README.md

@@ -19,6 +19,7 @@ Crucible is how the artifact repos turn source into artifacts.
 | `crucible-depot`   | `crucible depot`   | content-addressed depot blob verify/move        |
 | `crucible-hak`     | `crucible hak`     | ERF/HAK pack/unpack + hak manifests             |
 | `crucible-module`  | `crucible module`  | build/extract/validate/compare the `.mod`       |
+| `crucible-nwsync`  | `crucible nwsync`  | NWSync blob emit + manifest assemble            |
 | `crucible-topdata` | `crucible topdata` | compile 2da/tlk topdata + packages              |
 | `crucible-wiki`    | `crucible wiki`    | render + deploy mechanical wiki pages           |
 

cmd/crucible-nwsync/main.go

@@ -0,0 +1,11 @@
+// Command crucible-nwsync is the standalone nwsync builder (equivalent to
+// `crucible nwsync`). A single-token binary keeps consumer wrapper scripts simple.
+package main
+
+import (
+	"os"
+
+	"git.westgate.pw/ShadowsOverWestgate/sow-tools/internal/dispatch"
+)
+
+func main() { os.Exit(dispatch.RunBuilder("nwsync", os.Args[1:])) }

docs/command-surface.md

@@ -34,11 +34,20 @@ aliases.
 | `depot` | `verify` | Existence sweep plus sampled download re-hash. |
 | `depot` | `get` | Fetch one blob with sha re-verify. |
 | `depot` | `pull` | Incremental verified pull of every referenced blob. |
+| `nwsync` | `emit` | Explode one artifact into NWSync blobs plus its own NSYM manifest. |
+| `nwsync` | `assemble` | Merge per-artifact NSYM manifests into one merged manifest. |
 
 `depot status` and `depot get` pick their backend either with `--out DIR`, a
 depot tree on disk, or with `--target bunny|cdn`, a remote backend. The two
 flags are mutually exclusive.
 
+`nwsync emit` runs where an artifact is born (a `.hak`/`.erf`, or a loose file
+such as the TLK); `nwsync assemble` runs at module release and reads only the
+small per-artifact manifests. `--order` lists artifact names highest priority
+first: a resref in more than one artifact resolves to the earliest one, the way
+the game resolves it. `--group-id` is per channel — 1 is current, 2 is testing,
+and 0 leaves the field out of the sidecar.
+
 ## Hidden compatibility aliases
 
 Existing scripts may continue using these names indefinitely. They are accepted

flake.nix

@@ -22,12 +22,13 @@
             pname = "crucible";
             inherit version;
             src = ./.;
-            vendorHash = "sha256-hm6mrNAtXv0LidzHUfz4eukTFZouizGtxkZ8gKJFUVI=";
+            vendorHash = "sha256-0I8j7On9YGD2GK9xbj/KkgBrlkMJ6Y6XQv+KCLTgBBU=";
             subPackages = [
               "cmd/crucible"
               "cmd/crucible-depot"
               "cmd/crucible-hak"
               "cmd/crucible-module"
+              "cmd/crucible-nwsync"
               "cmd/crucible-topdata"
               "cmd/crucible-wiki"
             ];

go.mod

@@ -6,3 +6,5 @@ require (
 	golang.org/x/text v0.35.0
 	gopkg.in/yaml.v3 v3.0.1
 )
+
+require github.com/klauspost/compress v1.19.1

go.sum

@@ -1,3 +1,5 @@
+github.com/klauspost/compress v1.19.1 h1:VsB4HPswih7mmZ8WleSFQ75c/Ui1M4trX5oAsJnhSlk=
+github.com/klauspost/compress v1.19.1/go.mod h1:cwPg85FWrGar70rWktvGQj8/hthj3wpl0PGDogxkrSQ=
 golang.org/x/text v0.35.0 h1:JOVx6vVDFokkpaq1AEptVzLTpDe9KGpj5tR4/X+ybL8=
 golang.org/x/text v0.35.0/go.mod h1:khi/HExzZJ2pGnjenulevKNX1W67CUy0AsXcNubPGCA=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=

internal/dispatch/dispatch.go

@@ -21,6 +21,7 @@ import (
 	"git.westgate.pw/ShadowsOverWestgate/sow-tools/internal/buildinfo"
 	"git.westgate.pw/ShadowsOverWestgate/sow-tools/internal/depot"
 	"git.westgate.pw/ShadowsOverWestgate/sow-tools/internal/menu"
+	"git.westgate.pw/ShadowsOverWestgate/sow-tools/internal/nwsync"
 )
 
 // Exit codes follow the sysexits(3) convention so CI can distinguish
@@ -104,6 +105,16 @@ var Registry = []Builder{
 		},
 		Wired: true,
 	},
+	{
+		Name:    "nwsync",
+		Bin:     "crucible-nwsync",
+		Summary: "publish NWSync blobs and manifests (emit/assemble)",
+		Commands: []Command{
+			{Name: "emit", Summary: "explode one artifact into blobs plus its own NSYM manifest", Usage: "crucible nwsync emit <artifact> --out DIR"},
+			{Name: "assemble", Summary: "merge per-artifact NSYM manifests into one", Usage: "crucible nwsync assemble --order NAMES --entries DIR --out DIR [--group-id N]"},
+		},
+		Wired: true,
+	},
 	{
 		Name:    "assets",
 		Bin:     "crucible-assets",
@@ -241,7 +252,7 @@ var Registry = []Builder{
 					"2da-to-module [flags] <input.2da> [output.json]",
 					"json-to-2da <input.json> <output.2da>",
 				},
-				Aliases:    []CommandAlias{{Name: "convert-topdata"}},
+				Aliases: []CommandAlias{{Name: "convert-topdata"}},
 			},
 		},
 		Wired: true,
@@ -392,7 +403,7 @@ func runBuilder(name string, args []string, out, errw io.Writer) int {
 		}
 	}
 	if b.Wired {
-		// depot and assets are self-contained builders: they parse their own
+		// depot, assets and nwsync are self-contained builders: they parse their own
 		// subcommands and own their exit contract, bypassing the
 		// b.Commands/delegateLegacy legacy routing entirely.
 		switch b.Name {
@@ -400,6 +411,8 @@ func runBuilder(name string, args []string, out, errw io.Writer) int {
 			return depot.Run(args, out, errw, os.Getenv)
 		case "assets":
 			return assets.Run(args, out, errw, os.Getenv)
+		case "nwsync":
+			return nwsync.Run(args, out, errw)
 		}
 	}
 	if !b.Wired {

