Files
artifactapi/internal/provider/alpine/alpine.go
T
unkin-agent 73c0bfc670
ci/woodpecker/tag/docker Pipeline was successful
deb/apk: make local repodata deterministic (#119)
Part of #117. Local generated repodata must be byte-identical across the two no-affinity replicas and across every regeneration, so apt/apk never hit a checksum mismatch between an index's advertised hash and the bytes actually served. This does the deb+apk half (the rpm half landed in #118).

How:
- Derive the deb `Release` `Date:` from the newest persisted `created_at` (RFC1123Z, UTC) instead of `time.Now()`; carry `created_at` through the deb metadata SELECT and `DebMetadata`. An empty repo falls back to the Unix epoch. This also stops `Date:` running ahead of wall clock.
- Pin the apk `APKINDEX` tar header `ModTime` to the Unix epoch instead of the zero-value `time.Time`, so it is never wall-clock derived.
- Give both list queries a genuine total order by adding a `file_path` tiebreak (name/version/arch is not unique).
- Add guard tests: deb generators byte-identical across generations; the `Release` checksum/size matches the served `Packages`/`Packages.gz` bytes (the exact apt invariant); `Date:` pinned to `created_at`; apk index byte-identical and tar `ModTime` pinned to epoch.

Reviewed-on: #119
Co-authored-by: unkin-agent <unkin-agent@unkin.net>
Co-committed-by: unkin-agent <unkin-agent@unkin.net>
2026-08-12 23:32:07 +10:00

410 lines
12 KiB
Go

package alpine
import (
"bufio"
"bytes"
"compress/gzip"
"context"
"crypto/sha1"
"encoding/base64"
"errors"
"fmt"
"io"
"log/slog"
"net/http"
"path"
"strconv"
"strings"
"time"
"archive/tar"
"git.unkin.net/unkin/artifactapi/internal/auth"
"git.unkin.net/unkin/artifactapi/internal/provider"
"git.unkin.net/unkin/artifactapi/internal/storage"
"git.unkin.net/unkin/artifactapi/pkg/models"
)
func init() {
provider.Register(&Provider{})
}
type Provider struct{}
func (p *Provider) Type() models.PackageType { return models.PackageAlpine }
func (p *Provider) Classify(path string) provider.Mutability {
if strings.HasSuffix(path, "APKINDEX.tar.gz") {
return provider.Mutable
}
return provider.Immutable
}
func (p *Provider) ContentType(path string) string {
if strings.HasSuffix(path, ".apk") {
return "application/vnd.android.package-archive"
}
if strings.HasSuffix(path, ".tar.gz") {
return "application/gzip"
}
return "application/octet-stream"
}
func (p *Provider) UpstreamURL(remote models.Remote, path string) string {
return strings.TrimRight(remote.BaseURL, "/") + "/" + strings.TrimLeft(path, "/")
}
func (p *Provider) RewriteResponse(_ []byte, _ models.Remote, _ string) ([]byte, error) {
return nil, nil
}
func (p *Provider) AuthHeaders(_ context.Context, remote models.Remote) (http.Header, error) {
return auth.BasicHeaders(remote), nil
}
// --- LocalUploader: hosting real .apk packages -----------------------------
// ValidateUpload accepts any *.apk and preserves the client-supplied directory
// (the arch prefix) as the storage path, since arch cannot be parsed from the
// filename alone and the generic uploader hands us only the path. apk clients
// fetch packages at <arch>/<file>.apk, so publishers upload to that same path;
// AfterUpload records the true arch (from .PKGINFO) for index filtering.
