Restore the other files when a chunk is in no blob under --skip-errors (closes #326)

Restore planning stopped at the first file needing a chunk that no blob
holds, so --skip-errors never saw that file and nothing was restored.
The planner now leaves such a file out of the plan and returns it with
the reason, and each one goes through handleRestoreFileError like any
other failed file: skipped and counted under --skip-errors, fatal
without it. The fatal error now suggests --skip-errors, as other restore
failures do.

The skip-errors tests now share one backup helper; the new case removes
a blob from the snapshot metadata and exports it again.

Model: opus-5-5

README.md

@@ -293,8 +293,8 @@ needed.
 **`snapshot list`**: Show every snapshot known to this host — the union
 of the local index and the backup destination store — with timestamps
 and three sizes per snapshot (compressed remote size; total
-uncompressed chunk size; size of chunks newly referenced by that
-snapshot).
+uncompressed chunk size; size of the chunks that snapshot stored that
+were not stored before, recorded when its backup ran).
 
 Listing the destination store does **not** require the age secret key,
 so it works in vaultik's intended configuration, where the backed-up
@@ -307,8 +307,8 @@ only the hash, so hostname and snapshot name exist solely in the local
 index and in the encrypted per-snapshot database. Snapshots found only
 on the destination store are therefore listed as
 `<remote only:<abbreviated remote key>>` and show `<remote only>` in
-the uncompressed and "new chunk" columns, which can only be computed
-from the local index. Their timestamp and compressed size are real,
+the uncompressed and "new chunk" columns, which come only from the
+local index. Their timestamp and compressed size are real,
 read from the manifest.
 
 Snapshots in the local index with no counterpart on the destination

docs/DATAMODEL.md

@@ -119,6 +119,7 @@ Tracks backup snapshots.
 - `completed_at` (INTEGER) - Completion timestamp (NULL if in progress)
 - `file_count` (INTEGER) - Number of files in snapshot
 - `chunk_count` (INTEGER) - Number of chunks this snapshot stored that were not stored before
+- `new_chunk_size` (INTEGER) - Total size of those chunks before compression; `snapshot list` shows it as NEW CHUNK SIZE
 - `blob_count` (INTEGER) - Number of blobs this snapshot created
 - `total_size` (INTEGER) - Total size of all files
 - `blob_size` (INTEGER) - Total compressed size of all referenced blobs

internal/database/repositories.go

@@ -3,6 +3,7 @@ package database
 import (
 	"context"
 	"database/sql"
+	"errors"
 	"fmt"
 )
 
@@ -60,22 +61,24 @@ func (r *Repositories) WithTx(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

@@ -4,6 +4,7 @@ import (
 	"context"
 	"database/sql"
 	"errors"
+	"strings"
 	"testing"
 	"time"
 
@@ -235,6 +236,125 @@ func TestRepositoriesTransactionRollback(t *testing.T) {
 	}
 }
 