internal/dispatch/dispatch_test.go

@@ -148,6 +148,10 @@ func TestBuilderHelpIsOK(t *testing.T) {
 	}
 }
 
+// selfContained builders parse their own subcommands instead of delegating to
+// the legacy internal/app surface.
+var selfContained = map[string]bool{"depot": true, "assets": true, "nwsync": true}
+
 func TestCanonicalCommandSurface(t *testing.T) {
 	want := map[string][]string{
 		"depot":   {"status", "push", "verify", "get", "pull"},
@@ -156,6 +160,7 @@ func TestCanonicalCommandSurface(t *testing.T) {
 		"module":  {"build", "extract", "validate", "compare", "manifest"},
 		"topdata": {"validate", "build", "package", "compare", "convert"},
 		"wiki":    {"build", "deploy"},
+		"nwsync":  {"emit", "assemble"},
 	}
 	for _, builder := range Registry {
 		got := builder.subcommands()
@@ -173,10 +178,10 @@ func TestRegistryCommandNamesAndAliasesAreUnambiguous(t *testing.T) {
 	for _, builder := range Registry {
 		seen := map[string]bool{}
 		for _, command := range builder.Commands {
-			// depot and assets parse their own subcommands and bypass AppCommand
+			// depot, assets and nwsync parse their own subcommands and bypass
-			// routing entirely (see the self-contained-builder branch in
+			// AppCommand routing entirely (see the self-contained-builder branch
-			// runBuilder), so their Commands carry no AppCommand.
+			// in runBuilder), so their Commands carry no AppCommand.
-			requireAppCommand := builder.Name != "depot" && builder.Name != "assets"
+			requireAppCommand := !selfContained[builder.Name]
 			if command.Name == "" || command.Summary == "" || command.Usage == "" || (requireAppCommand && command.AppCommand == "") {
 				t.Errorf("%s has incomplete command metadata: %#v", builder.Name, command)
 			}

internal/nwsync/assemble.go

@@ -0,0 +1,111 @@
+package nwsync
+
+import (
+	"crypto/sha1"
+	"encoding/json"
+	"fmt"
+	"os"
+	"path/filepath"
+)
+
+// AssembleOptions describes one merged manifest.
+type AssembleOptions struct {
+	Order       []string // artifact names, highest priority first
+	EntriesDir  string   // directory holding <name>.nsym and <name>.nsym.json
+	OutDir      string   // repository root; the manifest lands in <out>/manifests
+	GroupID     int      // 1 = current, 2 = testing; 0 means absent
+	ModuleName  string
+	Description string
+}
+
+// AssembleResult reports what one assemble run produced.
+type AssembleResult struct {
+	SHA1         string
+	ManifestPath string
+	Entries      int
+}
+
+// Assemble merges the per-artifact NSYM manifests named by Order into one
+// manifest. It reads no bulk data at all — only the small index files.
+//
+// Merge rule is resref shadowing, not concatenation: a resref present in more
+// than one artifact resolves to the earliest artifact in Order, which is how
+// the game resolves it (upstream's resman adds haks in reverse and lets the
+// last one win). Get this backwards and the wrong texture ships silently.
+func Assemble(options AssembleOptions) (AssembleResult, error) {
+	if len(options.Order) == 0 {
+		return AssembleResult{}, fmt.Errorf("assemble: --order names no artifacts")
+	}
+
+	merged := make([]Entry, 0, 1024)
+	winner := make(map[Identity]bool, 1024)
+	var onDiskBytes int64
+
+	for _, name := range options.Order {
+		manifestPath := filepath.Join(options.EntriesDir, name+".nsym")
+		data, err := os.ReadFile(manifestPath)
+		if err != nil {
+			return AssembleResult{}, fmt.Errorf("assemble: no index for %q: %w", name, err)
+		}
+		entries, err := readManifest(data)
+		if err != nil {
+			return AssembleResult{}, fmt.Errorf("%s: %w", manifestPath, err)
+		}
+		sidecar, err := readSidecar(manifestPath + ".json")
+		if err != nil {
+			return AssembleResult{}, err
+		}
+		// Two producers of blobs means a skewed emitter can write blobs the
+		// merged manifest quietly disagrees with. Refuse to merge across
+		// mismatched emitter versions.
+		if sidecar.EmitterVersion != emitterVersion {
+			return AssembleResult{}, fmt.Errorf(
+				"assemble: emitter version mismatch: %s was emitted by emitter %q, this is emitter %q",
+				name, sidecar.EmitterVersion, emitterVersion)
+		}
+		// on_disk_bytes overcounts by the handful of cross-artifact
+		// duplicates. It is a display statistic; no dedupe pass for it.
+		onDiskBytes += sidecar.OnDiskBytes
+
+		for _, entry := range entries {
+			identity := entry.identity()
+			if winner[identity] {
+				continue
+			}
+			winner[identity] = true
+			merged = append(merged, entry)
+		}
+	}
+
+	if len(merged) == 0 {
+		return AssembleResult{}, fmt.Errorf("assemble: merged manifest is empty")
+	}
+
+	data, err := writeManifest(merged)
+	if err != nil {
+		return AssembleResult{}, err
+	}
+	sha1Hex := fmt.Sprintf("%x", sha1.Sum(data))
+	manifestPath := filepath.Join(options.OutDir, "manifests", sha1Hex)
+	sidecar := Sidecar{
+		ModuleName:  options.ModuleName,
+		Description: options.Description,
+		GroupID:     options.GroupID,
+	}
+	if err := writeManifestPair(manifestPath, data, merged, onDiskBytes, sidecar); err != nil {
+		return AssembleResult{}, err
+	}
+	return AssembleResult{SHA1: sha1Hex, ManifestPath: manifestPath, Entries: len(merged)}, nil
+}
+
+func readSidecar(path string) (Sidecar, error) {
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return Sidecar{}, fmt.Errorf("assemble: missing sidecar: %w", err)
+	}
+	var sidecar Sidecar
+	if err := json.Unmarshal(data, &sidecar); err != nil {
+		return Sidecar{}, fmt.Errorf("%s: %w", path, err)
+	}
+	return sidecar, nil
+}

