Add negative and boundary tests for blobgen and types (closes #170)

Test-only. internal/blobgen and internal/types had no negative or boundary coverage. Adds, in package blobgen_test: Writer-to-Reader round trips at the 64 KiB age-segment edges for random and compressible data, checking plaintext, byte counts and the reader/writer hashes by decrypting; a wrong-identity open; truncation and single-byte corruption of a multi-segment blob at every region; trailing bytes, empty input and garbage; rejected and accepted compression levels; nil, empty and invalid recipients; and a failing destination. In package types_test: Value/Scan round trips, NULL, wrong-type and malformed Scan, Parse and IsZero for FileID and BlobID.

The "cut right after the age header and nonce" truncation is excluded: it reads as valid and empty today and belongs to #152.

Model: opus-4-8

internal/blobgen/helpers_test.go

@@ -0,0 +1,119 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"crypto/rand"
+	"errors"
+	"io"
+	"testing"
+
+	"filippo.io/age"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// ageChunkSize is age's STREAM plaintext chunk size (64 KiB); each encrypted
+// chunk adds a 16-byte ChaCha20-Poly1305 tag.
+const (
+	ageChunkSize    = 64 * 1024
+	ageChunkTagSize = 16
+	ageSegmentSize  = ageChunkSize + ageChunkTagSize
+	ageNonceSize    = 16
+)
+
+// makeIdentity returns a fresh X25519 identity and its recipient string.
+func makeIdentity(t *testing.T) (*age.X25519Identity, string) {
+	t.Helper()
+
+	id, err := age.GenerateX25519Identity()
+	require.NoError(t, err)
+
+	return id, id.Recipient().String()
+}
+
+// randomBytes returns n cryptographically random bytes, which do not compress
+// so the encrypted payload spans multiple age segments.
+func randomBytes(t *testing.T, n int) []byte {
+	t.Helper()
+
+	b := make([]byte, n)
+	_, err := rand.Read(b)
+	require.NoError(t, err)
+
+	return b
+}
+
+// compressibleBytes returns n bytes of a repeating pattern, which zstd packs
+// down to a small payload.
+func compressibleBytes(n int) []byte {
+	pattern := bytes.Repeat([]byte("compressible-"), n/13+1)
+
+	return pattern[:n]
+}
+
+// encryptBlob compresses, encrypts and returns a blob for plaintext at
+// compression level 1.
+func encryptBlob(t *testing.T, plaintext []byte, recipients ...string) []byte {
+	t.Helper()
+
+	var buf bytes.Buffer
+
+	w, err := blobgen.NewWriter(&buf, 1, recipients)
+	require.NoError(t, err)
+
+	_, err = w.Write(plaintext)
+	require.NoError(t, err)
+	require.NoError(t, w.Close())
+
+	return buf.Bytes()
+}
+
+// ageHeaderLen returns the byte length of blob's age header, i.e. the offset
+// of the 16-byte payload nonce that follows it. The header ends with a MAC
+// line "--- <mac>\n"; the nonce begins right after that newline.
+func ageHeaderLen(t *testing.T, blob []byte) int {
+	t.Helper()
+
+	i := bytes.Index(blob, []byte("\n--- "))
+	require.GreaterOrEqual(t, i, 0, "age MAC footer line not found")
+
+	nl := bytes.IndexByte(blob[i+1:], '\n')
+	require.GreaterOrEqual(t, nl, 0, "newline ending MAC line not found")
+
+	return i + 1 + nl + 1
+}
+
+// requireBlobUnreadable asserts that data never decrypts to a plaintext with a
+// nil error: either NewReader fails, or reading it does.
+func requireBlobUnreadable(t *testing.T, data []byte, id age.Identity) {
+	t.Helper()
+
+	r, err := blobgen.NewReader(bytes.NewReader(data), id)
+	if err != nil {
+		return
+	}
+
+	_, err = io.ReadAll(r)
+	_ = r.Close()
+
+	require.Error(t, err, "reading a damaged blob must fail")
+}
+
+// errFailWriter is returned by failAfterWriter once its byte limit is passed.
+var errFailWriter = errors.New("destination write failed")
+
+// failAfterWriter accepts writes until more than limit bytes have been sent,
+// then fails every write. It models a destination that dies mid-blob.
+type failAfterWriter struct {
+	limit   int
+	written int
+}
+
+func (f *failAfterWriter) Write(p []byte) (int, error) {
+	f.written += len(p)
+	if f.written > f.limit {
+		return 0, errFailWriter
+	}
+
+	return len(p), nil
+}

