Return errors instead of exiting on rollback, rows close and blob temp file cleanup (closes #345)

WithReadTx ended the process when its rollback failed, including after
Ctrl-C, when database/sql has already rolled the transaction back and
Rollback returns sql.ErrTxDone. It now handles the rollback as WithTx
does. The query functions ended the process when closing their rows
failed; they now ignore the deferred close, because database/sql reports
a close failure through rows.Err(), which each returns, once the rows
are read to the end. The unused database.Fatalf is removed.
cleanupBlobTempFile returns its close or remove failure and the upload
step returns it, so such a failure now fails the backup.

No test makes a rows close fail: SQLite fails it only with the error the
last read already returned.

Model: opus-5-5

internal/database/blob_chunks.go

@@ -58,12 +58,7 @@ func (r *BlobChunkRepository) GetByBlobID(
 		return nil, fmt.Errorf("querying blob chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var blobChunks []*BlobChunk
 

internal/database/blobs.go

@@ -83,12 +83,7 @@ func (r *BlobRepository) GetAll(ctx context.Context) (map[string]*Blob, error) {
 		return nil, fmt.Errorf("querying blobs: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	out := make(map[string]*Blob)
 

internal/database/chunk_files.go

@@ -61,12 +61,7 @@ func (r *ChunkFileRepository) GetByChunkHash(
 		return nil, fmt.Errorf("querying chunk files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanChunkFiles(rows)
 }
@@ -87,12 +82,7 @@ func (r *ChunkFileRepository) GetByFilePath(
 		return nil, fmt.Errorf("querying chunk files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanChunkFiles(rows)
 }
@@ -112,12 +102,7 @@ func (r *ChunkFileRepository) GetByFileID(
 		return nil, fmt.Errorf("querying chunk files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanChunkFiles(rows)
 }

internal/database/chunks.go

@@ -107,12 +107,7 @@ func (r *ChunkRepository) GetByHashes(
 		return nil, fmt.Errorf("querying chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var chunks []*Chunk
 
@@ -152,12 +147,7 @@ func (r *ChunkRepository) ListUnpacked(
 		return nil, fmt.Errorf("querying unpacked chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var chunks []*Chunk
 

internal/database/chunks_ext.go

@@ -38,12 +38,7 @@ func (r *ChunkRepository) list(ctx context.Context, query string) ([]*Chunk, err
 		return nil, fmt.Errorf("querying chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var chunks []*Chunk
 

internal/database/errors.go

@@ -1,12 +0,0 @@
-package database
-
-import (
-	"fmt"
-	"os"
-)
-
-// Fatalf prints an error message to stderr and exits with status 1
-func Fatalf(format string, args ...any) {
-	fmt.Fprintf(os.Stderr, "FATAL: "+format+"\n", args...)
-	os.Exit(1)
-}

internal/database/file_chunks.go

@@ -62,12 +62,7 @@ func (r *FileChunkRepository) GetByPath(
 		return nil, fmt.Errorf("querying file chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanFileChunks(rows)
 }
@@ -88,12 +83,7 @@ func (r *FileChunkRepository) GetByFileID(
 		return nil, fmt.Errorf("querying file chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanFileChunks(rows)
 }
@@ -117,12 +107,7 @@ func (r *FileChunkRepository) GetByPathTx(
 		return nil, fmt.Errorf("querying file chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	fileChunks, err := r.scanFileChunks(rows)
 	LogSQL("GetByPathTx", "Complete", path, "count", len(fileChunks))

internal/database/files.go

@@ -175,12 +175,7 @@ func (r *FileRepository) ListModifiedSince(
 		return nil, fmt.Errorf("querying files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var files []*File
 
@@ -256,12 +251,7 @@ func (r *FileRepository) ListUnderPath(
 		return nil, fmt.Errorf("querying files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var files []*File
 
@@ -304,12 +294,7 @@ func (r *FileRepository) ListIDsWithChunksNotInUploadedBlobs(
 		return nil, fmt.Errorf("querying files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var ids []types.FileID
 