internal/nwsync/compressedbuf.go

@@ -0,0 +1,76 @@
+package nwsync
+
+import (
+	"bytes"
+	"encoding/binary"
+	"fmt"
+
+	"github.com/klauspost/compress/zstd"
+)
+
+// NWCompressedBuffer framing, as upstream's neverwinter/compressedbuf.nim
+// writes it for NWSync blobs. All fields are little-endian uint32:
+//
+//	magic "NSYC", version 3, algorithm 2 (zstd), uncompressed size,
+//	zstd header version 1, dictionary 0, then the raw zstd frame.
+const (
+	blobMagic       = 0x4359534E // "NSYC" little-endian
+	blobVersion     = 3
+	algorithmZstd   = 2
+	zstdHeaderVer   = 1
+	zstdDictionary  = 0
+	blobHeaderBytes = 24
+)
+
+var (
+	blobEncoder, _ = zstd.NewWriter(nil)
+	blobDecoder, _ = zstd.NewReader(nil)
+)
+
+// compressBlob wraps data in NWCompressedBuffer framing.
+func compressBlob(data []byte) []byte {
+	var out bytes.Buffer
+	header := []uint32{blobMagic, blobVersion, algorithmZstd, uint32(len(data)), zstdHeaderVer, zstdDictionary}
+	for _, field := range header {
+		_ = binary.Write(&out, binary.LittleEndian, field)
+	}
+	out.Write(blobEncoder.EncodeAll(data, nil))
+	return out.Bytes()
+}
+
+// decompressBlob unwraps NWCompressedBuffer framing. It exists so a blob this
+// package wrote — or one upstream wrote — can be compared by its uncompressed
+// bytes, which is the only comparison that is meaningful across zstd
+// implementations.
+func decompressBlob(blob []byte) ([]byte, error) {
+	if len(blob) < blobHeaderBytes {
+		return nil, fmt.Errorf("blob too small: %d bytes", len(blob))
+	}
+	header := make([]uint32, 6)
+	if err := binary.Read(bytes.NewReader(blob[:blobHeaderBytes]), binary.LittleEndian, header); err != nil {
+		return nil, fmt.Errorf("decode blob header: %w", err)
+	}
+	switch {
+	case header[0] != blobMagic:
+		return nil, fmt.Errorf("invalid blob magic: %#x", header[0])
+	case header[1] != blobVersion:
+		return nil, fmt.Errorf("unsupported blob version: %d", header[1])
+	case header[2] != algorithmZstd:
+		return nil, fmt.Errorf("unsupported compression algorithm: %d", header[2])
+	case header[4] != zstdHeaderVer:
+		return nil, fmt.Errorf("unsupported zstd header version: %d", header[4])
+	case header[5] != zstdDictionary:
+		return nil, fmt.Errorf("zstd dictionaries are not supported")
+	}
+	if header[3] == 0 {
+		return nil, nil
+	}
+	data, err := blobDecoder.DecodeAll(blob[blobHeaderBytes:], nil)
+	if err != nil {
+		return nil, fmt.Errorf("decompress blob: %w", err)
+	}
+	if uint32(len(data)) != header[3] {
+		return nil, fmt.Errorf("blob size mismatch: header says %d, got %d", header[3], len(data))
+	}
+	return data, nil
+}

internal/nwsync/emit.go

@@ -0,0 +1,247 @@
+package nwsync
+
+import (
+	"bytes"
+	"crypto/sha1"
+	"fmt"
+	"os"
+	"path/filepath"
+	"slices"
+	"sort"
+	"strconv"
+	"strings"
+	"time"
+
+	"git.westgate.pw/ShadowsOverWestgate/sow-tools/internal/buildinfo"
+	"git.westgate.pw/ShadowsOverWestgate/sow-tools/internal/erf"
+)
+
+// fileSizeLimit matches upstream's --limit-file-size default of 15 MB. A
+// resource over it is a hard failure, not a skip: upstream quit(1)s and so do
+// we. Our largest resource today is 13.66 MiB, so the headroom is thin.
+const fileSizeLimit = 15 * 1024 * 1024
+
+// skippedTypes are never published, matching upstream's GobalResTypeSkipList.
+var skippedTypes = resTypes("nss", "ndb", "gic")
+
+// emitterVersion identifies the blob/manifest byte format this package
+// produces. assemble refuses to merge indexes that disagree on it, because two
+// producers of blobs mean a skewed emitter can otherwise write blobs the merged
+// manifest quietly disagrees with. Bump it only when emitted bytes change — it
+// is deliberately not the build revision, which would invalidate every
+// published index on every unrelated commit.
+const emitterVersion = "1"
+
+// serverTypes are loaded only server-side; a manifest holding nothing else
+// has no client contents. Mirrors upstream's GlobalResTypeServerList, whose
+// trailing 0 is RESTYPE_INVALID.
+var serverTypes = append(resTypes(
+	"are", "dlg", "fac", "gic", "git", "ifo", "itp", "jrl", "ncs", "ndb",
+	"nss", "ptm", "utc", "utd", "ute", "uti", "utm", "utp", "uts", "utt", "utw",
+), 0)
+
+func resTypes(extensions ...string) []uint16 {
+	types := make([]uint16, 0, len(extensions))
+	for _, extension := range extensions {
+		restype, ok := erf.ResourceTypeForExtension(extension)
+		if !ok {
+			panic("nwsync: unknown restype " + extension)
+		}
+		types = append(types, restype)
+	}
+	return types
+}
+
+// EmitResult reports what one emit run produced.
+type EmitResult struct {
+	Name         string // artifact name, without extension
+	ManifestPath string
+	Entries      int
+	BlobsWritten int
+}
+
+// Emit explodes one artifact — a .hak/.erf/.mod or a loose file such as the
+// TLK — into NWSync blobs plus a NSYM manifest describing only that artifact.
+func Emit(artifactPath, outDir string) (EmitResult, error) {
+	resources, err := readArtifact(artifactPath)
+	if err != nil {
+		return EmitResult{}, err
+	}
+	name := strings.TrimSuffix(filepath.Base(artifactPath), filepath.Ext(artifactPath))
+
+	entries, blobs, onDiskBytes, err := emitResources(resources, outDir)
+	if err != nil {
+		return EmitResult{}, err
+	}
+	if len(entries) == 0 {
+		return EmitResult{}, fmt.Errorf("%s: nothing to index (no publishable resources)", artifactPath)
+	}
+
+	data, err := writeManifest(entries)
+	if err != nil {
+		return EmitResult{}, err
+	}
+	manifestPath := filepath.Join(outDir, name+".nsym")
+	if err := writeManifestPair(manifestPath, data, entries, onDiskBytes, Sidecar{ModuleName: name}); err != nil {
