Merge pull request #1 from tae2089/refactor/trim-peripheral-features

Trim peripheral features, improve search, and fix review findings

Author: tae2089

.github/workflows/ci.yml

@@ -30,6 +30,9 @@ jobs:
       - name: Run go vet
         run: make vet
 
+      - name: Run go test
+        run: CGO_ENABLED=1 go test -tags "fts5" ./... -count=1
+
       - name: Build release binaries
         run: make build
 

.gitignore

@@ -27,3 +27,4 @@ pyproject.toml
 /web/wiki/dist/
 /web/wiki/node_modules/
 search-retrieval.md
+_workspace/

cmd/ccg/main_postgres_test.go

@@ -157,7 +157,7 @@ func TestRunMigrations_PostgresDownRestoresNullableColumns(t *testing.T) {
 		t.Fatalf("run down migration: %v", err)
 	}
 
-	var version migrateSchemaVersion
+	var version migration.MigrationSchemaVersion
 	if err := db.Table("schema_migrations").First(&version).Error; err != nil {
 		t.Fatalf("load schema version: %v", err)
 	}
@@ -190,7 +190,7 @@ func TestRunMigrations_PostgresDownFromVersionThreeDropsPolicyTables(t *testing.
 		t.Fatalf("run down migration: %v", err)
 	}
 
-	var version migrateSchemaVersion
+	var version migration.MigrationSchemaVersion
 	if err := db.Table("schema_migrations").First(&version).Error; err != nil {
 		t.Fatalf("load schema version: %v", err)
 	}

docker-compose.langfuse.yaml

@@ -37,7 +37,10 @@ func NewExecGitClient() *ExecGitClient {
 // @requires repoDir points to a valid git working tree
 // @ensures returned file paths are trimmed and exclude blank lines
 func (g *ExecGitClient) ChangedFiles(ctx context.Context, repoDir, baseRef string) ([]string, error) {
-	out, err := runGitLimited(ctx, repoDir, []string{"diff", "--name-only", baseRef})
+	if err := validateBaseRef(baseRef); err != nil {
+		return nil, err
+	}
+	out, err := runGitLimited(ctx, repoDir, []string{"-c", "core.quotePath=false", "diff", "--name-only", baseRef, "--"})
 	if err != nil {
 		return nil, trace.Wrap(err, "git diff --name-only")
 	}
@@ -66,7 +69,10 @@ func (g *ExecGitClient) ChangedFiles(ctx context.Context, repoDir, baseRef strin
 // @requires repoDir points to a valid git working tree
 // @ensures each returned hunk has a file path and inclusive start/end lines
 func (g *ExecGitClient) DiffHunks(ctx context.Context, repoDir, baseRef string, paths []string) ([]Hunk, error) {
-	args := []string{"diff", "-U0", baseRef, "--"}
+	if err := validateBaseRef(baseRef); err != nil {
+		return nil, err
+	}
+	args := []string{"-c", "core.quotePath=false", "diff", "-U0", baseRef, "--"}
 	args = append(args, paths...)
 	out, err := runGitLimited(ctx, repoDir, args)
 	if err != nil {
@@ -98,6 +104,19 @@ func (g *ExecGitClient) DiffHunks(ctx context.Context, repoDir, baseRef string,
 	return hunks, nil
 }
 
+// validateBaseRef rejects base revisions that git would parse as command options.
+// @intent block git argument injection through the caller-supplied base ref, which sits
+// before the "--" separator and would otherwise be interpreted as a diff flag.
+func validateBaseRef(baseRef string) error {
+	if baseRef == "" {
+		return fmt.Errorf("base ref must not be empty")
+	}
+	if strings.HasPrefix(baseRef, "-") {
+		return fmt.Errorf("invalid base ref %q: must not start with '-'", baseRef)
+	}
+	return nil
+}
+
 // runGitLimited runs a git subcommand with the default output size cap.
 // @intent share a single bounded git invocation helper across diff operations
 func runGitLimited(ctx context.Context, repoDir string, args []string) ([]byte, error) {

internal/analysis/changes/git.go

@@ -29,6 +29,46 @@ func TestGitClient_ChangedFiles(t *testing.T) {
 	}
 }
 
+func TestGitClient_RejectsOptionLikeBaseRef(t *testing.T) {
+	dir := initTestRepo(t)
+	writeFile(t, dir, "hello.go", "package main\n")
+	gitCommit(t, dir, "initial")
+
+	git := NewExecGitClient()
+	for _, base := range []string{"--output=/tmp/pwned", "-U9999", "--no-index"} {
+		if _, err := git.ChangedFiles(context.Background(), dir, base); err == nil {
+			t.Errorf("ChangedFiles accepted option-like base %q", base)
+		}
+		if _, err := git.DiffHunks(context.Background(), dir, base, []string{"hello.go"}); err == nil {
+			t.Errorf("DiffHunks accepted option-like base %q", base)
+		}
+	}
+}
+
+func TestGitClient_ChangedFiles_NonASCIIPath(t *testing.T) {
+	dir := initTestRepo(t)
+	writeFile(t, dir, "한글.go", "package main\n")
+	gitCommit(t, dir, "initial")
+	writeFile(t, dir, "한글.go", "package main\n\nfunc Hello() {}\n")
+
+	git := NewExecGitClient()
+	files, err := git.ChangedFiles(context.Background(), dir, "HEAD")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(files) != 1 || files[0] != "한글.go" {
+		t.Fatalf("expected unquoted 한글.go, got %v", files)
+	}
+
+	hunks, err := git.DiffHunks(context.Background(), dir, "HEAD", []string{"한글.go"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(hunks) == 0 || hunks[0].FilePath != "한글.go" {
+		t.Fatalf("expected hunk for unquoted 한글.go, got %+v", hunks)
+	}
+}
+
 func TestGitClient_DiffHunks(t *testing.T) {
 	dir := initTestRepo(t)
 	writeFile(t, dir, "hello.go", "package main\n\nfunc Old() {}\n")