internal/blobgen/negative_test.go

@@ -0,0 +1,190 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"fmt"
+	"testing"
+
+	"filippo.io/age"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// TestNewReaderWrongIdentity covers issue case 4: opening a blob with an
+// identity other than the recipient reports no matching identity.
+func TestNewReaderWrongIdentity(t *testing.T) {
+	t.Parallel()
+
+	_, recipient := makeIdentity(t)
+	other, _ := makeIdentity(t)
+
+	blob := encryptBlob(t, []byte("secret payload"), recipient)
+
+	_, err := blobgen.NewReader(bytes.NewReader(blob), other)
+	require.Error(t, err)
+
+	var noMatch *age.NoIdentityMatchError
+	assert.ErrorAs(t, err, &noMatch)
+}
+
+// TestNewReaderTruncated covers issue case 6: a multi-segment blob cut at
+// several points must never read back as valid data. The point immediately
+// after the header and nonce is intentionally excluded: it reads as a valid
+// empty blob today and is the regression case for
+// https://git.eeqj.de/sneak/vaultik/issues/152.
+func TestNewReaderTruncated(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+	blob := encryptBlob(t, randomBytes(t, 4*65536+123), recipient)
+	h := ageHeaderLen(t, blob)
+
+	require.Greater(t, len(blob), h+ageNonceSize+ageSegmentSize,
+		"test needs a blob of at least two age segments")
+
+	cases := []struct {
+		name string
+		size int
+	}{
+		{"inside header", h / 2},
+		{"inside nonce", h + 8},
+		{"inside first segment", h + ageNonceSize + 100},
+		{"end of first full segment", h + ageNonceSize + ageSegmentSize},
+		{"last byte removed", len(blob) - 1},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			requireBlobUnreadable(t, blob[:tc.size], id)
+		})
+	}
+}
+
+// TestNewReaderCorrupted covers issue case 7: one flipped byte in each region
+// of a multi-segment blob makes it unreadable.
+func TestNewReaderCorrupted(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+	blob := encryptBlob(t, randomBytes(t, 4*65536+123), recipient)
+	h := ageHeaderLen(t, blob)
+
+	firstNL := bytes.IndexByte(blob, '\n')
+	require.Positive(t, firstNL, "header must have a version line")
+
+	cases := []struct {
+		name string
+		pos  int
+	}{
+		{"header stanza", firstNL + 5},
+		{"header MAC line", h - 2},
+		{"nonce", h + 4},
+		{"body segment", h + ageNonceSize + 50},
+		{"final tag", len(blob) - 1},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			corrupt := append([]byte(nil), blob...)
+			corrupt[tc.pos] ^= 0xff
+			requireBlobUnreadable(t, corrupt, id)
+		})
+	}
+}
+
+// TestNewReaderTrailingAndGarbage covers issue case 8: bytes appended after a