+		return EmitResult{}, err
+	}
+	return EmitResult{Name: name, ManifestPath: manifestPath, Entries: len(entries), BlobsWritten: blobs}, nil
+}
+
+// readArtifact returns the resources of an ERF/HAK/MOD, or the single resource
+// a loose file represents. Upstream's resman does the same dispatch on the
+// file's first three bytes.
+func readArtifact(path string) ([]erf.Resource, error) {
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return nil, fmt.Errorf("read artifact: %w", err)
+	}
+	if len(data) >= 3 {
+		switch string(data[:3]) {
+		case "ERF", "HAK":
+			archive, err := erf.Read(bytes.NewReader(data))
+			if err != nil {
+				return nil, fmt.Errorf("%s: %w", path, err)
+			}
+			return archive.Resources, nil
+		case "MOD":
+			// A persistent world never publishes module contents, so the .mod
+			// contributes no bytes to a manifest — it only says which haks and
+			// which TLK the manifest covers.
+			return nil, fmt.Errorf("%s: a module is never emitted; a manifest is haks plus the TLK", path)
+		}
+	}
+	base := filepath.Base(path)
+	extension := filepath.Ext(base)
+	restype, ok := erf.ResourceTypeForExtension(extension)
+	if !ok {
+		return nil, fmt.Errorf("%s: unknown resource type %q", path, extension)
+	}
+	return []erf.Resource{{
+		Name: strings.TrimSuffix(base, extension),
+		Type: restype,
+		Data: data,
+	}}, nil
+}
+
+func emitResources(resources []erf.Resource, outDir string) ([]Entry, int, int64, error) {
+	// A resref appearing twice inside one artifact resolves to the last one,
+	// the way resman lets the last container added win.
+	order := make([]Identity, 0, len(resources))
+	latest := make(map[Identity]erf.Resource, len(resources))
+	var tooBig []string
+	for _, resource := range resources {
+		if _, ok := erf.ExtensionForResourceType(resource.Type); !ok {
+			return nil, 0, 0, fmt.Errorf("resref %s is not resolvable (unknown restype %d)", resource.Name, resource.Type)
+		}
+		if slices.Contains(skippedTypes, resource.Type) {
+			continue
+		}
+		if len(resource.Data) > fileSizeLimit {
+			tooBig = append(tooBig, fmt.Sprintf("%s: %d bytes > %d", resource.Name, len(resource.Data), fileSizeLimit))
+			continue
+		}
+		identity := Identity{ResRef: strings.ToLower(resource.Name), ResType: resource.Type}
+		if _, seen := latest[identity]; !seen {
+			order = append(order, identity)
+		}
+		latest[identity] = resource
+	}
+	if len(tooBig) > 0 {
+		sort.Strings(tooBig)
+		return nil, 0, 0, fmt.Errorf("resources exceed the file size limit:\n  %s", strings.Join(tooBig, "\n  "))
+	}
+
+	entries := make([]Entry, 0, len(order))
+	var blobs int
+	var onDiskBytes int64
+	for _, identity := range order {
+		resource := latest[identity]
+		sum := sha1.Sum(resource.Data)
+		entries = append(entries, Entry{
+			SHA1:    sum,
+			Size:    uint32(len(resource.Data)),
+			ResRef:  identity.ResRef,
+			ResType: identity.ResType,
+		})
+		written, err := writeBlob(outDir, sum, resource.Data)
+		if err != nil {
+			return nil, 0, 0, err
+		}
+		if written > 0 {
+			blobs++
+			onDiskBytes += written
+		}
+	}
+	return entries, blobs, onDiskBytes, nil
+}
+
+// created is the sidecar timestamp. SOURCE_DATE_EPOCH pins it so a build can
+// be reproduced byte for byte; the manifest itself is deterministic already.
+func created() int64 {
+	if raw := os.Getenv("SOURCE_DATE_EPOCH"); raw != "" {
+		if seconds, err := strconv.ParseInt(raw, 10, 64); err == nil {
+			return seconds
+		}
+	}
+	return time.Now().Unix()
+}
+
+// writeBlob writes one NWCompressedBuffer blob and returns its size on disk,
+// or 0 if the blob already existed. Blob names are content hashes, so an
+// existing name is existing content.
+func writeBlob(outDir string, sum [20]byte, data []byte) (int64, error) {
+	path := blobPath(outDir, fmt.Sprintf("%x", sum))
+	if _, err := os.Stat(path); err == nil {
+		return 0, nil
+	}
+	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
+		return 0, fmt.Errorf("create blob directory: %w", err)
+	}
+	blob := compressBlob(data)
+	if err := os.WriteFile(path, blob, 0o644); err != nil {
+		return 0, fmt.Errorf("write blob: %w", err)
+	}
+	return int64(len(blob)), nil
+}
+
+// writeManifestPair writes a NSYM manifest and its .json sidecar. The caller
+// supplies the sidecar fields it knows; the rest are derived from the entries.
+// data must be the serialised form of entries.
+func writeManifestPair(manifestPath string, data []byte, entries []Entry, onDiskBytes int64, sidecar Sidecar) error {
+	if err := os.MkdirAll(filepath.Dir(manifestPath), 0o755); err != nil {
+		return fmt.Errorf("create manifest directory: %w", err)
+	}
+	if err := os.WriteFile(manifestPath, data, 0o644); err != nil {
+		return fmt.Errorf("write manifest: %w", err)
+	}
+
+	var totalBytes int64
+	clientContents := false
+	for _, entry := range entries {
+		totalBytes += int64(entry.Size)
+		if !slices.Contains(serverTypes, entry.ResType) {
+			clientContents = true
+		}
+	}
+	sidecar.Version = manifestVersion
+	sidecar.SHA1 = fmt.Sprintf("%x", sha1.Sum(data))
+	sidecar.HashTreeDepth = hashTreeDepth
+	sidecar.IncludesModuleContents = false
+	sidecar.IncludesClientContents = clientContents
+	sidecar.TotalFiles = len(entries)
+	sidecar.TotalBytes = totalBytes
+	sidecar.OnDiskBytes = onDiskBytes
+	sidecar.Created = created()
+	sidecar.CreatedWith = buildinfo.String()
+	sidecar.EmitterVersion = emitterVersion
+
+	body, err := marshalSidecar(sidecar)
+	if err != nil {
+		return err
+	}
+	if err := os.WriteFile(manifestPath+".json", body, 0o644); err != nil {
+		return fmt.Errorf("write sidecar: %w", err)
+	}
+	return nil
+}