+// When the context is cancelled during the body, database/sql rolls the
+// transaction back itself, so WithTx's own rollback finds it already
+// done. WithTx must return the body's error alone, without that rollback.
+func TestRepositoriesTransactionContextCancelled(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.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		file := &database.File{
+			Path:  "/test/cancelled_file.txt",
+			MTime: time.Now().Truncate(time.Second),
+			Size:  1024,
+			Mode:  0644,
+			UID:   1000,
+			GID:   1000,
+		}
+
+		err := repos.Files.Create(ctx, tx, file)
+		if err != nil {
+			return err
+		}
+
+		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)
+	}
+
+	file, err := repos.Files.GetByPath(
+		context.Background(), "/test/cancelled_file.txt",
+	)
+	if err != nil {
+		t.Fatalf("error checking for file: %v", err)
+	}
+
+	if file != nil {
+		t.Error("file should not exist after the cancelled transaction")
+	}
+}
+
+// A body that ends the transaction itself with ROLLBACK makes WithTx's
+// own rollback fail. WithTx must return that failure together with the
+// body's error.
+func TestRepositoriesTransactionRollbackFails(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repos := database.NewRepositories(db)
+
+	err := repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		_, err := tx.ExecContext(ctx, "ROLLBACK")
+		if err != nil {
+			return err
+		}
+
+		return errIntentionalRollback
+	})
+	if !errors.Is(err, errIntentionalRollback) {
+		t.Fatalf("expected the body's error, got: %v", err)
+	}
+
+	if !strings.Contains(err.Error(), "rolling back transaction") {
+		t.Errorf("expected the failed rollback in the error, got: %v", err)
+	}
+}
+
+// When the body ends the transaction itself with ROLLBACK and then
+// panics, WithTx's own rollback fails; the panic must still reach the
+// caller.
+func TestRepositoriesTransactionPanicRollbackFails(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repos := database.NewRepositories(db)
+
+	defer func() {
+		if recover() == nil {
+			t.Error("expected the body's panic to reach the caller")
+		}
+	}()
+
+	_ = repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		_, err := tx.ExecContext(ctx, "ROLLBACK")
+		if err != nil {
+			return err
+		}
+
+		panic(errIntentionalRollback)
+	})
+}
+
 func TestRepositoriesReadTransaction(t *testing.T) {
 	t.Parallel()
 

internal/database/schema/001.sql

@@ -86,6 +86,7 @@ CREATE TABLE IF NOT EXISTS snapshots (
     completed_at INTEGER,
     file_count INTEGER NOT NULL DEFAULT 0,
     chunk_count INTEGER NOT NULL DEFAULT 0,
+    new_chunk_size INTEGER NOT NULL DEFAULT 0,
     blob_count INTEGER NOT NULL DEFAULT 0,
     total_size INTEGER NOT NULL DEFAULT 0,
     blob_size INTEGER NOT NULL DEFAULT 0,

internal/database/snapshots.go

@@ -129,6 +129,7 @@ func (r *SnapshotRepository) UpdateExtendedStats(
 	compressionLevel int,
 	uploadBytes int64,
 	uploadDurationMs int64,
+	newChunkSize int64,
 ) error {
 	compressionRatio, err := r.extendedCompressionRatio(
 		ctx, tx, snapshotID, blobUncompressedSize,
@@ -143,18 +144,19 @@ func (r *SnapshotRepository) UpdateExtendedStats(
 		    compression_ratio = ?,
 		    compression_level = ?,
 		    upload_bytes = ?,
-		    upload_duration_ms = ?
+		    upload_duration_ms = ?,
+		    new_chunk_size = ?
 		WHERE id = ?
 	`
 
 	if tx != nil {
 		_, err = tx.ExecContext(ctx, query,
 			blobUncompressedSize, compressionRatio, compressionLevel,
-			uploadBytes, uploadDurationMs, snapshotID)
+			uploadBytes, uploadDurationMs, newChunkSize, snapshotID)
 	} else {
 		_, err = r.db.ExecWithLog(ctx, query,
 			blobUncompressedSize, compressionRatio, compressionLevel,
-			uploadBytes, uploadDurationMs, snapshotID)
+			uploadBytes, uploadDurationMs, newChunkSize, snapshotID)
 	}
 
 	if err != nil {
@@ -594,45 +596,24 @@ func (r *SnapshotRepository) GetSnapshotUncompressedChunkSize(
 	return totalSize, nil
 }
 