@@ -340,12 +325,7 @@ func (r *FileRepository) ListAll(ctx context.Context) ([]*File, error) {
 		return nil, fmt.Errorf("querying files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var files []*File
 

internal/database/repositories.go

@@ -106,22 +106,24 @@ func (r *Repositories) WithReadTx(ctx context.Context, fn TxFunc) error {
 
 	defer func() {
 		if p := recover(); p != nil {
-			rollbackErr := tx.Rollback()
-			if rollbackErr != nil {
-				Fatalf("failed to rollback transaction: %v", rollbackErr)
-			}
+			// The panic is the failure to report, so a rollback error
+			// is dropped.
+			_ = tx.Rollback()
 
 			panic(p)
-		} else if err != nil {
-			rollbackErr := tx.Rollback()
-			if rollbackErr != nil {
-				Fatalf("failed to rollback transaction: %v", rollbackErr)
-			}
 		}
 	}()
 
 	err = fn(ctx, tx)
 	if err != nil {
+		// Once ctx is cancelled, database/sql rolls the transaction
+		// back itself and Rollback returns sql.ErrTxDone.
+		rollbackErr := tx.Rollback()
+		if rollbackErr != nil && !errors.Is(rollbackErr, sql.ErrTxDone) {
+			return fmt.Errorf("%w; rolling back transaction: %w",
+				err, rollbackErr)
+		}
+
 		return err
 	}
 

internal/database/repositories_test.go

@@ -411,3 +411,41 @@ func TestRepositoriesReadTransaction(t *testing.T) {
 		t.Error("expected to retrieve file in read transaction")
 	}
 }
+
+// When the context is cancelled during a read transaction's body,
+// database/sql rolls the transaction back itself, so WithReadTx's own
+// rollback finds it already done. WithReadTx must return the body's error
+// alone, without that rollback.
+func TestRepositoriesReadTransactionContextCancelled(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	repos := database.NewRepositories(db)
+
+	err := repos.WithReadTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		cancel()
+
+		// database/sql rolls back from a goroutine of its own; wait
+		// until it has, so the test does not depend on that race.
+		for {
+			_, err := tx.ExecContext(context.WithoutCancel(ctx), "SELECT 1")
+			if errors.Is(err, sql.ErrTxDone) {
+				return ctx.Err()
+			}
+
+			time.Sleep(time.Millisecond)
+		}
+	})
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("expected context.Canceled, got: %v", err)
+	}
+
+	if errors.Is(err, sql.ErrTxDone) {
+		t.Fatalf("the already-done rollback was reported: %v", err)
+	}
+}

internal/database/snapshots.go

@@ -239,12 +239,7 @@ func (r *SnapshotRepository) GetOnlySnapshot(ctx context.Context) (*Snapshot, er
 		return nil, fmt.Errorf("querying snapshots: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	snapshots, err := r.scanSnapshotRows(rows)
 	if err != nil {
@@ -279,12 +274,7 @@ func (r *SnapshotRepository) ListRecent(
 		return nil, fmt.Errorf("querying snapshots: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanSnapshotRows(rows)
 }
@@ -499,12 +489,7 @@ func (r *SnapshotRepository) GetBlobHashes(
 		return nil, fmt.Errorf("querying blob hashes: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var blobs []string
 
@@ -653,12 +638,7 @@ func (r *SnapshotRepository) GetIncompleteSnapshots(
 		return nil, fmt.Errorf("querying incomplete snapshots: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanSnapshotRows(rows)
 }
@@ -681,12 +661,7 @@ func (r *SnapshotRepository) GetIncompleteByHostname(
 		return nil, fmt.Errorf("querying incomplete snapshots: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	// Same column set as every other multi-row snapshot query, so the
 	// shared scanner applies — including its timestamp normalization.