internal/nwsync/manifest.go

@@ -0,0 +1,200 @@
+package nwsync
+
+import (
+	"bytes"
+	"encoding/binary"
+	"encoding/hex"
+	"encoding/json"
+	"fmt"
+	"io"
+	"path/filepath"
+	"sort"
+	"strings"
+)
+
+// NSYM manifest, version 3, exactly as upstream neverwinter/nwsync.nim writes
+// it. Everything is little-endian:
+//
+//	"NSYM", uint32 version, uint32 entry count, uint32 mapping count,
+//	entries:  byte[20] raw sha1, uint32 size, char[16] resref, uint16 restype
+//	mappings: uint32 entry index,             char[16] resref, uint16 restype
+//
+// Entries are sorted by lowercase sha1 hex then resref, and a resource whose
+// sha1 was already written becomes a mapping instead of a second entry.
+const (
+	manifestVersion = 3
+	hashTreeDepth   = 2
+	resRefBytes     = 16
+)
+
+// Entry is one resource in a manifest.
+type Entry struct {
+	SHA1    [20]byte
+	Size    uint32
+	ResRef  string // lowercase, no extension, at most 16 characters
+	ResType uint16
+}
+
+func (e Entry) sha1Hex() string { return hex.EncodeToString(e.SHA1[:]) }
+
+// Identity is what a resref resolves by: name plus type. It is the merge key
+// inside one artifact and across artifacts alike.
+type Identity struct {
+	ResRef  string
+	ResType uint16
+}
+
+func (e Entry) identity() Identity { return Identity{ResRef: e.ResRef, ResType: e.ResType} }
+
+// writeManifest serialises entries into NSYM v3 bytes.
+func writeManifest(entries []Entry) ([]byte, error) {
+	sorted := make([]Entry, len(entries))
+	copy(sorted, entries)
+	sort.SliceStable(sorted, func(i, j int) bool {
+		left, right := sorted[i].sha1Hex(), sorted[j].sha1Hex()
+		if left != right {
+			return left < right
+		}
+		return sorted[i].ResRef < sorted[j].ResRef
+	})
+
+	var body, mappings bytes.Buffer
+	seen := make(map[string]uint32, len(sorted))
+	var entryCount, mappingCount uint32
+	for _, entry := range sorted {
+		padded, err := padResRef(entry.ResRef)
+		if err != nil {
+			return nil, err
+		}
+		if index, ok := seen[entry.sha1Hex()]; ok {
+			_ = binary.Write(&mappings, binary.LittleEndian, index)
+			mappings.Write(padded)
+			_ = binary.Write(&mappings, binary.LittleEndian, entry.ResType)
+			mappingCount++
+			continue
+		}
+		seen[entry.sha1Hex()] = entryCount
+		entryCount++
+		body.Write(entry.SHA1[:])
+		_ = binary.Write(&body, binary.LittleEndian, entry.Size)
+		body.Write(padded)
+		_ = binary.Write(&body, binary.LittleEndian, entry.ResType)
+	}
+
+	var out bytes.Buffer
+	out.WriteString("NSYM")
+	for _, field := range []uint32{manifestVersion, entryCount, mappingCount} {
+		_ = binary.Write(&out, binary.LittleEndian, field)
+	}
+	out.Write(body.Bytes())
+	out.Write(mappings.Bytes())
+	return out.Bytes(), nil
+}
+
+// readManifest parses NSYM v3 bytes. Mappings are expanded back into entries,
+// the way upstream's reader does, so a caller sees one entry per resref.
+func readManifest(data []byte) ([]Entry, error) {
+	reader := bytes.NewReader(data)
+	magic := make([]byte, 4)
+	if _, err := io.ReadFull(reader, magic); err != nil || string(magic) != "NSYM" {
+		return nil, fmt.Errorf("not a manifest (invalid magic bytes)")
+	}
+	var version, entryCount, mappingCount uint32
+	for _, field := range []*uint32{&version, &entryCount, &mappingCount} {
+		if err := binary.Read(reader, binary.LittleEndian, field); err != nil {
+			return nil, fmt.Errorf("truncated manifest header: %w", err)
+		}
+	}
+	if version != manifestVersion {
+		return nil, fmt.Errorf("unsupported manifest version %d", version)
+	}
+
+	entries := make([]Entry, 0, entryCount+mappingCount)
+	for i := uint32(0); i < entryCount; i++ {
+		var entry Entry
+		if _, err := io.ReadFull(reader, entry.SHA1[:]); err != nil {
+			return nil, fmt.Errorf("truncated entry %d: %w", i, err)
+		}
+		if err := binary.Read(reader, binary.LittleEndian, &entry.Size); err != nil {
+			return nil, fmt.Errorf("truncated entry %d: %w", i, err)
+		}
+		resref, restype, err := readResRef(reader)
+		if err != nil {
+			return nil, fmt.Errorf("truncated entry %d: %w", i, err)
+		}
+		entry.ResRef, entry.ResType = resref, restype
+		entries = append(entries, entry)
+	}
+	for i := uint32(0); i < mappingCount; i++ {
+		var index uint32
+		if err := binary.Read(reader, binary.LittleEndian, &index); err != nil {
+			return nil, fmt.Errorf("truncated mapping %d: %w", i, err)
+		}
+		if index >= entryCount {
+			return nil, fmt.Errorf("mapping %d references non-existent entry %d", i, index)
+		}
+		resref, restype, err := readResRef(reader)
+		if err != nil {
+			return nil, fmt.Errorf("truncated mapping %d: %w", i, err)
+		}
+		target := entries[index]
+		entries = append(entries, Entry{SHA1: target.SHA1, Size: target.Size, ResRef: resref, ResType: restype})
+	}
+	return entries, nil
+}
+
+func readResRef(reader *bytes.Reader) (string, uint16, error) {
+	raw := make([]byte, resRefBytes)
+	if _, err := io.ReadFull(reader, raw); err != nil {
+		return "", 0, err
+	}
+	var restype uint16
+	if err := binary.Read(reader, binary.LittleEndian, &restype); err != nil {
+		return "", 0, err
+	}
+	return strings.ToLower(string(bytes.TrimRight(raw, "\x00"))), restype, nil
+}
+
+func padResRef(resref string) ([]byte, error) {
+	if len(resref) > resRefBytes {
+		return nil, fmt.Errorf("resref %q exceeds %d characters", resref, resRefBytes)
+	}
+	padded := make([]byte, resRefBytes)
+	copy(padded, strings.ToLower(resref))
+	return padded, nil
+}
+
+// Sidecar is the .json file written next to every manifest. Clients fetch it
+// (nwn_nwsync_fetch.nim), so the field names and order match upstream.
+// Upstream omits an integer meta field whose value is 0, hence group_id's
+// omitempty: 0 means absent, not "group zero".
+type Sidecar struct {
+	Version                int    `json:"version"`
+	SHA1                   string `json:"sha1"`
+	HashTreeDepth          int    `json:"hash_tree_depth"`
+	ModuleName             string `json:"module_name"`
+	Description            string `json:"description"`
+	IncludesModuleContents bool   `json:"includes_module_contents"`
+	IncludesClientContents bool   `json:"includes_client_contents"`
+	TotalFiles             int    `json:"total_files"`
+	TotalBytes             int64  `json:"total_bytes"`
+	OnDiskBytes            int64  `json:"on_disk_bytes"`
+	Created                int64  `json:"created"`
+	CreatedWith            string `json:"created_with"`
+	EmitterVersion         string `json:"emitter_version,omitempty"`
+	GroupID                int    `json:"group_id,omitempty"`
+}
+
+func marshalSidecar(sidecar Sidecar) ([]byte, error) {
+	body, err := json.MarshalIndent(sidecar, "", "  ")
+	if err != nil {
+		return nil, err
+	}
+	// Upstream terminates the file with CRLF; match it.
+	return append(body, '\r', '\n'), nil
+}
+
+// blobPath is the data store path for a blob, hash tree depth 2.
+func blobPath(root, sha1Hex string) string {
+	return filepath.Join(root, "data", "sha1", sha1Hex[0:2], sha1Hex[2:4], sha1Hex)
+}