-// GetSnapshotNewChunkSize returns the sum of plaintext sizes of chunks that are
-// referenced by this snapshot but not by any earlier completed snapshot known to
-// the local database. The result is the marginal uncompressed data this snapshot
-// added to the dedup pool — i.e., the delta from prior snapshots.
+// GetSnapshotNewChunkSize returns the total plaintext size of the chunks the
+// snapshot's backup stored that were not stored before, as the backup
+// recorded it. It is not derived from the snapshot's files: a file's row is
+// shared by every snapshot that lists it and is rewritten when the file
+// changes, so it shows only the file's latest chunks.
 func (r *SnapshotRepository) GetSnapshotNewChunkSize(
 	ctx context.Context, snapshotID string,
 ) (int64, error) {
-	query := `
-		WITH this_snap_chunks AS (
-			SELECT DISTINCT fc.chunk_hash
-			FROM snapshot_files sf
-			JOIN file_chunks fc ON sf.file_id = fc.file_id
-			WHERE sf.snapshot_id = ?
-		),
-		prior_chunks AS (
-			SELECT DISTINCT fc.chunk_hash
-			FROM snapshots s
-			JOIN snapshot_files sf ON sf.snapshot_id = s.id
-			JOIN file_chunks fc ON fc.file_id = sf.file_id
-			WHERE s.completed_at IS NOT NULL
-			  AND s.id != ?
-			  AND s.started_at < (SELECT started_at FROM snapshots WHERE id = ?)
-		)
-		SELECT COALESCE(SUM(c.size), 0)
-		FROM chunks c
-		JOIN this_snap_chunks t ON c.chunk_hash = t.chunk_hash
-		WHERE c.chunk_hash NOT IN (SELECT chunk_hash FROM prior_chunks)
-	`
+	query := `SELECT new_chunk_size FROM snapshots WHERE id = ?`
 
-	var totalSize int64
+	var newChunkSize int64
 
-	err := r.db.conn.QueryRowContext(
-		ctx, query, snapshotID, snapshotID, snapshotID,
-	).Scan(&totalSize)
+	err := r.db.conn.QueryRowContext(ctx, query, snapshotID).Scan(&newChunkSize)
 	if err != nil {
 		return 0, fmt.Errorf("querying new chunk size: %w", err)
 	}
 
-	return totalSize, nil
+	return newChunkSize, nil
 }
 
 // GetIncompleteSnapshots returns all snapshots that haven't been completed

internal/snapshot/scanner.go

@@ -135,7 +135,9 @@ type ScannerConfig struct {
 // are counted per file: BytesScanned is the size of the new and changed
 // files, BytesSkipped that of the unchanged ones. FilesFailed counts the
 // new and changed files that phase 2 could not store; FilesScanned
-// includes them and BytesScanned does not.
+// includes them and BytesScanned does not. ChunksCreated counts the
+// chunks that were not already stored, and NewChunkSize is their total
+// size.
 type ScanResult struct {
 	FilesScanned   int
 	FilesSkipped   int
@@ -145,6 +147,7 @@ type ScanResult struct {
 	BytesSkipped   int64
 	BytesDeleted   int64
 	ChunksCreated  int
+	NewChunkSize   int64
 	BlobsCreated   int
 	BlobsUploaded  int
 	BytesUploaded  int64
@@ -1866,11 +1869,12 @@ func (s *Scanner) processFileStreaming(
 	return nil
 }
 
-// updateChunkStats counts a chunk that was not already stored. The scan
-// result's file counts, BytesScanned and BytesSkipped are not touched
-// here: the scan phase counts each file once.
+// updateChunkStats counts a chunk that was not already stored, and its
+// size. The scan result's file counts, BytesScanned and BytesSkipped are
+// not touched here: the scan phase counts each file once.
 func (s *Scanner) updateChunkStats(chunkSize int64, result *ScanResult) {
 	result.ChunksCreated++
+	result.NewChunkSize += chunkSize
 
 	if s.progress != nil {
 		s.progress.GetStats().ChunksCreated.Add(1)

internal/snapshot/snapshot.go

@@ -199,6 +199,7 @@ func (sm *SnapshotManager) UpdateSnapshotStatsExtended(
 			stats.CompressionLevel,
 			stats.BytesUploaded,
 			stats.UploadDurationMs,
+			stats.NewChunkSize,
 		)
 	})
 }
@@ -906,6 +907,7 @@ type ExtendedBackupStats struct {
 	BlobUncompressedSize int64 // Total uncompressed size of all referenced blobs
 	CompressionLevel     int   // Compression level used for this snapshot
 	UploadDurationMs     int64 // Total milliseconds spent uploading to S3
+	NewChunkSize         int64 // Total size of the chunks counted in ChunksCreated
 }
 