+// valid blob, empty input, and random garbage each fail to read.
+func TestNewReaderTrailingAndGarbage(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+
+	valid := encryptBlob(t, []byte("small payload"), recipient)
+	appended := append(append([]byte(nil), valid...), []byte("trailing junk")...)
+
+	t.Run("appended bytes", func(t *testing.T) {
+		t.Parallel()
+		requireBlobUnreadable(t, appended, id)
+	})
+	t.Run("empty input", func(t *testing.T) {
+		t.Parallel()
+		requireBlobUnreadable(t, []byte{}, id)
+	})
+	t.Run("random garbage", func(t *testing.T) {
+		t.Parallel()
+		requireBlobUnreadable(t, randomBytes(t, 512), id)
+	})
+}
+
+// TestNewWriterInvalidLevel covers the rejected end of issue case 9: an
+// out-of-range compression level errors and writes nothing to the destination.
+func TestNewWriterInvalidLevel(t *testing.T) {
+	t.Parallel()
+
+	_, recipient := makeIdentity(t)
+
+	for _, level := range []int{0, -1, 20} {
+		t.Run(fmt.Sprintf("level%d", level), func(t *testing.T) {
+			t.Parallel()
+
+			var buf bytes.Buffer
+
+			w, err := blobgen.NewWriter(&buf, level, []string{recipient})
+			require.ErrorIs(t, err, blobgen.ErrInvalidCompressionLevel)
+			assert.Nil(t, w)
+			assert.Zero(t, buf.Len(), "nothing written on an invalid level")
+		})
+	}
+}
+
+// TestNewWriterInvalidRecipients covers issue case 10: nil and empty recipient
+// lists and an unparsable recipient string each error.
+func TestNewWriterInvalidRecipients(t *testing.T) {
+	t.Parallel()
+
+	cases := []struct {
+		name       string
+		recipients []string
+	}{
+		{"nil list", nil},
+		{"empty list", []string{}},
+		{"invalid recipient string", []string{"not-a-recipient"}},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			var buf bytes.Buffer
+
+			w, err := blobgen.NewWriter(&buf, 1, tc.recipients)
+			require.Error(t, err)
+			assert.Nil(t, w)
+		})
+	}
+}
+
+// TestNewWriterFailingDestination covers issue case 11: a destination that
+// fails mid-blob surfaces its error from Write or Close.
+func TestNewWriterFailingDestination(t *testing.T) {
+	t.Parallel()
+
+	_, recipient := makeIdentity(t)
+
+	// The limit clears the age header and nonce so NewWriter succeeds, then
+	// trips once the compressed body starts flowing.
+	dst := &failAfterWriter{limit: 512}
+
+	w, err := blobgen.NewWriter(dst, 1, []string{recipient})
+	require.NoError(t, err)
+
+	_, writeErr := w.Write(randomBytes(t, 256*1024))
+	closeErr := w.Close()
+
+	assert.True(t, writeErr != nil || closeErr != nil,
+		"destination failure must surface from Write or Close")
+}