internal/nwsync/nwsync_test.go

@@ -0,0 +1,450 @@
+package nwsync
+
+import (
+	"bytes"
+	"crypto/sha1"
+	"encoding/binary"
+	"encoding/hex"
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"git.westgate.pw/ShadowsOverWestgate/sow-tools/internal/erf"
+)
+
+func restype(t *testing.T, extension string) uint16 {
+	t.Helper()
+	value, ok := erf.ResourceTypeForExtension(extension)
+	if !ok {
+		t.Fatalf("unknown restype %q", extension)
+	}
+	return value
+}
+
+// writeHak builds a HAK fixture from resref.ext => body pairs.
+func writeHak(t *testing.T, path string, contents map[string][]byte) {
+	t.Helper()
+	resources := make([]erf.Resource, 0, len(contents))
+	for name, body := range contents {
+		stem, extension, _ := strings.Cut(name, ".")
+		resources = append(resources, erf.Resource{
+			Name: stem,
+			Type: restype(t, extension),
+			Data: body,
+			Size: int64(len(body)),
+		})
+	}
+	var out bytes.Buffer
+	if err := erf.Write(&out, erf.New("HAK", resources)); err != nil {
+		t.Fatalf("write hak: %v", err)
+	}
+	if err := os.WriteFile(path, out.Bytes(), 0o644); err != nil {
+		t.Fatalf("write hak file: %v", err)
+	}
+}
+
+func TestBlobFramingRoundTrips(t *testing.T) {
+	data := []byte("the quick brown fox jumps over the lazy dog, repeatedly and at length")
+	blob := compressBlob(data)
+
+	header := make([]uint32, 6)
+	if err := binary.Read(bytes.NewReader(blob[:blobHeaderBytes]), binary.LittleEndian, header); err != nil {
+		t.Fatalf("read header: %v", err)
+	}
+	want := []uint32{blobMagic, 3, 2, uint32(len(data)), 1, 0}
+	for i := range want {
+		if header[i] != want[i] {
+			t.Errorf("header field %d = %d, want %d", i, header[i], want[i])
+		}
+	}
+	if string(blob[:4]) != "NSYC" {
+		t.Errorf("magic bytes = %q, want NSYC", blob[:4])
+	}
+
+	got, err := decompressBlob(blob)
+	if err != nil {
+		t.Fatalf("decompress: %v", err)
+	}
+	if !bytes.Equal(got, data) {
+		t.Errorf("round trip mismatch: %q", got)
+	}
+}
+
+func TestManifestBytesMatchUpstreamLayout(t *testing.T) {
+	shared := sha1.Sum([]byte("shared"))
+	other := sha1.Sum([]byte("other"))
+	// Deliberately out of order, and with two resrefs sharing one sha1: the
+	// second one must become a mapping, not a second entry.
+	entries := []Entry{
+		{SHA1: other, Size: 5, ResRef: "zzz", ResType: 1},
+		{SHA1: shared, Size: 6, ResRef: "bbb", ResType: 2},
+		{SHA1: shared, Size: 6, ResRef: "aaa", ResType: 3},
+	}
+	data, err := writeManifest(entries)
+	if err != nil {
+		t.Fatalf("write manifest: %v", err)
+	}
+
+	if string(data[:4]) != "NSYM" {
+		t.Fatalf("magic = %q", data[:4])
+	}
+	var version, entryCount, mappingCount uint32
+	reader := bytes.NewReader(data[4:16])
+	for _, field := range []*uint32{&version, &entryCount, &mappingCount} {
+		_ = binary.Read(reader, binary.LittleEndian, field)
+	}
+	if version != 3 || entryCount != 2 || mappingCount != 1 {
+		t.Fatalf("header = version %d, %d entries, %d mappings; want 3/2/1", version, entryCount, mappingCount)
+	}
+	wantSize := 16 + int(entryCount)*(20+4+16+2) + int(mappingCount)*(4+16+2)
+	if len(data) != wantSize {
+		t.Fatalf("manifest is %d bytes, want %d", len(data), wantSize)
+	}
+
+	// Sorted by sha1 hex then resref, so the shared hash's "aaa" is the entry
+	// and "bbb" is demoted to a mapping.
+	round, err := readManifest(data)
+	if err != nil {
+		t.Fatalf("read manifest: %v", err)
+	}
+	if len(round) != 3 {
+		t.Fatalf("round trip returned %d entries, want 3", len(round))
+	}
+	byResRef := map[string]Entry{}
+	for _, entry := range round {
+		byResRef[entry.ResRef] = entry
+	}
+	for _, entry := range entries {
+		got, ok := byResRef[entry.ResRef]
+		if !ok {
+			t.Fatalf("resref %q lost in round trip", entry.ResRef)
+		}
+		if got != entry {
+			t.Errorf("resref %q = %+v, want %+v", entry.ResRef, got, entry)
+		}
+	}
+	if round[0].ResRef != "aaa" && round[1].ResRef != "aaa" {
+		t.Errorf("entries are not sorted by sha1 then resref: %+v", round)
+	}
+}
+
+func TestEmitWritesBlobsAndManifest(t *testing.T) {
+	dir := t.TempDir()
+	hak := filepath.Join(dir, "sow_test_01.hak")
+	body := []byte("texture bytes")
+	writeHak(t, hak, map[string][]byte{
+		"bloodstain1.tga": body,
+		"copy1.txi":       body, // same content, different resref: one blob
+		"script1.nss":     []byte("void main() {}"),
+		"debug1.ndb":      []byte("debug"),
+		"comment1.gic":    []byte("comment"),
+	})
+
+	out := filepath.Join(dir, "out")
+	result, err := Emit(hak, out)
+	if err != nil {
+		t.Fatalf("emit: %v", err)
+	}
+	if result.Entries != 2 {
+		t.Errorf("emitted %d entries, want 2 (nss/ndb/gic are always skipped)", result.Entries)
+	}
+	if result.BlobsWritten != 1 {
+		t.Errorf("wrote %d blobs, want 1 (identical content shares a blob)", result.BlobsWritten)
+	}
+
+	// The blob is named by the sha1 of the uncompressed bytes, under a depth-2
+	// hash tree, and decompresses back to exactly those bytes.
+	sum := sha1.Sum(body)
+	name := hex.EncodeToString(sum[:])
+	path := filepath.Join(out, "data", "sha1", name[0:2], name[2:4], name)
+	blob, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("blob missing at %s: %v", path, err)
+	}
+	got, err := decompressBlob(blob)
+	if err != nil {
+		t.Fatalf("decompress blob: %v", err)
+	}
+	if !bytes.Equal(got, body) {
+		t.Errorf("blob decompressed to %q, want %q", got, body)
+	}
+
+	entries := readEmitted(t, out, "sow_test_01")
+	for _, entry := range entries {
+		if entry.ResType == restype(t, "nss") || entry.ResType == restype(t, "ndb") || entry.ResType == restype(t, "gic") {
+			t.Errorf("skipped restype leaked into the manifest: %+v", entry)
+		}
+	}
+
+	var sidecar Sidecar
+	body2, err := os.ReadFile(result.ManifestPath + ".json")
+	if err != nil {
+		t.Fatalf("sidecar missing: %v", err)
+	}
+	if err := json.Unmarshal(body2, &sidecar); err != nil {
+		t.Fatalf("sidecar json: %v", err)
+	}
+	if sidecar.TotalFiles != 2 || sidecar.HashTreeDepth != 2 || sidecar.Version != 3 {
+		t.Errorf("sidecar = %+v", sidecar)
+	}
+	if sidecar.IncludesModuleContents {
+		t.Error("sidecar claims module contents; a published manifest never has them")
+	}
+	if strings.Contains(string(body2), "group_id") {
+		t.Error("per-artifact sidecar should omit group_id (0 means absent)")
+	}
+	if _, err := os.Stat(filepath.Join(out, "latest")); err == nil {
+		t.Error("a latest file was written; there must never be one")
+	}
+}
+
+func readEmitted(t *testing.T, dir, name string) []Entry {
+	t.Helper()
+	data, err := os.ReadFile(filepath.Join(dir, name+".nsym"))
+	if err != nil {
+		t.Fatalf("read emitted manifest: %v", err)
+	}
+	entries, err := readManifest(data)
+	if err != nil {
+		t.Fatalf("parse emitted manifest: %v", err)
+	}
+	return entries
+}
+
+func TestEmitFailsClosedOnOversizeResource(t *testing.T) {
+	dir := t.TempDir()
+	hak := filepath.Join(dir, "big.hak")
+	writeHak(t, hak, map[string][]byte{"huge1.tga": make([]byte, fileSizeLimit+1)})
+
+	if _, err := Emit(hak, filepath.Join(dir, "out")); err == nil {