 // deleteOtherSnapshots deletes all snapshots except the current one

internal/vaultik/new_chunk_size_test.go

@@ -0,0 +1,133 @@
+package vaultik_test
+
+import (
+	"bytes"
+	"context"
+	"maps"
+	"path/filepath"
+	"testing"
+
+	"github.com/spf13/afero"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/chunker"
+	"sneak.berlin/go/vaultik/internal/database"
+	"sneak.berlin/go/vaultik/internal/log"
+	"sneak.berlin/go/vaultik/internal/storage"
+	"sneak.berlin/go/vaultik/internal/vaultik"
+)
+
+// TestNewChunkSizeOfChangedFileStaysFixed covers
+// https://git.eeqj.de/sneak/vaultik/issues/311. A backup whose only change
+// is a modified file lists the size of the chunks that change added as its
+// NEW CHUNK SIZE, and a later backup that changes the file again leaves
+// every earlier snapshot's figure as it was. The file's row in the local
+// index is shared by all three snapshots and rewritten by each backup.
+func TestNewChunkSizeOfChangedFileStaysFixed(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	fs := afero.NewOsFs()
+	tempDir := t.TempDir()
+	dataDir := filepath.Join(tempDir, "src")
+	dbPath := filepath.Join(tempDir, "index.sqlite")
+	changedPath := filepath.Join(dataDir, "changed.bin")
+
+	ctx := context.Background()
+	files := writeFaultSourceTree(t, fs, dataDir)
+	cfg := changedFileConfig(dataDir, dbPath)
+
+	appendRandomBytes(t, fs, files, changedPath, appendedFileSize)
+
+	store, err := storage.NewFileStorer(filepath.Join(tempDir, "remote"))
+	require.NoError(t, err)
+
+	db, err := database.New(ctx, dbPath)
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = db.Close() })
+
+	repos := database.NewRepositories(db)
+	v := newBackupVaultik(ctx, cfg, store, repos, db, fs)
+
+	require.NoError(t, backUp(v, "first"))
+
+	firstFiles := maps.Clone(files)
+	firstID := localSnapshotID(ctx, t, repos, "first")
+
+	appendRandomBytes(t, fs, files, changedPath, faultChunkSize)
+	require.NoError(t, backUp(v, "second"))
+
+	secondID := localSnapshotID(ctx, t, repos, "second")
+	secondAdded := sizeOfNewChunks(t, files, firstFiles)
+	require.Positive(t, secondAdded)
+	assert.Equal(t, secondAdded, listedNewChunkSizes(t, v)[secondID],
+		"the second snapshot must count the chunks the change added")
+
+	appendRandomBytes(t, fs, files, changedPath, faultChunkSize)
+	require.NoError(t, backUp(v, "third"))
+
+	sizes := listedNewChunkSizes(t, v)
+	assert.Equal(t, sizeOfNewChunks(t, firstFiles, nil), sizes[firstID],
+		"a later backup must not change the first snapshot's figure")
+	assert.Equal(t, secondAdded, sizes[secondID],
+		"a later backup must not change the second snapshot's figure")
+}
+
+// sizeOfNewChunks returns the total size of the distinct chunks of the
+// files in after that no file in before has, cut the way the backup cuts
+// them.
+func sizeOfNewChunks(t *testing.T, after, before map[string][]byte) int64 {
+	t.Helper()
+
+	known := make(map[string]bool)
+	for _, chunk := range chunksOf(t, before) {
+		known[chunk.Hash] = true
+	}
+
+	var total int64
+
+	for _, chunk := range chunksOf(t, after) {
+		if !known[chunk.Hash] {
+			known[chunk.Hash] = true
+			total += chunk.Size
+		}
+	}
+
+	return total
+}
+
+// chunksOf returns the chunks of every file in files.
+func chunksOf(t *testing.T, files map[string][]byte) []chunker.Chunk {
+	t.Helper()
+
+	c := chunker.NewChunker(faultChunkSize)
+
+	var all []chunker.Chunk
+
+	for _, content := range files {
+		chunks, err := c.ChunkReader(bytes.NewReader(content))
+		require.NoError(t, err)
+
+		all = append(all, chunks...)
+	}
+
+	return all
+}
+
+// listedNewChunkSizes runs `snapshot list --json` and returns each listed
+// snapshot's new_chunk_size by snapshot ID.
+func listedNewChunkSizes(t *testing.T, v *vaultik.Vaultik) map[string]int64 {
+	t.Helper()
+
+	var out bytes.Buffer
+
+	v.Stdout = &out
+	require.NoError(t, v.ListSnapshots(true))
+
+	sizes := make(map[string]int64)
+	for _, row := range decodeListJSON(t, out.String()) {
+		sizes[row.ID] = row.NewChunkSize
+	}
+
+	return sizes
+}