internal/blobgen/roundtrip_test.go

@@ -0,0 +1,132 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"crypto/sha256"
+	"fmt"
+	"io"
+	"testing"
+
+	"filippo.io/age"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// checkRoundTrip writes input through a Writer, reads it back through a Reader,
+// and verifies the plaintext, the byte counts, and the content hashes.
+func checkRoundTrip(
+	t *testing.T, id *age.X25519Identity, recipient string,
+	level int, input []byte,
+) {
+	t.Helper()
+
+	var buf bytes.Buffer
+
+	w, err := blobgen.NewWriter(&buf, level, []string{recipient})
+	require.NoError(t, err)
+
+	n, err := w.Write(input)
+	require.NoError(t, err)
+	assert.Equal(t, len(input), n)
+	require.NoError(t, w.Close())
+	require.Equal(t, int64(len(input)), w.BytesWritten())
+
+	r, err := blobgen.NewReader(bytes.NewReader(buf.Bytes()), id)
+	require.NoError(t, err)
+
+	got, err := io.ReadAll(r)
+	require.NoError(t, err)
+	require.NoError(t, r.Close())
+
+	assert.Equal(t, input, got, "decrypted output must equal input")
+	require.Equal(t, int64(len(input)), r.BytesRead())
+
+	// The hash values are checked by decrypting: the reader's single SHA-256
+	// is the hash of the plaintext, and hashing it once more (DoubleSHA256)
+	// gives the writer's ContentID.
+	single := sha256.Sum256(got)
+	assert.Equal(t, single[:], r.Sum256())
+	assert.Equal(t, blobgen.DoubleSHA256(r.Sum256()), w.ContentID())
+}
+
+// TestWriterReaderRoundTrip covers issue cases 1 and 2: every size round trips
+// for both random and compressible data, and the reader hash, its double hash
+// and the byte counts all agree.
+func TestWriterReaderRoundTrip(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+
+	// Sizes exercise the age segment boundary (64 KiB) from just below to a
+	// few segments above it, plus the empty and single-byte edges.
+	sizes := []int{0, 1, 65535, 65536, 65537, 4*65536 + 123}
+
+	kinds := []struct {
+		name string
+		fill func(*testing.T, int) []byte
+	}{
+		{"random", randomBytes},
+		{"compressible", func(_ *testing.T, n int) []byte {
+			return compressibleBytes(n)
+		}},
+	}
+
+	for _, k := range kinds {
+		for _, size := range sizes {
+			name := fmt.Sprintf("%s/%d", k.name, size)
+			t.Run(name, func(t *testing.T) {
+				t.Parallel()
+				checkRoundTrip(t, id, recipient, 1, k.fill(t, size))
+			})
+		}
+	}
+}
+
+// TestZeroLengthNoWrite covers issue case 3: a Writer closed with no Write at
+// all produces the double hash of the empty input, and the blob reads back as
+// empty with no error.
+func TestZeroLengthNoWrite(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+
+	var buf bytes.Buffer
+
+	w, err := blobgen.NewWriter(&buf, 1, []string{recipient})
+	require.NoError(t, err)
+	require.NoError(t, w.Close())
+	assert.Equal(t, int64(0), w.BytesWritten())
+
+	empty := sha256.Sum256(nil)
+	doubled := sha256.Sum256(empty[:])
+	assert.Equal(t, doubled[:], w.ContentID(),
+		"ContentID of empty input is SHA256(SHA256(\"\"))")
+
+	r, err := blobgen.NewReader(bytes.NewReader(buf.Bytes()), id)
+	require.NoError(t, err)
+
+	got, err := io.ReadAll(r)
+	require.NoError(t, err)
+	require.NoError(t, r.Close())
+
+	assert.Empty(t, got, "empty blob decrypts to empty output")
+	assert.Equal(t, int64(0), r.BytesRead())
+	assert.Equal(t, empty[:], r.Sum256())
+}
+
+// TestNewWriterValidLevelsRoundTrip covers the accepted end of issue case 9:
+// the boundary compression levels 1 and 19 both round trip.
+func TestNewWriterValidLevelsRoundTrip(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+	input := randomBytes(t, 4096)
+
+	for _, level := range []int{1, 19} {
+		t.Run(fmt.Sprintf("level%d", level), func(t *testing.T) {
+			t.Parallel()
+			checkRoundTrip(t, id, recipient, level, input)
+		})
+	}
+}

internal/blobgen/writer_test.go

@@ -13,8 +13,9 @@ import (
 )
 
 // TestWriterHashIsDoubleHash verifies that Writer.ContentID() returns
-// the double hash SHA256(SHA256(plaintext)) for security.
-// Double hashing prevents attackers from confirming existence of known content.
+// SHA256(SHA256(plaintext)). Stored objects are named by this second hash so a
+// name is not the plaintext's own SHA-256; this does not stop someone who
+// already holds the plaintext from confirming it.
 func TestWriterHashIsDoubleHash(t *testing.T) {
 	t.Parallel()
 
@@ -60,11 +61,11 @@ func TestWriterHashIsDoubleHash(t *testing.T) {
 
 	// The writer hash should match the double hash
 	assert.Equal(t, expectedDoubleHash, writerHash,
-		"Writer.ContentID() should return SHA256(SHA256(plaintext)) for security")
+		"Writer.ContentID() must be SHA256(SHA256(plaintext))")
 
-	// Verify it's NOT the single hash (would leak information)
+	// It must be the second hash, not the plaintext's own SHA-256.
 	assert.NotEqual(t, singleHashStr, writerHash,
-		"Writer hash should not be single hash (would allow content confirmation attacks)")
+		"Writer hash must be the double hash, not the single SHA-256")
 }
 
