feat(nwsync): emit blobs and per-artifact NSYM, assemble merged manifests #71

Merged
archvillainette merged 2 commits from feat/nwsync-emit-assemble into main 2026-07-29 10:50:05 +00:00
6 changed files with 96 additions and 30 deletions
Showing only changes of commit 978e2913f6 - Show all commits
+1
View File
@@ -28,6 +28,7 @@
"cmd/crucible-depot" "cmd/crucible-depot"
"cmd/crucible-hak" "cmd/crucible-hak"
"cmd/crucible-module" "cmd/crucible-module"
"cmd/crucible-nwsync"
"cmd/crucible-topdata" "cmd/crucible-topdata"
"cmd/crucible-wiki" "cmd/crucible-wiki"
]; ];
+1 -1
View File
@@ -7,4 +7,4 @@ require (
gopkg.in/yaml.v3 v3.0.1 gopkg.in/yaml.v3 v3.0.1
) )
require github.com/klauspost/compress v1.19.1 // indirect require github.com/klauspost/compress v1.19.1
+5 -13
View File
@@ -37,15 +37,9 @@ func Assemble(options AssembleOptions) (AssembleResult, error) {
return AssembleResult{}, fmt.Errorf("assemble: --order names no artifacts") return AssembleResult{}, fmt.Errorf("assemble: --order names no artifacts")
} }
type key struct {
resref string
restype uint16
}
merged := make([]Entry, 0, 1024) merged := make([]Entry, 0, 1024)
winner := make(map[key]bool, 1024) winner := make(map[Identity]bool, 1024)
var onDiskBytes int64 var onDiskBytes int64
emitterVersion := ""
emitterSource := ""
for _, name := range options.Order { for _, name := range options.Order {
manifestPath := filepath.Join(options.EntriesDir, name+".nsym") manifestPath := filepath.Join(options.EntriesDir, name+".nsym")
@@ -64,19 +58,17 @@ func Assemble(options AssembleOptions) (AssembleResult, error) {
// Two producers of blobs means a skewed emitter can write blobs the // Two producers of blobs means a skewed emitter can write blobs the
// merged manifest quietly disagrees with. Refuse to merge across // merged manifest quietly disagrees with. Refuse to merge across
// mismatched emitter versions. // mismatched emitter versions.
if emitterVersion == "" { if sidecar.EmitterVersion != emitterVersion {
emitterVersion, emitterSource = sidecar.CreatedWith, name
} else if sidecar.CreatedWith != emitterVersion {
return AssembleResult{}, fmt.Errorf( return AssembleResult{}, fmt.Errorf(
"assemble: emitter version mismatch: %s was emitted with %q, %s with %q", "assemble: emitter version mismatch: %s was emitted by emitter %q, this is emitter %q",
emitterSource, emitterVersion, name, sidecar.CreatedWith) name, sidecar.EmitterVersion, emitterVersion)
} }
// on_disk_bytes overcounts by the handful of cross-artifact // on_disk_bytes overcounts by the handful of cross-artifact
// duplicates. It is a display statistic; no dedupe pass for it. // duplicates. It is a display statistic; no dedupe pass for it.
onDiskBytes += sidecar.OnDiskBytes onDiskBytes += sidecar.OnDiskBytes
for _, entry := range entries { for _, entry := range entries {
identity := key{resref: entry.ResRef, restype: entry.ResType} identity := entry.identity()
if winner[identity] { if winner[identity] {
continue continue
} }
+35 -12
View File
@@ -8,6 +8,7 @@ import (
"path/filepath" "path/filepath"
"slices" "slices"
"sort" "sort"
"strconv"
"strings" "strings"
"time" "time"
@@ -23,8 +24,17 @@ const fileSizeLimit = 15 * 1024 * 1024
// skippedTypes are never published, matching upstream's GobalResTypeSkipList. // skippedTypes are never published, matching upstream's GobalResTypeSkipList.
var skippedTypes = resTypes("nss", "ndb", "gic") 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 // serverTypes are loaded only server-side; a manifest holding nothing else
// has no client contents. Mirrors upstream's GlobalResTypeServerList. // has no client contents. Mirrors upstream's GlobalResTypeServerList, whose
// trailing 0 is RESTYPE_INVALID.