internal/vaultik/restore.go

@@ -309,11 +309,22 @@ func (v *Vaultik) restoreAllFiles(
 		return nil, err
 	}
 
-	plan, err := newRestorePlan(v.ctx, repos, files, chunkToBlobMap, blobIDToHash)
+	plan, unplanned, err := newRestorePlan(
+		v.ctx, repos, files, chunkToBlobMap, blobIDToHash)
 	if err != nil {
 		return nil, fmt.Errorf("building restore plan: %w", err)
 	}
 
+	// Ranging over files rather than the map reports them in a stable order.
+	for _, file := range files {
+		if planErr, ok := unplanned[file.ID]; ok {
+			err = v.handleRestoreFileError(plan, opts, result, file, file.ID, planErr)
+			if err != nil {
+				return nil, err
+			}
+		}
+	}
+
 	filesByID, totalBytesExpected := indexRestoreFiles(files)
 
 	v.UI.Beginf("Restoring %s files (%s) to %s.",

internal/vaultik/restore_malformed_db_test.go

@@ -61,8 +61,8 @@ func craftedFile(
 }
 
 // TestRestoreShortChunkHashInNoBlob proves a file whose short chunk hash
-// has no blob_chunks row fails restore planning and the chunk write with
-// an error.
+// has no blob_chunks row is left out of the restore plan with an error,
+// and fails the chunk write with an error.
 func TestRestoreShortChunkHashInNoBlob(t *testing.T) {
 	t.Parallel()
 
@@ -77,9 +77,11 @@ func TestRestoreShortChunkHashInNoBlob(t *testing.T) {
 	chunkToBlobMap, err := v.buildChunkToBlobMap(ctx, repos)
 	require.NoError(t, err)
 
-	_, err = newRestorePlan(ctx, repos, []*database.File{file},
+	plan, unplanned, err := newRestorePlan(ctx, repos, []*database.File{file},
 		chunkToBlobMap, map[string]string{})
-	require.ErrorIs(t, err, errPlanChunkMissing)
+	require.NoError(t, err)
+	require.ErrorIs(t, unplanned[file.ID], errPlanChunkMissing)
+	require.False(t, plan.hasPending())
 
 	fileChunks, err := repos.FileChunks.GetByFileID(ctx, file.ID)
 	require.NoError(t, err)

