, mine: 1, reviewed: 9 }, filtersApplied: true }); + } + if (p === "/api/human-item-summaries") + return json({ + summaries: + mode === "acceptance" && u.searchParams.get("pool") === "mine" && !u.searchParams.get("search") + ? [ + { + kind: "pool-summary", + id: "owned-acceptance", + organizationId: "org", + projectId: "p", + workflowId: "w", + runId: "r", + batchId: "b", + pluginId: "annotation-qa", + currentStageId: "prepackage_acceptance", + status: "claimed", + assignedUserId: "reviewer", + processingProfile: "raw_human", + version: 1, + createdAt: "2026-11-15T00:00:00Z", + subject: { itemKey: "synthetic-capture" }, + display: { captureDirectories: ["synthetic-capture"], projectName: "Synthetic project" }, + durationSeconds: 8, + }, + ] + : [], + hasMore: false, + filtersApplied: true, + }); + return json({ items: [], hasMore: false, labelIds: [], sliceLabelIds: [] }); + }); + await page.goto( + `${server.resolvedUrls.local[0]}plugins/annotation-qa/workflows/all/${mode}?projectId=p&pool=mine`, + { waitUntil: "domcontentloaded" }, + ); + const mine = page.getByTestId("human-task-pool-mine"); + await mine.waitFor(); + await page + .getByRole("alert") + .filter({ hasText: /系统暂时无法完成操作|任务数量暂时无法读取/ }) + .waitFor(); + assert.match(await mine.innerText(), /—/); + if (mode === "acceptance") { + const action = page.getByTestId("acceptance-task-card").getByRole("button", { name: "继续处理", exact: true }); + await action.waitFor(); + assert(!(await action.isDisabled()), "count failure must not block continuing an owned acceptance task"); + } + // Red on the screenshot symptom: the count error currently has no scoped recovery. + const retry = page.getByRole("button", { name: "重试任务数量", exact: true }); + await retry.waitFor({ timeout: 1500 }); + assert.match(await page.getByTestId("review-pool-count-error").innerText(), /任务数量暂时无法读取/); + await page.screenshot({ path: `${out}/${mode}-count-failed.png`, fullPage: true }); + fail = false; + const before = reads; + await retry.click(); + await mine.filter({ hasText: /我的任务\s+1/ }).waitFor(); + assert.equal(await mine.getAttribute("aria-pressed"), "true", "count retry must not switch task pools"); + assert.equal( + await page.getByTestId("review-pool-count-error").count(), + 0, + "successful refresh clears the count error", + ); + assert(reads > before); + assert.match(await page.getByTestId("human-task-pool-available").innerText(), /7/); + assert.match(await page.getByTestId("human-task-pool-reviewed").innerText(), /9/); + await page.screenshot({ path: `${out}/${mode}-count-recovered.png`, fullPage: true }); + const search = page.getByPlaceholder("搜索 Capture"); + await search.fill("slow"); + await slowSeen.promise; + await search.fill("fast"); + await mine.filter({ hasText: /我的任务\s+4/ }).waitFor(); + releaseSlow.resolve(); + await page.waitForTimeout(100); + assert.match(await mine.innerText(), /我的任务\s+4/, "obsolete count response cannot overwrite the current search"); + await search.fill("incomplete"); + await page.getByTestId("review-pool-count-error").waitFor(); + assert.match( + await mine.innerText(), + /—/, + "incomplete totals remain unknown, never fabricated zero or stale totals", + ); + await search.fill(""); + await mine.filter({ hasText: /我的任务\s+1/ }).waitFor(); + assert.equal(await page.getByTestId("review-pool-count-error").count(), 0); + assert.deepEqual(errors, []); + await page.close(); + } + console.log( + "PASS acceptance and quality: count timeout remains unknown, explicit retry restores scoped totals and clears error without changing selection or writing", + ); +} finally { + await closeBrowserFixture({ browser, vite: server }); +}