internal/snapshot/blob_temp_file_test.go

@@ -0,0 +1,62 @@
+//nolint:testpackage // exercises the unexported cleanupBlobTempFile
+package snapshot
+
+import (
+	"errors"
+	"os"
+	"testing"
+
+	"github.com/spf13/afero"
+	"sneak.berlin/go/vaultik/internal/blob"
+)
+
+// A blob temp file that cannot be closed or removed must make the cleanup
+// return an error, not end the process.
+func TestCleanupBlobTempFileReturnsErrors(t *testing.T) {
+	t.Parallel()
+
+	fs := afero.NewOsFs()
+	s := &Scanner{fs: fs}
+
+	tests := []struct {
+		name string
+		// prepare runs before the cleanup and makes its close or remove
+		// fail.
+		prepare func(tempFile afero.File) error
+		want    error
+	}{
+		{
+			name:    "close fails",
+			prepare: func(tempFile afero.File) error { return tempFile.Close() },
+			want:    os.ErrClosed,
+		},
+		{
+			name: "remove fails",
+			prepare: func(tempFile afero.File) error {
+				return fs.Remove(tempFile.Name())
+			},
+			want: os.ErrNotExist,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			tempFile, err := afero.TempFile(fs, t.TempDir(), "blob-*.tmp")
+			if err != nil {
+				t.Fatal(err)
+			}
+
+			err = tt.prepare(tempFile)
+			if err != nil {
+				t.Fatal(err)
+			}
+
+			err = s.cleanupBlobTempFile(&blob.WithReader{TempFile: tempFile})
+			if !errors.Is(err, tt.want) {
+				t.Fatalf("expected an error wrapping %v, got: %v", tt.want, err)
+			}
+		})
+	}
+}

internal/snapshot/scanner.go

@@ -1553,19 +1553,23 @@ func (s *Scanner) handleBlobReady(
 	blobExists, err := s.uploadBlobIfNeeded(
 		ctx, blobPath, blobWithReader, startTime, result)
 	if err != nil {
-		s.cleanupBlobTempFile(blobWithReader)
+		cleanupErr := s.cleanupBlobTempFile(blobWithReader)
 
-		return fmt.Errorf("uploading blob %s: %w", finishedBlob.Hash, err)
+		return errors.Join(
+			fmt.Errorf("uploading blob %s: %w", finishedBlob.Hash, err), cleanupErr)
 	}
 
 	err = s.recordBlobMetadata(ctx, finishedBlob, blobExists, startTime)
 	if err != nil {
-		s.cleanupBlobTempFile(blobWithReader)
+		cleanupErr := s.cleanupBlobTempFile(blobWithReader)
 
-		return err
+		return errors.Join(err, cleanupErr)
 	}
 
-	s.cleanupBlobTempFile(blobWithReader)
+	err = s.cleanupBlobTempFile(blobWithReader)
+	if err != nil {
+		return err
+	}
 
 	// Chunks from this blob are now committed to DB - remove from pending set
 	s.removePendingChunkHashes(blobWithReader.InsertedChunkHashes)
@@ -1759,20 +1763,22 @@ func (s *Scanner) recordBlobMetadata(
 }
 
 // cleanupBlobTempFile closes and removes the blob's temporary file
-func (s *Scanner) cleanupBlobTempFile(blobWithReader *blob.WithReader) {
+func (s *Scanner) cleanupBlobTempFile(blobWithReader *blob.WithReader) error {
 	if blobWithReader.TempFile != nil {
 		tempName := blobWithReader.TempFile.Name()
 
 		err := blobWithReader.TempFile.Close()
 		if err != nil {
-			log.Fatal("Failed to close temp file", "file", tempName, "error", err)
+			return fmt.Errorf("closing blob temp file: %w", err)
 		}
 
 		err = s.fs.Remove(tempName)
 		if err != nil {
-			log.Fatal("Failed to remove temp file", "file", tempName, "error", err)
+			return fmt.Errorf("removing blob temp file: %w", err)
 		}
 	}
+
+	return nil
 }
 
 // streamingChunkInfo tracks chunk metadata collected during streaming