internal/vaultik/restore_plan.go

@@ -41,20 +41,25 @@ type restorePlan struct {
 
 // newRestorePlan builds the file→blob index for the given files. Files
 // whose chunks reference no blobs (symlinks, directories) start in the
-// ready queue immediately.
+// ready queue immediately. A file with a chunk that no blob holds cannot
+// be restored: it is left out of the plan and returned, by ID, with the
+// reason.
 func newRestorePlan(
 	ctx context.Context,
 	repos *database.Repositories,
 	files []*database.File,
 	chunkToBlobMap map[string]*database.BlobChunk,
 	blobIDToHash map[string]string,
-) (*restorePlan, error) {
+) (*restorePlan, map[types.FileID]error, error) {
 	p := &restorePlan{
 		fileBlobs: make(map[types.FileID]map[string]struct{}, len(files)),
 		blobFiles: make(map[string]map[types.FileID]struct{}),
 		ready:     make([]types.FileID, 0, len(files)),
 		cached:    make(map[string]struct{}),
 	}
+	unplanned := make(map[types.FileID]error)
+
+nextFile:
 	for _, f := range files {
 		if f.IsSymlink() || f.Mode&uint32(os.ModeDir) != 0 {
 			// No chunks to fetch — restore can run immediately.
@@ -66,7 +71,7 @@ func newRestorePlan(
 
 		fileChunks, err := repos.FileChunks.GetByFileID(ctx, f.ID)
 		if err != nil {
-			return nil, fmt.Errorf("planning %s: %w", f.Path, err)
+			return nil, nil, fmt.Errorf("planning %s: %w", f.Path, err)
 		}
 
 		blobs := make(map[string]struct{})
@@ -74,14 +79,18 @@ func newRestorePlan(
 		for _, fc := range fileChunks {
 			bc, ok := chunkToBlobMap[fc.ChunkHash.String()]
 			if !ok {
-				return nil, fmt.Errorf("planning %s: %w: %s",
-					f.Path, errPlanChunkMissing, shortHash(fc.ChunkHash.String()))
+				unplanned[f.ID] = fmt.Errorf("%w: %s",
+					errPlanChunkMissing, shortHash(fc.ChunkHash.String()))
+
+				continue nextFile
 			}
 
 			hash, ok := blobIDToHash[bc.BlobID.String()]
 			if !ok {
-				return nil, fmt.Errorf("planning %s: %w: %s",
-					f.Path, errPlanBlobIDMissing, bc.BlobID)
+				unplanned[f.ID] = fmt.Errorf("%w: %s",
+					errPlanBlobIDMissing, bc.BlobID)
+
+				continue nextFile
 			}
 
 			blobs[hash] = struct{}{}
@@ -103,7 +112,7 @@ func newRestorePlan(
 		}
 	}
 
-	return p, nil
+	return p, unplanned, nil
 }
 
 // markBlobCached records that the named blob is now resident in the

internal/vaultik/restore_skip_errors_test.go

@@ -28,7 +28,8 @@ const (
 )
 
 // missingBlobBackup is a snapshot of single-chunk files spread over two
-// blobs, with one of those blobs deleted from the store.
+// blobs, with one of those blobs lost: deleted from the store, or missing
+// from the snapshot metadata.
 type missingBlobBackup struct {
 	fs         afero.Fs
 	cfg        *config.Config
@@ -37,20 +38,39 @@ type missingBlobBackup struct {
 	restoreDir string
 	// files holds the original content by source path.
 	files map[string][]byte
-	// lost holds the source paths whose chunk is in the deleted blob.
+	// lost holds the source paths whose chunk is in the lost blob.
 	lost map[string]bool
 }
 
 // TestRestoreSkipErrorsSkipsFilesOfMissingBlob restores with SkipErrors
-// after one blob of a two-blob snapshot was deleted. Every file stored in
-// that blob must be reported as failed and left absent, every other file
-// must be restored intact, and Restore must still return an error.
+// after one blob of a two-blob snapshot was deleted from the store.
 func TestRestoreSkipErrorsSkipsFilesOfMissingBlob(t *testing.T) {
 	log.Initialize(log.Config{})
 	t.Parallel()
 