 // TestWriterDeterministicHash verifies that the same input always produces

internal/types/types_test.go

@@ -0,0 +1,136 @@
+package types_test
+
+import (
+	"database/sql"
+	"database/sql/driver"
+	"fmt"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/types"
+)
+
+// scannableID is the shared behaviour of the UUID-backed id types. A pointer
+// to FileID or BlobID satisfies it, so both are tested through one set of
+// cases.
+type scannableID interface {
+	driver.Valuer
+	sql.Scanner
+	fmt.Stringer
+	IsZero() bool
+}
+
+// idKind adapts one id type to the generic tests below.
+type idKind struct {
+	name      string
+	newZero   func() scannableID
+	newRandom func() scannableID
+	parse     func(string) (scannableID, error)
+}
+
+func idKinds() []idKind {
+	return []idKind{
+		{
+			name:    "FileID",
+			newZero: func() scannableID { return &types.FileID{} },
+			newRandom: func() scannableID {
+				id := types.NewFileID()
+
+				return &id
+			},
+			parse: func(s string) (scannableID, error) {
+				id, err := types.ParseFileID(s)
+
+				return &id, err
+			},
+		},
+		{
+			name:    "BlobID",
+			newZero: func() scannableID { return &types.BlobID{} },
+			newRandom: func() scannableID {
+				id := types.NewBlobID()
+
+				return &id
+			},
+			parse: func(s string) (scannableID, error) {
+				id, err := types.ParseBlobID(s)
+
+				return &id, err
+			},
+		},
+	}
+}
+
+// TestIDValueScan checks that Value then Scan round trips from both a string
+// and a []byte, that a NULL scans to the zero id, and that a non-string type
+// and malformed text are rejected.
+func TestIDValueScan(t *testing.T) {
+	t.Parallel()
+
+	for _, k := range idKinds() {
+		t.Run(k.name, func(t *testing.T) {
+			t.Parallel()
+
+			orig := k.newRandom()
+			v, err := orig.Value()
+			require.NoError(t, err)
+
+			s, ok := v.(string)
+			require.True(t, ok, "Value must yield a string")
+
+			fromString := k.newZero()
+			require.NoError(t, fromString.Scan(s))
+			assert.Equal(t, orig.String(), fromString.String())
+			assert.False(t, fromString.IsZero())
+
+			fromBytes := k.newZero()
+			require.NoError(t, fromBytes.Scan([]byte(s)))
+			assert.Equal(t, orig.String(), fromBytes.String())
+
+			nulled := k.newRandom()
+			require.NoError(t, nulled.Scan(nil))
+			assert.True(t, nulled.IsZero(), "NULL scans to the zero id")
+
+			require.Error(t, k.newZero().Scan(42),
+				"a non-string type must be rejected")
+			assert.Error(t, k.newZero().Scan("not-a-uuid"),
+				"malformed text must be rejected")
+		})
+	}
+}
+
+// TestIDParse checks that the Parse function accepts a canonical id and
+// rejects malformed text.
+func TestIDParse(t *testing.T) {
+	t.Parallel()
+
+	for _, k := range idKinds() {
+		t.Run(k.name, func(t *testing.T) {
+			t.Parallel()
+
+			canonical := k.newRandom().String()
+
+			parsed, err := k.parse(canonical)
+			require.NoError(t, err)
+			assert.Equal(t, canonical, parsed.String())
+
+			_, err = k.parse("not-a-uuid")
+			assert.Error(t, err)
+		})
+	}
+}
+
+// TestIDIsZero checks IsZero on the zero and on a freshly generated id.
+func TestIDIsZero(t *testing.T) {
+	t.Parallel()
+
+	for _, k := range idKinds() {
+		t.Run(k.name, func(t *testing.T) {
+			t.Parallel()
+
+			assert.True(t, k.newZero().IsZero())
+			assert.False(t, k.newRandom().IsZero())
+		})
+	}
+}