var serverTypes = append(resTypes( var serverTypes = append(resTypes(
"are", "dlg", "fac", "gic", "git", "ifo", "itp", "jrl", "ncs", "ndb", "are", "dlg", "fac", "gic", "git", "ifo", "itp", "jrl", "ncs", "ndb",
"nss", "ptm", "utc", "utd", "ute", "uti", "utm", "utp", "uts", "utt", "utw", "nss", "ptm", "utc", "utd", "ute", "uti", "utm", "utp", "uts", "utt", "utw",
@@ -88,12 +98,17 @@ func readArtifact(path string) ([]erf.Resource, error) {
} }
if len(data) >= 3 { if len(data) >= 3 {
switch string(data[:3]) { switch string(data[:3]) {
case "ERF", "HAK", "MOD": case "ERF", "HAK":
archive, err := erf.Read(bytes.NewReader(data)) archive, err := erf.Read(bytes.NewReader(data))
if err != nil { if err != nil {
return nil, fmt.Errorf("%s: %w", path, err) return nil, fmt.Errorf("%s: %w", path, err)
} }
return archive.Resources, nil 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) base := filepath.Base(path)
@@ -112,12 +127,8 @@ func readArtifact(path string) ([]erf.Resource, error) {
func emitResources(resources []erf.Resource, outDir string) ([]Entry, int, int64, error) { func emitResources(resources []erf.Resource, outDir string) ([]Entry, int, int64, error) {
// A resref appearing twice inside one artifact resolves to the last one, // A resref appearing twice inside one artifact resolves to the last one,
// the way resman lets the last container added win. // the way resman lets the last container added win.
type key struct { order := make([]Identity, 0, len(resources))
resref string latest := make(map[Identity]erf.Resource, len(resources))
restype uint16
}
order := make([]key, 0, len(resources))
latest := make(map[key]erf.Resource, len(resources))
var tooBig []string var tooBig []string
for _, resource := range resources { for _, resource := range resources {
if _, ok := erf.ExtensionForResourceType(resource.Type); !ok { if _, ok := erf.ExtensionForResourceType(resource.Type); !ok {
@@ -130,7 +141,7 @@ func emitResources(resources []erf.Resource, outDir string) ([]Entry, int, int64
tooBig = append(tooBig, fmt.Sprintf("%s: %d bytes > %d", resource.Name, len(resource.Data), fileSizeLimit)) tooBig = append(tooBig, fmt.Sprintf("%s: %d bytes > %d", resource.Name, len(resource.Data), fileSizeLimit))
continue continue
} }
identity := key{resref: strings.ToLower(resource.Name), restype: resource.Type} identity := Identity{ResRef: strings.ToLower(resource.Name), ResType: resource.Type}
if _, seen := latest[identity]; !seen { if _, seen := latest[identity]; !seen {
order = append(order, identity) order = append(order, identity)
} }
@@ -150,8 +161,8 @@ func emitResources(resources []erf.Resource, outDir string) ([]Entry, int, int64
entries = append(entries, Entry{ entries = append(entries, Entry{
SHA1: sum, SHA1: sum,
Size: uint32(len(resource.Data)), Size: uint32(len(resource.Data)),
ResRef: identity.resref, ResRef: identity.ResRef,
ResType: identity.restype, ResType: identity.ResType,
}) })
written, err := writeBlob(outDir, sum, resource.Data) written, err := writeBlob(outDir, sum, resource.Data)
if err != nil { if err != nil {
@@ -165,6 +176,17 @@ func emitResources(resources []erf.Resource, outDir string) ([]Entry, int, int64
return entries, blobs, onDiskBytes, nil 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, // 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 // or 0 if the blob already existed. Blob names are content hashes, so an
// existing name is existing content. // existing name is existing content.
@@ -210,8 +232,9 @@ func writeManifestPair(manifestPath string, data []byte, entries []Entry, onDisk
sidecar.TotalFiles = len(entries) sidecar.TotalFiles = len(entries)
sidecar.TotalBytes = totalBytes sidecar.TotalBytes = totalBytes
sidecar.OnDiskBytes = onDiskBytes sidecar.OnDiskBytes = onDiskBytes
sidecar.Created = time.Now().Unix() sidecar.Created = created()
sidecar.CreatedWith = buildinfo.String() sidecar.CreatedWith = buildinfo.String()
sidecar.EmitterVersion = emitterVersion
body, err := marshalSidecar(sidecar) body, err := marshalSidecar(sidecar)
if err != nil { if err != nil {
+14 -3
View File
@@ -6,6 +6,7 @@ import (
"encoding/hex" "encoding/hex"
"encoding/json" "encoding/json"
"fmt" "fmt"
"io"
"path/filepath" "path/filepath"
"sort" "sort"
"strings" "strings"
@@ -36,6 +37,15 @@ type Entry struct {
func (e Entry) sha1Hex() string { return hex.EncodeToString(e.SHA1[:]) } 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. // writeManifest serialises entries into NSYM v3 bytes.