 	ctx := context.Background()
-	backup := backupThenDeleteOneBlob(ctx, t)
+	assertRestoreSkipsLostFiles(ctx, t, backupThenDeleteOneBlob(ctx, t))
+}
+
+// TestRestoreSkipErrorsSkipsFilesWithChunkInNoBlob restores with
+// SkipErrors a snapshot whose metadata lists chunks that no blob holds, as
+// earlier backup defects left behind.
+func TestRestoreSkipErrorsSkipsFilesWithChunkInNoBlob(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	ctx := context.Background()
+	assertRestoreSkipsLostFiles(ctx, t, backupThenDropOneBlobFromMetadata(ctx, t))
+}
+
+// assertRestoreSkipsLostFiles restores backup with SkipErrors. Every file
+// stored in the lost blob must be reported as failed and left absent,
+// every other file must be restored intact, and Restore must still return
+// an error.
+func assertRestoreSkipsLostFiles(
+	ctx context.Context, t *testing.T, backup *missingBlobBackup,
+) {
+	t.Helper()
 
 	var out bytes.Buffer
 
@@ -72,14 +92,14 @@ func TestRestoreSkipErrorsSkipsFilesOfMissingBlob(t *testing.T) {
 
 		if backup.lost[path] {
 			assert.Containsf(t, out.String(), path,
-				"%s needs the deleted blob and must be reported", path)
+				"%s needs the lost blob and must be reported", path)
 			assert.NoFileExists(t, restored)
 
 			continue
 		}
 
 		got, err := afero.ReadFile(backup.fs, restored)
-		require.NoErrorf(t, err, "%s does not need the deleted blob", path)
+		require.NoErrorf(t, err, "%s does not need the lost blob", path)
 		assert.Equalf(t, content, got, "%s restored with wrong content", path)
 	}
 }
@@ -103,6 +123,30 @@ func TestRestoreMissingBlobAbortsWithoutSkipErrors(t *testing.T) {
 	require.ErrorIs(t, err, storage.ErrNotFound)
 }
 
+// TestRestoreChunkInNoBlobAbortsWithoutSkipErrors checks that a chunk held
+// by no blob still ends the restore, before any file is written, when
+// SkipErrors is not set.
+func TestRestoreChunkInNoBlobAbortsWithoutSkipErrors(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	ctx := context.Background()
+	backup := backupThenDropOneBlobFromMetadata(ctx, t)
+
+	v := newReaderVaultik(ctx, backup.cfg, backup.storer, nil, backup.fs)
+
+	err := v.Restore(&vaultik.RestoreOptions{
+		SnapshotID: backup.snapshotID,
+		TargetDir:  backup.restoreDir,
+	})
+
+	require.ErrorContains(t, err, "chunk missing from blob map")
+
+	for path := range backup.files {
+		assert.NoFileExists(t, filepath.Join(backup.restoreDir, path))
+	}
+}
+
 // backupThenDeleteOneBlob backs up missingBlobFileCount single-chunk
 // files, then deletes from the store the blob holding the first of them.
 func backupThenDeleteOneBlob(
@@ -110,6 +154,54 @@ func backupThenDeleteOneBlob(
 ) *missingBlobBackup {
 	t.Helper()
 