+		t.Fatal("emit accepted a resource over the 15 MB limit")
+	}
+}
+
+func TestEmitLooseFile(t *testing.T) {
+	dir := t.TempDir()
+	tlk := filepath.Join(dir, "sow_tlk.tlk")
+	if err := os.WriteFile(tlk, []byte("TLK V3.0 payload"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	out := filepath.Join(dir, "out")
+	result, err := Emit(tlk, out)
+	if err != nil {
+		t.Fatalf("emit tlk: %v", err)
+	}
+	entries := readEmitted(t, out, "sow_tlk")
+	if len(entries) != 1 || entries[0].ResRef != "sow_tlk" || entries[0].ResType != restype(t, "tlk") {
+		t.Fatalf("tlk emitted as %+v", entries)
+	}
+	if result.BlobsWritten != 1 {
+		t.Errorf("wrote %d blobs, want 1", result.BlobsWritten)
+	}
+}
+
+// emitFixture emits two haks that share a resref, so the merge rule is
+// observable: "top" holds the winning body, "assets" the shadowed one.
+func emitFixture(t *testing.T) (string, []byte, []byte) {
+	t.Helper()
+	dir := t.TempDir()
+	topBody := []byte("2da from sow_top")
+	assetBody := []byte("2da from the asset hak")
+	writeHak(t, filepath.Join(dir, "sow_top.hak"), map[string][]byte{"appearance.2da": topBody})
+	writeHak(t, filepath.Join(dir, "sow_core_01.hak"), map[string][]byte{
+		"appearance.2da":  assetBody,
+		"bloodstain1.tga": []byte("blood"),
+	})
+	out := filepath.Join(dir, "out")
+	for _, name := range []string{"sow_top", "sow_core_01"} {
+		if _, err := Emit(filepath.Join(dir, name+".hak"), out); err != nil {
+			t.Fatalf("emit %s: %v", name, err)
+		}
+	}
+	return out, topBody, assetBody
+}
+
+func TestAssembleShadowsByOrder(t *testing.T) {
+	entriesDir, topBody, assetBody := emitFixture(t)
+	out := t.TempDir()
+
+	result, err := Assemble(AssembleOptions{
+		Order:      []string{"sow_top", "sow_core_01"},
+		EntriesDir: entriesDir,
+		OutDir:     out,
+		GroupID:    2,
+	})
+	if err != nil {
+		t.Fatalf("assemble: %v", err)
+	}
+	if result.Entries != 2 {
+		t.Fatalf("merged %d entries, want 2 (appearance.2da is shadowed, not duplicated)", result.Entries)
+	}
+
+	data, err := os.ReadFile(result.ManifestPath)
+	if err != nil {
+		t.Fatalf("read merged manifest: %v", err)
+	}
+	merged := sha1.Sum(data)
+	if hex.EncodeToString(merged[:]) != result.SHA1 || filepath.Base(result.ManifestPath) != result.SHA1 {
+		t.Errorf("manifest is not named by its own sha1: %s", result.ManifestPath)
+	}
+	entries, err := readManifest(data)
+	if err != nil {
+		t.Fatalf("parse merged manifest: %v", err)
+	}
+	for _, entry := range entries {
+		if entry.ResRef != "appearance" {
+			continue
+		}
+		if entry.SHA1 != sha1.Sum(topBody) {
+			t.Errorf("appearance.2da resolved to the wrong hak; want the earliest in --order")
+		}
+		if entry.SHA1 == sha1.Sum(assetBody) {
+			t.Error("appearance.2da resolved to the shadowed hak")
+		}
+	}
+
+	sidecar := Sidecar{}
+	body, err := os.ReadFile(result.ManifestPath + ".json")
+	if err != nil {
+		t.Fatalf("merged sidecar missing: %v", err)
+	}
+	if err := json.Unmarshal(body, &sidecar); err != nil {
+		t.Fatalf("merged sidecar json: %v", err)
+	}
+	if sidecar.GroupID != 2 {
+		t.Errorf("group_id = %d, want 2 (testing)", sidecar.GroupID)
+	}
+	if sidecar.SHA1 != result.SHA1 {
+		t.Errorf("sidecar sha1 = %s, want %s", sidecar.SHA1, result.SHA1)
+	}
+	if sidecar.TotalFiles != 2 {
+		t.Errorf("total_files = %d, want 2", sidecar.TotalFiles)
+	}
+}
+
+func TestAssembleReversedOrderPicksTheOtherHak(t *testing.T) {
+	entriesDir, topBody, assetBody := emitFixture(t)
+	out := t.TempDir()
+
+	result, err := Assemble(AssembleOptions{
+		Order:      []string{"sow_core_01", "sow_top"},
+		EntriesDir: entriesDir,
+		OutDir:     out,
+	})
+	if err != nil {
+		t.Fatalf("assemble: %v", err)
+	}
+	data, _ := os.ReadFile(result.ManifestPath)
+	entries, err := readManifest(data)
+	if err != nil {
+		t.Fatalf("parse merged manifest: %v", err)
+	}
+	for _, entry := range entries {
+		if entry.ResRef == "appearance" && entry.SHA1 != sha1.Sum(assetBody) {
+			t.Errorf("appearance.2da did not follow --order; still resolves to %x", entry.SHA1)
+		}
+	}
+	_ = topBody
+}
+
+func TestAssembleRefusesMismatchedEmitterVersions(t *testing.T) {
+	entriesDir, _, _ := emitFixture(t)
+
+	path := filepath.Join(entriesDir, "sow_core_01.nsym.json")
+	var sidecar Sidecar
+	body, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := json.Unmarshal(body, &sidecar); err != nil {
+		t.Fatal(err)
+	}
+	sidecar.EmitterVersion = "0"
+	patched, err := marshalSidecar(sidecar)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(path, patched, 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	_, err = Assemble(AssembleOptions{
+		Order:      []string{"sow_top", "sow_core_01"},
+		EntriesDir: entriesDir,
+		OutDir:     t.TempDir(),
+	})
+	if err == nil || !strings.Contains(err.Error(), "emitter version mismatch") {
+		t.Fatalf("assemble merged across emitter versions: %v", err)
+	}
+}
+
+func TestEmitHonoursSourceDateEpoch(t *testing.T) {
+	t.Setenv("SOURCE_DATE_EPOCH", "1700000000")
+	dir := t.TempDir()
+	hak := filepath.Join(dir, "pinned.hak")
+	writeHak(t, hak, map[string][]byte{"one1.tga": []byte("body")})
+	result, err := Emit(hak, filepath.Join(dir, "out"))
+	if err != nil {
+		t.Fatalf("emit: %v", err)
+	}
+	body, err := os.ReadFile(result.ManifestPath + ".json")
+	if err != nil {
+		t.Fatal(err)
+	}
+	var sidecar Sidecar
+	if err := json.Unmarshal(body, &sidecar); err != nil {
+		t.Fatal(err)
+	}
+	if sidecar.Created != 1700000000 {
+		t.Errorf("created = %d, want the pinned SOURCE_DATE_EPOCH", sidecar.Created)
+	}
+}
+
+func TestEmitRejectsAModule(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "sow.mod")
+	var out bytes.Buffer
+	if err := erf.Write(&out, erf.New("MOD", []erf.Resource{
+		{Name: "module", Type: restype(t, "ifo"), Data: []byte("ifo"), Size: 3},
+	})); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(path, out.Bytes(), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := Emit(path, filepath.Join(dir, "out")); err == nil {
+		t.Fatal("emit accepted a .mod; a manifest never carries module contents")
+	}
+}
+
+func TestAssembleFailsClosedOnMissingIndex(t *testing.T) {
+	entriesDir, _, _ := emitFixture(t)
+	_, err := Assemble(AssembleOptions{
+		Order:      []string{"sow_top", "sow_never_published"},
+		EntriesDir: entriesDir,
+		OutDir:     t.TempDir(),
+	})
+	if err == nil || !strings.Contains(err.Error(), "sow_never_published") {
+		t.Fatalf("assemble did not fail closed and name the missing artifact: %v", err)
+	}
+}
+
+func TestRunUsageErrors(t *testing.T) {
+	cases := [][]string{
+		nil,
+		{"nope"},
+		{"emit"},
+		{"emit", "artifact.hak"},
+		{"assemble", "--entries", "x", "--out", "y"},
+		{"assemble", "--order", "a", "--out", "y"},
+		{"assemble", "--order", "a", "--entries", "x"},
+	}
+	for _, args := range cases {
+		var out, errw bytes.Buffer
+		if code := Run(args, &out, &errw); code != exitUsage {
+			t.Errorf("Run(%v) exit=%d, want %d", args, code, exitUsage)
+		}
+	}
+}