internal/analysis/changes/git_test.go

@@ -86,13 +86,17 @@ func (s *Service) FindSuspectsPage(ctx context.Context, opts Options) (Result, e
 	results := make([]SuspectEdge, 0, len(edges))
 	for _, edge := range edges {
 		source, err := s.store.GetNodeByID(ctx, edge.FromNodeID)
-		if err != nil || source == nil {
+		if err != nil {
 			return Result{}, err
 		}
 		target, err := s.store.GetNodeByID(ctx, edge.ToNodeID)
-		if err != nil || target == nil {
+		if err != nil {
 			return Result{}, err
 		}
+		// A dangling edge (endpoint node missing) must not silently drop the whole page.
+		if source == nil || target == nil {
+			continue
+		}
 		sourceAnn, err := s.store.GetAnnotation(ctx, source.ID)
 		if err != nil {
 			return Result{}, err

internal/analysis/fallback/service.go

@@ -63,6 +63,44 @@ func TestSuspectFallbackEdges_FlagsDisjointIntentAndDomainRule(t *testing.T) {
 	}
 }
 
+func TestSuspectFallbackEdges_SkipsEdgeWithMissingEndpointInsteadOfDroppingPage(t *testing.T) {
+	db := setupFallbackDB(t)
+	ctx := context.Background()
+	store := gormstore.New(db)
+
+	source := model.Node{QualifiedName: "pkg.A", Kind: model.NodeKindFunction, Name: "A", FilePath: "a.go", StartLine: 1, EndLine: 2, Language: "go"}
+	target := model.Node{QualifiedName: "pkg.B", Kind: model.NodeKindFunction, Name: "B", FilePath: "b.go", StartLine: 1, EndLine: 2, Language: "go"}
+	if err := store.UpsertNodes(ctx, []model.Node{source, target}); err != nil {
+		t.Fatal(err)
+	}
+	sourceNode, _ := store.GetNode(ctx, "pkg.A")
+	targetNode, _ := store.GetNode(ctx, "pkg.B")
+	if err := store.UpsertEdges(ctx, []model.Edge{
+		// dangling edge: endpoint node id does not exist
+		{FromNodeID: 99999, ToNodeID: targetNode.ID, Kind: model.EdgeKindFallbackCalls, Fingerprint: "a-dangling"},
+		{FromNodeID: sourceNode.ID, ToNodeID: targetNode.ID, Kind: model.EdgeKindFallbackCalls, Fingerprint: "b-valid"},
+	}); err != nil {
+		t.Fatal(err)
+	}
+	if err := store.UpsertAnnotation(ctx, &model.Annotation{NodeID: sourceNode.ID, Summary: "auth", Tags: []model.DocTag{{Kind: model.TagIntent, Value: "verify credentials", Ordinal: 0}}}); err != nil {
+		t.Fatal(err)
+	}
+	if err := store.UpsertAnnotation(ctx, &model.Annotation{NodeID: targetNode.ID, Summary: "billing", Tags: []model.DocTag{{Kind: model.TagIntent, Value: "render invoice", Ordinal: 0}}}); err != nil {
+		t.Fatal(err)
+	}
+
+	results, err := New(db, store).FindSuspects(ctx, Options{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(results) != 1 {
+		t.Fatalf("dangling edge must be skipped, not drop the whole page: got %d results", len(results))
+	}
+	if results[0].Source.QualifiedName != "pkg.A" {
+		t.Fatalf("expected surviving valid edge, got %+v", results[0])
+	}
+}
+
 func TestSuspectFallbackEdges_IgnoresOverlappingAnnotationContext(t *testing.T) {
 	db := setupFallbackDB(t)
 	ctx := context.Background()

internal/analysis/fallback/service_test.go

@@ -143,7 +143,7 @@ func (t *Tracer) TraceFlowBounded(ctx context.Context, startNodeID uint, opts Tr
 		Truncated:             truncated,
 		MaxNodes:              opts.MaxNodes,
 		ReturnedNodes:         len(members),
-		ContainsFallbackCalls:  containsFallbackCalls,
+		ContainsFallbackCalls: containsFallbackCalls,
 		FallbackEdgesCount:    fallbackEdges,
 	}, nil
 }

internal/analysis/flows/flows.go

@@ -84,7 +84,7 @@ func stripBlockDelimiters(text string, language string) string {
 }
 
 // stripPythonDocstringDelimiters removes triple-quote wrappers from a Python docstring body.
-// @intent expose the raw docstring text by trying both """ and ''' triple-quote forms.
+// @intent expose the raw docstring text by trying both """ and ”' triple-quote forms.
 func stripPythonDocstringDelimiters(text string) (string, bool) {
 	for _, quote := range []string{"\"\"\"", "'''"} {
 		if stripped, ok := stripPythonQuotedString(text, quote); ok {