func (p *Provider) ValidateUpload(filePath string) (storagePath, contentType string, err error) {
clean := strings.TrimPrefix(path.Clean("/"+filePath), "/")
filename := clean
if i := strings.LastIndex(clean, "/"); i >= 0 {
filename = clean[i+1:]
}
if !strings.HasSuffix(strings.ToLower(filename), ".apk") {
return "", "", fmt.Errorf("file must be a .apk package")
}
return clean, "application/vnd.android.package-archive", nil
}
func (p *Provider) UploadResponse(storagePath, contentHash string, sizeBytes int64) map[string]any {
filename := storagePath
if i := strings.LastIndex(storagePath, "/"); i >= 0 {
filename = storagePath[i+1:]
}
return map[string]any{
"filename": filename,
"content_hash": contentHash,
"size_bytes": sizeBytes,
}
}
func (p *Provider) AfterUpload(ctx context.Context, repoName, storagePath, contentHash string, blobs provider.BlobReader, db provider.MetadataStore) {
s3Key := storage.BlobKey(strings.TrimPrefix(contentHash, "sha256:"))
reader, blobSize, err := blobs.Download(ctx, s3Key)
if err != nil {
slog.Error("alpine metadata: download failed", "repo", repoName, "path", storagePath, "error", err)
return
}
defer reader.Close()
raw, err := io.ReadAll(reader)
if err != nil {
slog.Error("alpine metadata: read failed", "repo", repoName, "path", storagePath, "error", err)
return
}
meta, err := parseApk(raw)
if err != nil {
slog.Error("alpine metadata: parse failed", "repo", repoName, "path", storagePath, "error", err)
return
}
meta.RepoName = repoName
meta.FilePath = storagePath
meta.ContentHash = contentHash
meta.DownloadSize = blobSize
if meta.Name == "" || meta.Arch == "" {
slog.Error("alpine metadata: .PKGINFO missing pkgname/arch", "repo", repoName, "path", storagePath)
return
}
store, ok := db.(provider.AlpineMetadataStore)
if !ok {
slog.Error("alpine metadata: store does not support alpine metadata", "repo", repoName)
return
}
if err := store.InsertAlpineMetadata(ctx, meta); err != nil {
slog.Error("alpine metadata: insert failed", "repo", repoName, "path", storagePath, "error", err)
return
}
slog.Info("alpine metadata: parsed", "repo", repoName, "name", meta.Name, "version", meta.Version, "arch", meta.Arch)
}
func (p *Provider) AfterDelete(ctx context.Context, repoName, storagePath string, db provider.MetadataDeleter) error {
deleter, ok := db.(provider.AlpineMetadataDeleter)
if !ok {
return nil
}
if err := deleter.DeleteAlpineMetadata(ctx, repoName, storagePath); err != nil {
slog.Error("alpine metadata: delete failed", "repo", repoName, "path", storagePath, "error", err)
return err
}
slog.Info("alpine metadata: deleted", "repo", repoName, "path", storagePath)
return nil
}
// --- LocalIndexer: generating a per-arch APKINDEX.tar.gz -------------------
// normalizeIndexPath collapses apk's dot-segment prefix: an /etc/apk/repositories
// line of "<url>/api/v1/local/<name>" makes apk request "./<arch>/APKINDEX.tar.gz".
// Mirrors deb's flat-repo normalization.
func normalizeIndexPath(p string) string {
return strings.TrimPrefix(path.Clean("/"+p), "/")
}
func (p *Provider) ServeLocalIndex(w http.ResponseWriter, r *http.Request, files provider.FileStore, repoName, reqPath string) bool {
clean := normalizeIndexPath(reqPath)
if !strings.HasSuffix(clean, "APKINDEX.tar.gz") {
return false
}
arch := strings.TrimSuffix(clean, "APKINDEX.tar.gz")
arch = strings.Trim(arch, "/")
if arch == "" || strings.Contains(arch, "/") {
http.Error(w, "APKINDEX must be requested per-arch: <arch>/APKINDEX.tar.gz", http.StatusNotFound)
return true
}
reader, ok := files.(provider.AlpineMetadataReader)
if !ok {
http.Error(w, "alpine metadata not available", http.StatusInternalServerError)
return true
}
metas, err := reader.ListAlpineMetadataEntries(r.Context(), repoName)
if err != nil {
if errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded) {
slog.Warn("alpine: metadata read canceled", "repo", repoName, "error", err)
http.Error(w, "metadata read canceled", http.StatusServiceUnavailable)
return true
}
http.Error(w, err.Error(), http.StatusInternalServerError)
return true
}
var filtered []provider.AlpineMetadata
for _, m := range metas {
if m.Arch == arch {
filtered = append(filtered, m)
}
}
w.Header().Set("Content-Type", "application/gzip")
w.WriteHeader(http.StatusOK)
w.Write(generateAPKIndex(filtered))
return true
}
func (p *Provider) GenerateLocalIndex(ctx context.Context, files provider.FileStore, repoName, path string) ([]byte, error) {
return nil, fmt.Errorf("alpine local index generation for virtual repos not supported")
}
// --- pure-Go .apk parsing --------------------------------------------------
// parseApk reads an .apk (up to three concatenated, independently gzipped tar
// streams: optional signature, control, data). It locates the control stream by
// its .PKGINFO member, computes the apk pull checksum C: = "Q1" +
// base64(sha1(<control gzip stream bytes>)), and reads the .PKGINFO fields.
func parseApk(raw []byte) (*provider.AlpineMetadata, error) {
members, err := gzipMembers(raw)
if err != nil {
return nil, err
}
for _, m := range members {
pkginfo, ok := pkginfoFromTar(m.tar)
if !ok {
continue
}
meta := parsePkginfo(pkginfo)
sum := sha1.Sum(m.raw)
meta.Checksum = "Q1" + base64.StdEncoding.EncodeToString(sum[:])
return meta, nil
}
return nil, errors.New("no .PKGINFO found in any .apk gzip stream")
}
type gzMember struct {
raw []byte // the raw bytes of this gzip stream (for the Q1 checksum)
tar []byte // the decompressed tar payload
}
// gzipMembers splits the concatenated gzip streams, returning each stream's raw
// bytes alongside its decompressed tar. It relies on bytes.Reader being an
// io.ByteReader (so compress/gzip does not over-read past a member's trailer)
// to recover exact stream boundaries via Multistream(false)+Reset.