internal/nwsync/run.go

@@ -0,0 +1,129 @@
+// Package nwsync publishes NWSync repository data: blobs and a per-artifact
+// NSYM manifest at each artifact's birth (emit), and one merged manifest at
+// module release (assemble).
+//
+// The split exists because upstream nwn_nwsync_write wants every hak, the TLK
+// and the module present in one run on one disk, which our build hosts cannot
+// hold. Upstream stays the conformance oracle: manifests compare byte for
+// byte, blobs compare after decompression.
+package nwsync
+
+import (
+	"flag"
+	"fmt"
+	"io"
+	"strings"
+)
+
+const (
+	exitOK       = 0
+	exitUsage    = 64
+	exitInternal = 70
+)
+
+// Run executes an nwsync subcommand. args[0] is the subcommand (emit|assemble);
+// returns the process exit code.
+func Run(args []string, stdout, stderr io.Writer) int {
+	if len(args) == 0 {
+		printRunUsage(stderr)
+		return exitUsage
+	}
+	switch args[0] {
+	case "emit":
+		return runEmit(args[1:], stdout, stderr)
+	case "assemble":
+		return runAssemble(args[1:], stdout, stderr)
+	case "-h", "--help", "help":
+		printRunUsage(stdout)
+		return exitOK
+	default:
+		fmt.Fprintf(stderr, "nwsync: unknown subcommand %q\n\n", args[0])
+		printRunUsage(stderr)
+		return exitUsage
+	}
+}
+
+func printRunUsage(w io.Writer) {
+	fmt.Fprint(w, `usage:
+  nwsync emit     <artifact> --out DIR
+  nwsync assemble --order NAMES --entries DIR --out DIR [--group-id N]
+
+emit explodes one .hak/.erf or one loose file (the TLK) into NWSync blobs plus
+a NSYM manifest covering only that artifact. assemble merges those per-artifact
+manifests into one, reading no bulk data.
+`)
+}
+
+func runEmit(args []string, stdout, stderr io.Writer) int {
+	fs := flag.NewFlagSet("emit", flag.ContinueOnError)
+	fs.SetOutput(stderr)
+	out := fs.String("out", "", "output directory (blobs plus the per-artifact manifest)")
+	if err := fs.Parse(args); err != nil {
+		return exitUsage
+	}
+	if fs.NArg() != 1 {
+		fmt.Fprintf(stderr, "nwsync emit: exactly one artifact is required\n")
+		return exitUsage
+	}
+	if *out == "" {
+		fmt.Fprintf(stderr, "nwsync emit: --out is required\n")
+		return exitUsage
+	}
+
+	result, err := Emit(fs.Arg(0), *out)
+	if err != nil {
+		fmt.Fprintf(stderr, "nwsync emit: %v\n", err)
+		return exitInternal
+	}
+	fmt.Fprintf(stdout, "emitted %s: %d resources, %d new blobs, manifest %s\n",
+		result.Name, result.Entries, result.BlobsWritten, result.ManifestPath)
+	return exitOK
+}
+
+func runAssemble(args []string, stdout, stderr io.Writer) int {
+	fs := flag.NewFlagSet("assemble", flag.ContinueOnError)
+	fs.SetOutput(stderr)
+	order := fs.String("order", "", "comma-separated artifact names, highest priority first")
+	entries := fs.String("entries", "", "directory holding the per-artifact .nsym files")
+	out := fs.String("out", "", "repository root; the manifest lands in <out>/manifests")
+	groupID := fs.Int("group-id", 0, "NWSync group id (1 = current, 2 = testing; 0 omits it)")
+	moduleName := fs.String("module-name", "", "module name recorded in the sidecar")
+	description := fs.String("description", "", "description recorded in the sidecar")
+	if err := fs.Parse(args); err != nil {
+		return exitUsage
+	}
+	switch {
+	case *order == "":
+		fmt.Fprintf(stderr, "nwsync assemble: --order is required\n")
+		return exitUsage
+	case *entries == "":
+		fmt.Fprintf(stderr, "nwsync assemble: --entries is required\n")
+		return exitUsage
+	case *out == "":
+		fmt.Fprintf(stderr, "nwsync assemble: --out is required\n")
+		return exitUsage
+	}
+
+	names := make([]string, 0, 8)
+	for _, name := range strings.Split(*order, ",") {
+		if name = strings.TrimSpace(name); name != "" {
+			names = append(names, name)
+		}
+	}
+
+	result, err := Assemble(AssembleOptions{
+		Order:       names,
+		EntriesDir:  *entries,
+		OutDir:      *out,
+		GroupID:     *groupID,
+		ModuleName:  *moduleName,
+		Description: *description,
+	})
+	if err != nil {
+		fmt.Fprintf(stderr, "nwsync assemble: %v\n", err)
+		return exitInternal
+	}
+	fmt.Fprintf(stdout, "assembled %s: %d resources from %d artifacts\n",
+		result.SHA1, result.Entries, len(names))
+	return exitOK
+}

scripts/crucible-smoke.sh

@@ -18,7 +18,7 @@ bin=bin
 
 # Keep in sync with internal/dispatch.Registry (Wired flag).
 unwired=()
-wired=(assets depot hak module topdata wiki)
+wired=(assets depot hak module nwsync topdata wiki)
 
 exit_of() { set +e; "$@" >/dev/null 2>&1; local c=$?; set -e; echo "${c}"; }