Move asset verification into the artifact package

internal/artifact/assets.go

@@ -0,0 +1,141 @@
+package artifact
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+	"path"
+	"path/filepath"
+	"sort"
+	"strings"
+)
+
+var (
+	// ErrInvalidAsset identifies an invalid asset declaration or unreadable
+	// local path.
+	ErrInvalidAsset = errors.New("invalid asset")
+	// ErrAssetConflict identifies local content that does not match its
+	// declared identity.
+	ErrAssetConflict = errors.New("asset content conflict")
+)
+
+// ExpectedAsset describes the local content declared for hosted delivery.
+// Content is populated for inline assets such as the generated checksum
+// manifest; Path identifies file content for later streaming uploads.
+type ExpectedAsset struct {
+	Filename string
+	SHA256   string
+	Size     int64
+	Path     string
+	Content  []byte
+}
+
+// BuildManifest creates the deterministic checksum file for the declared
+// binaries. An empty binary set produces no manifest.
+func BuildManifest(binaries []ExpectedAsset) (ExpectedAsset, error) {
+	if len(binaries) == 0 {
+		return ExpectedAsset{}, nil
+	}
+	ordered := append([]ExpectedAsset(nil), binaries...)
+	sort.Slice(ordered, func(i, j int) bool { return ordered[i].Filename < ordered[j].Filename })
+	var body strings.Builder
+	seen := make(map[string]struct{}, len(ordered))
+	for _, binary := range ordered {
+		if !validAssetName(binary.Filename) || binary.Filename == ManifestFilename || !validSHA256(binary.SHA256) {
+			return ExpectedAsset{}, fmt.Errorf("%w: invalid binary identity for %q", ErrInvalidAsset, binary.Filename)
+		}
+		if _, exists := seen[binary.Filename]; exists {
+			return ExpectedAsset{}, fmt.Errorf("%w: duplicate binary filename %q", ErrInvalidAsset, binary.Filename)
+		}
+		seen[binary.Filename] = struct{}{}
+		fmt.Fprintf(&body, "%s  %s\n", binary.SHA256, binary.Filename)
+	}
+	content := []byte(body.String())
+	return ExpectedAsset{Filename: ManifestFilename, SHA256: digest(content), Size: int64(len(content)), Content: content}, nil
+}
+
+// ValidateExpectedAssets checks the declarations before they are compared
+// with remote content or uploaded.
+func ValidateExpectedAssets(expected []ExpectedAsset) error {
+	seen := make(map[string]struct{}, len(expected))
+	for _, item := range expected {
+		if !validAssetName(item.Filename) || !validSHA256(item.SHA256) || item.Size < 0 {
+			return fmt.Errorf("%w: %q", ErrInvalidAsset, item.Filename)
+		}
+		if _, duplicate := seen[item.Filename]; duplicate {
+			return fmt.Errorf("%w: duplicate expected asset %q", ErrInvalidAsset, item.Filename)
+		}
+		seen[item.Filename] = struct{}{}
+		if len(item.Content) > 0 && (int64(len(item.Content)) != item.Size || digest(item.Content) != item.SHA256) {
+			return fmt.Errorf("%w: expected content does not match identity for %q", ErrInvalidAsset, item.Filename)
+		}
+	}
+	return nil
+}
+
+// VerifyLocalAssets checks that every expected file still matches its declared
+// size and digest before hosted writes begin.
+func VerifyLocalAssets(assets []ExpectedAsset) error {
+	if err := ValidateExpectedAssets(assets); err != nil {
+		return err
+	}
+	for _, asset := range assets {
+		if err := verifyLocalAsset(asset); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func verifyLocalAsset(asset ExpectedAsset) error {
+	if len(asset.Content) > 0 {
+		if int64(len(asset.Content)) != asset.Size || digest(asset.Content) != asset.SHA256 {
+			return fmt.Errorf("%w: local content does not match declared asset %q", ErrAssetConflict, asset.Filename)
+		}
+		return nil
+	}
+	if asset.Path == "" || filepath.Base(asset.Path) != asset.Filename {
+		return fmt.Errorf("%w: local path for %q does not match its filename", ErrInvalidAsset, asset.Filename)
+	}
+	file, err := os.Open(asset.Path)
+	if err != nil {
+		return fmt.Errorf("%w: open %q: %w", ErrInvalidAsset, asset.Filename, err)
+	}
+	defer file.Close()
+	info, err := file.Stat()
+	if err != nil {
+		return fmt.Errorf("%w: stat %q: %w", ErrInvalidAsset, asset.Filename, err)
+	}
+	if !info.Mode().IsRegular() || info.Size() != asset.Size {
+		return fmt.Errorf("%w: local file does not match declared size for %q", ErrAssetConflict, asset.Filename)
+	}
+	hasher := sha256.New()
+	size, err := io.Copy(hasher, file)
+	if err != nil {
+		return fmt.Errorf("%w: read %q: %w", ErrInvalidAsset, asset.Filename, err)
+	}
+	if size != asset.Size || hex.EncodeToString(hasher.Sum(nil)) != asset.SHA256 {
+		return fmt.Errorf("%w: local content does not match declared asset %q", ErrAssetConflict, asset.Filename)
+	}
+	return nil
+}
+
+func validAssetName(name string) bool {
+	return name != "" && name != "." && name != ".." && path.Base(name) == name && !strings.ContainsAny(name, "\\\r\n\x00")
+}
+
+func validSHA256(value string) bool {
+	if len(value) != sha256.Size*2 {
+		return false
+	}
+	_, err := hex.DecodeString(value)
+	return err == nil && strings.ToLower(value) == value
+}
+
+func digest(content []byte) string {
+	sum := sha256.Sum256(content)
+	return hex.EncodeToString(sum[:])
+}