diff --git a/CHANGELOG.md b/CHANGELOG.md index 3d51f15..01b54cf 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,5 +1,12 @@ # Changelog +## 开发快照 snapshot/v0.1.98 - 2026-10-28 + +### Fixed + +- 修复整库候选封存只单向校验搜索索引的问题:数据库标记为已索引的分块必须在导出索引中存在,否则拒绝封存;按分块身份而非总数量核对,避免重复索引掩盖缺失。仅存储的父分块仍允许不建索引。 +- 本修复增强候选完整性,不代表整库业务入口、生产同步恢复或语义验收已经完成。 + ## 开发快照 snapshot/v0.1.97 - 2026-10-28 ### Fixed diff --git a/internal/releaseadapter/v071/library_archive.go b/internal/releaseadapter/v071/library_archive.go index 6f94dd5..0953955 100644 --- a/internal/releaseadapter/v071/library_archive.go +++ b/internal/releaseadapter/v071/library_archive.go @@ -116,6 +116,7 @@ type libraryRow struct { OwnerType string `json:"owner_type"` OwnerID string `json:"owner_id"` Content string `json:"content"` + Status int `json:"status"` } func validateLibraryCandidate(identity PostgresIdentity, pg PostgresSnapshot, es SearchSnapshot) error { @@ -220,6 +221,7 @@ func validateLibraryCandidate(identity PostgresIdentity, pg PostgresSnapshot, es return errors.New("invalid library index digest") } seen := map[string]bool{} + indexedChunks := map[string]bool{} for _, document := range es.Documents { var row libraryRow if json.Unmarshal(document.Source, &row) != nil || !validSearchIdentifier(document.ID) || seen[document.ID] { @@ -233,6 +235,15 @@ func validateLibraryCandidate(identity PostgresIdentity, pg PostgresSnapshot, es if !searchHitMatches(document.Source, SearchIdentity{TenantID: identity.TenantID, KnowledgeBaseID: identity.KnowledgeBaseID, KnowledgeID: row.KnowledgeID}) { return errors.New("cross-scope library index") } + indexedChunks[row.ChunkID] = true + } + // v0.7.1 ChunkStatusIndexed is 2. One chunk can have several search + // records, so compare identities rather than counts. Stored-only parent + // chunks intentionally have no index and must not be rejected as missing. + for id, chunk := range tables["chunks"] { + if chunk.Status == 2 && !indexedChunks[id] { + return errors.New("missing indexed library chunk") + } } return nil } diff --git a/internal/releaseadapter/v071/library_index_completeness_test.go b/internal/releaseadapter/v071/library_index_completeness_test.go new file mode 100644 index 0000000..762ec64 --- /dev/null +++ b/internal/releaseadapter/v071/library_index_completeness_test.go @@ -0,0 +1,71 @@ +package v071 + +import ( + "context" + "encoding/json" + "os" + "strings" + "testing" + "time" + + "knowledge-governance/internal/librarysnapshot" +) + +func TestLibraryCandidateIndexCompletenessUsesChunkIdentity(t *testing.T) { + for _, mode := range []string{"stored-parent", "multiple-records-one-chunk", "same-count-missing-chunk"} { + t.Run(mode, func(t *testing.T) { + pg, es, _ := libraryCandidateFixture() + rows := []json.RawMessage{json.RawMessage(`{"id":"chunk-safe","tenant_id":10000,"knowledge_base_id":"kb-safe","knowledge_id":"doc-safe","content":"chunk","status":2,"chunk_type":"text"}`)} + if mode == "stored-parent" { + rows = append(rows, json.RawMessage(`{"id":"parent-safe","tenant_id":10000,"knowledge_base_id":"kb-safe","knowledge_id":"doc-safe","content":"context","status":1,"chunk_type":"parent_text"}`)) + } else { + es.Documents = append(es.Documents, SearchDocument{ID: "es-second", Source: append(json.RawMessage(nil), es.Documents[0].Source...)}) + if mode == "same-count-missing-chunk" { + rows = append(rows, json.RawMessage(`{"id":"chunk-second","tenant_id":10000,"knowledge_base_id":"kb-safe","knowledge_id":"doc-safe","content":"second","status":2,"chunk_type":"text"}`)) + } + } + pg.Tables[1] = newTableSnapshot("chunks", rows) + es.DocumentsSHA256, _ = searchDocumentsDigest(es.Documents) + err := validateLibraryCandidate(PostgresIdentity{TenantID: 10000, KnowledgeBaseID: "kb-safe"}, pg, es) + if mode == "same-count-missing-chunk" { + if err == nil || !strings.Contains(err.Error(), "missing indexed library chunk") { + t.Fatalf("same counts masked a missing chunk: %v", err) + } + } else if err != nil { + t.Fatalf("valid index shape rejected: %v", err) + } + }) + } +} + +func TestLibraryCandidateRejectsMissingIndexedChunk(t *testing.T) { + pg, es, objects := libraryCandidateFixture() + pg.Tables[1] = newTableSnapshot("chunks", []json.RawMessage{json.RawMessage(`{"id":"chunk-safe","tenant_id":10000,"knowledge_base_id":"kb-safe","knowledge_id":"doc-safe","content":"chunk","status":2,"chunk_type":"text"}`)}) + identity := PostgresIdentity{TenantID: 10000, KnowledgeBaseID: "kb-safe"} + if err := validateLibraryCandidate(identity, pg, es); err != nil { + t.Fatalf("complete indexed chunk rejected: %v", err) + } + es.Documents = nil + es.DocumentsSHA256, _ = searchDocumentsDigest(es.Documents) + root := t.TempDir() + if err := os.Chmod(root, 0700); err != nil { + t.Fatal(err) + } + archive, err := librarysnapshot.Open(root, 1024*1024) + if err != nil { + t.Fatal(err) + } + defer archive.Close() + contract := librarysnapshot.Compatibility{WeKnoraVersion: "v0.7.1", SchemaVersion: 74, RetrieverEngine: "elasticsearch_v8", EmbeddingModel: "fixture-model", EmbeddingDimension: 1024, IndexMappingSHA256: es.MappingSHA256} + id, err := ArchiveLibraryCandidate(context.Background(), archive, "internal", identity, contract, pg, es, objects, time.Now().UTC()) + if err == nil || id != "" { + t.Fatal("sealed a supposedly complete library with an indexed PostgreSQL chunk missing from Elasticsearch") + } + if !strings.Contains(err.Error(), "missing indexed library chunk") { + t.Fatalf("unexpected refusal: %v", err) + } + versions, err := os.ReadDir(root + "/versions") + if err != nil || len(versions) != 0 { + t.Fatalf("incomplete candidate became visible: versions=%d err=%v", len(versions), err) + } +}