+	backup, db, lostBlob := backupSingleChunkFiles(ctx, t)
+	require.NoError(t, db.Close())
+
+	require.NoError(t, backup.storer.Delete(ctx, fmt.Sprintf(
+		"blobs/%s/%s/%s", lostBlob[:2], lostBlob[2:4], lostBlob)))
+
+	return backup
+}
+
+// backupThenDropOneBlobFromMetadata backs up missingBlobFileCount
+// single-chunk files, then removes the blob holding the first of them from
+// the snapshot and exports the snapshot metadata again. The exported
+// metadata then lists chunks that no blob holds.
+func backupThenDropOneBlobFromMetadata(
+	ctx context.Context, t *testing.T,
+) *missingBlobBackup {
+	t.Helper()
+
+	backup, db, lostBlob := backupSingleChunkFiles(ctx, t)
+
+	// The uploads row refers to the blob row, so it would keep the export
+	// from deleting it.
+	for _, query := range []string{
+		"DELETE FROM snapshot_blobs WHERE blob_hash = ?",
+		"DELETE FROM uploads WHERE blob_hash = ?",
+	} {
+		_, err := db.Conn().ExecContext(ctx, query, lostBlob)
+		require.NoError(t, err)
+	}
+
+	sm := newFaultSnapshotManager(
+		backup.fs, backup.storer, backup.cfg, database.NewRepositories(db))
+	require.NoError(t,
+		sm.ExportSnapshotMetadata(ctx, db.Path(), backup.snapshotID))
+	require.NoError(t, db.Close())
+
+	return backup
+}
+
+// backupSingleChunkFiles backs up missingBlobFileCount single-chunk files.
+// It returns the backup, with lost naming the files stored in the same
+// blob as the first of them, the still open local index, and the hash of
+// that blob.
+func backupSingleChunkFiles(
+	ctx context.Context, t *testing.T,
+) (*missingBlobBackup, *database.DB, string) {
+	t.Helper()
+
 	fs := afero.NewOsFs()
 	tempDir := t.TempDir()
 	dataDir := filepath.Join(tempDir, "src")
@@ -143,16 +235,11 @@ func backupThenDeleteOneBlob(
 		blobOfFile[path] = blobHashOfSingleChunkFile(ctx, t, repos, path)
 	}
 
-	require.NoError(t, db.Close())
-
-	deleted := blobOfFile[filepath.Join(dataDir, "file-00.bin")]
-	require.NoError(t, storer.Delete(ctx, fmt.Sprintf(
-		"blobs/%s/%s/%s", deleted[:2], deleted[2:4], deleted)))
-
+	lostBlob := blobOfFile[filepath.Join(dataDir, "file-00.bin")]
 	lost := make(map[string]bool)
 
 	for path, hash := range blobOfFile {
-		if hash == deleted {
+		if hash == lostBlob {
 			lost[path] = true
 		}
 	}
@@ -168,7 +255,7 @@ func backupThenDeleteOneBlob(
 		restoreDir: filepath.Join(tempDir, "restored"),
 		files:      files,
 		lost:       lost,
-	}
+	}, db, lostBlob
 }
 
 // blobHashOfSingleChunkFile returns the hash of the blob holding the one

internal/vaultik/snapshot.go

@@ -209,6 +209,7 @@ type snapshotStats struct {
 	totalFiles         int
 	totalBytes         int64
 	totalChunks        int
+	newChunkSize       int64
 	totalBlobs         int
 	totalBytesSkipped  int64
 	totalFilesSkipped  int
@@ -341,6 +342,7 @@ func (v *Vaultik) scanAllDirectories(
 		stats.totalFiles += result.FilesScanned
 		stats.totalBytes += result.BytesScanned
 		stats.totalChunks += result.ChunksCreated
+		stats.newChunkSize += result.NewChunkSize
 		stats.totalBlobs += result.BlobsCreated
 		stats.totalFilesSkipped += result.FilesSkipped
 		stats.totalFilesFailed += result.FilesFailed
@@ -403,6 +405,7 @@ func (v *Vaultik) finalizeSnapshotMetadata(
 		BlobUncompressedSize: stats.blobUncompressedSize,
 		CompressionLevel:     v.Config.CompressionLevel,
 		UploadDurationMs:     stats.uploadDuration.Milliseconds(),
+		NewChunkSize:         stats.newChunkSize,
 	}
 
 	err = v.SnapshotManager.UpdateSnapshotStatsExtended(v.ctx, snapshotID, extStats)

internal/vaultik/snapshot_list_test.go

@@ -487,6 +487,7 @@ type listJSONRow struct {
 	RemoteKey      string `json:"remote_key"`
 	Timestamp      string `json:"timestamp"`
 	CompressedSize int64  `json:"compressed_size"`
+	NewChunkSize   int64  `json:"new_chunk_size"`
 	LocallyTracked bool   `json:"locally_tracked"`
 	RemotePresent  *bool  `json:"remote_present"`
 }