func writeManifest(entries []Entry) ([]byte, error) { func writeManifest(entries []Entry) ([]byte, error) {
sorted := make([]Entry, len(entries)) sorted := make([]Entry, len(entries))
@@ -86,7 +96,7 @@ func writeManifest(entries []Entry) ([]byte, error) {
func readManifest(data []byte) ([]Entry, error) { func readManifest(data []byte) ([]Entry, error) {
reader := bytes.NewReader(data) reader := bytes.NewReader(data)
magic := make([]byte, 4) magic := make([]byte, 4)
if _, err := reader.Read(magic); err != nil || string(magic) != "NSYM" { if _, err := io.ReadFull(reader, magic); err != nil || string(magic) != "NSYM" {
return nil, fmt.Errorf("not a manifest (invalid magic bytes)") return nil, fmt.Errorf("not a manifest (invalid magic bytes)")
} }
var version, entryCount, mappingCount uint32 var version, entryCount, mappingCount uint32
@@ -102,7 +112,7 @@ func readManifest(data []byte) ([]Entry, error) {
entries := make([]Entry, 0, entryCount+mappingCount) entries := make([]Entry, 0, entryCount+mappingCount)
for i := uint32(0); i < entryCount; i++ { for i := uint32(0); i < entryCount; i++ {
var entry Entry var entry Entry
if _, err := reader.Read(entry.SHA1[:]); err != nil { if _, err := io.ReadFull(reader, entry.SHA1[:]); err != nil {
return nil, fmt.Errorf("truncated entry %d: %w", i, err) return nil, fmt.Errorf("truncated entry %d: %w", i, err)
} }
if err := binary.Read(reader, binary.LittleEndian, &entry.Size); err != nil { if err := binary.Read(reader, binary.LittleEndian, &entry.Size); err != nil {
@@ -135,7 +145,7 @@ func readManifest(data []byte) ([]Entry, error) {
func readResRef(reader *bytes.Reader) (string, uint16, error) { func readResRef(reader *bytes.Reader) (string, uint16, error) {
raw := make([]byte, resRefBytes) raw := make([]byte, resRefBytes)
if _, err := reader.Read(raw); err != nil { if _, err := io.ReadFull(reader, raw); err != nil {
return "", 0, err return "", 0, err
} }
var restype uint16 var restype uint16
@@ -171,6 +181,7 @@ type Sidecar struct {
OnDiskBytes int64 `json:"on_disk_bytes"` OnDiskBytes int64 `json:"on_disk_bytes"`
Created int64 `json:"created"` Created int64 `json:"created"`
CreatedWith string `json:"created_with"` CreatedWith string `json:"created_with"`
EmitterVersion string `json:"emitter_version,omitempty"`
GroupID int `json:"group_id,omitempty"` GroupID int `json:"group_id,omitempty"`
} }
+40 -1
View File
@@ -361,7 +361,7 @@ func TestAssembleRefusesMismatchedEmitterVersions(t *testing.T) {
if err := json.Unmarshal(body, &sidecar); err != nil { if err := json.Unmarshal(body, &sidecar); err != nil {
t.Fatal(err) t.Fatal(err)
} }
sidecar.CreatedWith = "crucible some-other-build" sidecar.EmitterVersion = "0"
patched, err := marshalSidecar(sidecar) patched, err := marshalSidecar(sidecar)
if err != nil { if err != nil {
t.Fatal(err) t.Fatal(err)
@@ -380,6 +380,45 @@ func TestAssembleRefusesMismatchedEmitterVersions(t *testing.T) {
} }
} }
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) { func TestAssembleFailsClosedOnMissingIndex(t *testing.T) {
entriesDir, _, _ := emitFixture(t) entriesDir, _, _ := emitFixture(t)
_, err := Assemble(AssembleOptions{ _, err := Assemble(AssembleOptions{