func gzipMembers(data []byte) ([]gzMember, error) {
br := bytes.NewReader(data)
zr, err := gzip.NewReader(br)
if err != nil {
return nil, err
}
var members []gzMember
prev := 0
for {
zr.Multistream(false)
out, err := io.ReadAll(zr)
if err != nil {
return nil, err
}
end := len(data) - br.Len()
members = append(members, gzMember{raw: data[prev:end], tar: out})
prev = end
if err := zr.Reset(br); err != nil {
if err == io.EOF {
break
}
return nil, err
}
}
return members, nil
}
func pkginfoFromTar(tarBytes []byte) (string, bool) {
tr := tar.NewReader(bytes.NewReader(tarBytes))
for {
hdr, err := tr.Next()
if err != nil {
return "", false
}
if strings.TrimPrefix(hdr.Name, "./") == ".PKGINFO" {
b, err := io.ReadAll(tr)
if err != nil {
return "", false
}
return string(b), true
}
}
}
// parsePkginfo reads the "key = value" .PKGINFO text, collecting the repeated
// depend/provides/install_if keys into slices.
func parsePkginfo(text string) *provider.AlpineMetadata {
m := &provider.AlpineMetadata{}
sc := bufio.NewScanner(strings.NewReader(text))
sc.Buffer(make([]byte, 0, 64*1024), 1024*1024)
for sc.Scan() {
line := strings.TrimSpace(sc.Text())
if line == "" || strings.HasPrefix(line, "#") {
continue
}
idx := strings.Index(line, "=")
if idx < 0 {
continue
}
key := strings.TrimSpace(line[:idx])
val := strings.TrimSpace(line[idx+1:])
switch key {
case "pkgname":
m.Name = val
case "pkgver":
m.Version = val
case "arch":
m.Arch = val
case "pkgdesc":
m.Description = val
case "url":
m.URL = val
case "license":
m.License = val
case "origin":
m.Origin = val
case "maintainer":
m.Maintainer = val
case "builddate":
if n, err := strconv.ParseInt(val, 10, 64); err == nil {
m.BuildTime = n
}
case "commit":
m.Commit = val
case "size":
if n, err := strconv.ParseInt(val, 10, 64); err == nil {
m.InstalledSize = n
}
case "provider_priority":
m.ProviderPriority = val
case "depend":
if val != "" {
m.Depends = append(m.Depends, val)
}
case "provides":
if val != "" {
m.Provides = append(m.Provides, val)
}
case "install_if":
if val != "" {
m.InstallIf = append(m.InstallIf, val)
}
}
}
return m
}
// generateAPKIndex builds the APKINDEX.tar.gz = gzip(tar(APKINDEX)) for the
// given (already arch-filtered) rows. Records are blank-line separated; fields
// follow the canonical C/P/V/A/S/I/T/U/L/o/m/t/c/k/D/p/i order and empties are
// omitted. Unsigned (clients use --allow-untrusted), matching rpm gpgcheck=0.
func generateAPKIndex(metas []provider.AlpineMetadata) []byte {
var idx bytes.Buffer
for i, m := range metas {
if i > 0 {
idx.WriteString("\n")
}
writeField(&idx, "C", m.Checksum)
writeField(&idx, "P", m.Name)
writeField(&idx, "V", m.Version)
writeField(&idx, "A", m.Arch)
writeField(&idx, "S", intField(m.DownloadSize))
writeField(&idx, "I", intField(m.InstalledSize))
writeField(&idx, "T", m.Description)
writeField(&idx, "U", m.URL)
writeField(&idx, "L", m.License)
writeField(&idx, "o", m.Origin)
writeField(&idx, "m", m.Maintainer)
writeField(&idx, "t", intField(m.BuildTime))
writeField(&idx, "c", m.Commit)
writeField(&idx, "k", m.ProviderPriority)
writeField(&idx, "D", strings.Join(m.Depends, " "))
writeField(&idx, "p", strings.Join(m.Provides, " "))
writeField(&idx, "i", strings.Join(m.InstallIf, " "))
}
var tarBuf bytes.Buffer
tw := tar.NewWriter(&tarBuf)
body := idx.Bytes()
// ModTime is pinned to the Unix epoch (never wall clock) so APKINDEX.tar.gz
// is byte-identical across replicas and regenerations (issue #117); apk
// clients ignore the tar mtime.
tw.WriteHeader(&tar.Header{Name: "APKINDEX", Mode: 0o644, Size: int64(len(body)), Typeflag: tar.TypeReg, ModTime: time.Unix(0, 0)})
tw.Write(body)
tw.Close()
var gzBuf bytes.Buffer
gz := gzip.NewWriter(&gzBuf)
gz.Write(tarBuf.Bytes())
gz.Close()
return gzBuf.Bytes()
}
func writeField(b *bytes.Buffer, key, val string) {
if val == "" {
return
}
b.WriteString(key)
b.WriteString(":")
b.WriteString(val)
b.WriteString("\n")
}
func intField(n int64) string {
if n == 0 {
return ""
}
return strconv.FormatInt(n, 10)
}