Add deb virtual repositories (#138)
ci/woodpecker/tag/docker Pipeline was successful

Virtual repos merge helm, pypi, rpm, npm, gems, ansible, goproxy and puppet but not deb, so several apt repos (e.g. github_deb remotes plus a local deb) cannot be served as one apt source.

- merge member Packages for flat (`./`) and `dists/<suite>` layouts; member order wins duplicate (Package, Version, Architecture)
- generate an unsigned Release with Acquire-By-Hash, serving indexes by name and by SHA256 from both retained generations
- prefix Filename with the member name and 302 downloads to the member route
- cache merges per suite on the shared virtual cache, capped at 32 suites per virtual (LRU)
- add docs/deb/virtual.md, unit tests and dockerised apt e2e for flat and suite layouts

Reviewed-on: #138
Co-authored-by: unkin-agent <unkin-agent@unkin.net>
Co-committed-by: unkin-agent <unkin-agent@unkin.net>
This commit was merged in pull request #138.
This commit is contained in:
2026-10-24 16:21:33 +11:00
committed by BenVincent
parent ca269c5046
commit 7fa5434dbc
7 changed files with 840 additions and 10 deletions
+107 -9
View File
@@ -9,6 +9,7 @@ import (
"net/http"
"net/http/httptest"
"net/url"
"path"
"slices"
"strings"
"sync"
@@ -26,6 +27,7 @@ type Engine struct {
proxyEngine *proxy.Engine
getRemote func(context.Context, string) (*models.Remote, error)
rpmMember func(context.Context, string) (*RPMMember, error)
debMember func(ctx context.Context, name, base string) (*DebMember, error)
npmMember func(ctx context.Context, name, pkg string) (*NPMMember, error)
// NPMLocal returns a local npm repo's packument with tarball paths relative
@@ -62,19 +64,25 @@ type mergedRepo interface {
func (r *RPMRepo) index() []byte { return r.Repomd }
func (r *RPMRepo) files() map[string][]byte { return r.Files }
func (r *DebRepo) index() []byte { return r.Release }
func (r *DebRepo) files() map[string][]byte { return r.Files }
// repoGen holds a virtual's current merge plus the previous one, so a client
// holding the previous index can still fetch the files it references.
type repoGen struct {
cur, prev mergedRepo
at time.Time
at, used time.Time
}
const rpmMergeTTL = 60 * time.Second
// maxDebSuites bounds the client-chosen suites cached per deb virtual.
const maxDebSuites = 32
func NewEngine(db *database.DB, proxyEngine *proxy.Engine) *Engine {
e := &Engine{db: db, proxyEngine: proxyEngine, getRemote: db.GetRemote, mergeTTL: rpmMergeTTL}
e.rpmMember = e.fetchRPMMember
e.debMember = e.fetchDebMember
e.npmMember = e.fetchNPMMember
e.gemsMember = e.fetchGemsMember
e.ansibleDoc = e.ansibleMemberDoc
@@ -84,8 +92,11 @@ func NewEngine(db *database.DB, proxyEngine *proxy.Engine) *Engine {
}
func (e *Engine) Fetch(ctx context.Context, virt models.Virtual, path string, proxyBaseURL string) ([]byte, string, error) {
if virt.PackageType == models.PackageRPM {
switch virt.PackageType {
case models.PackageRPM:
return e.fetchRPM(ctx, virt, path)
case models.PackageDeb:
return e.fetchDeb(ctx, virt, path)
}
merger, err := GetMerger(virt.PackageType)
@@ -204,16 +215,22 @@ var (
ErrBadPath = errors.New("invalid path")
)
// MemberRedirect maps an rpm, npm or ansible virtual package path
// (<member>/<href>, as written by MergeRPM/MergeNPM/AnsibleVersion) to the
// owning member's route. ok is false for any other path; ErrBadPath means the
// path is absolute or escapes the member.
// MemberRedirect maps an rpm, deb, npm or ansible virtual package path
// (<member>/<href>, as written by MergeRPM/MergeDeb/MergeNPM/AnsibleVersion)
// to the owning member's route. ok is false for any other path; ErrBadPath
// means the path is absolute or escapes the member.
func (e *Engine) MemberRedirect(ctx context.Context, virt models.Virtual, path, rawQuery, proxyBaseURL string) (string, bool, error) {
switch virt.PackageType {
case models.PackageRPM:
if strings.HasPrefix(path, "repodata/") {
return "", false, nil
}
case models.PackageDeb:
// apt requests a flat repo's files under its "./" dist verbatim.
path = strings.TrimPrefix(path, "./")
if strings.HasPrefix(path, "dists/") {
return "", false, nil
}
case models.PackageNPM, models.PackageAnsible:
default:
return "", false, nil
@@ -294,6 +311,7 @@ func (e *Engine) merged(ctx context.Context, key string, merge func(context.Cont
e.mu.Lock()
g := e.gens[key]
if g != nil && time.Since(g.at) < e.mergeTTL {
g.used = time.Now()
e.mu.Unlock()
return g.cur, nil
}
@@ -312,12 +330,12 @@ func (e *Engine) merged(ctx context.Context, key string, merge func(context.Cont
g := e.gens[key]
switch {
case g == nil:
e.gens[key] = &repoGen{cur: repo, at: time.Now()}
e.gens[key] = &repoGen{cur: repo, at: time.Now(), used: time.Now()}
case string(g.cur.index()) == string(repo.index()):
g.at = time.Now()
g.at, g.used = time.Now(), time.Now()
repo = g.cur
default:
g.prev, g.cur, g.at = g.cur, repo, time.Now()
g.prev, g.cur, g.at, g.used = g.cur, repo, time.Now(), time.Now()
}
return repo, nil
})
@@ -364,6 +382,85 @@ func fetchMembers[T any](virt models.Virtual, fetch func(name string) (*T, error
return members, nil
}
func (e *Engine) fetchDeb(ctx context.Context, virt models.Virtual, reqPath string) ([]byte, string, error) {
base, name, ok := debIndexPath(strings.TrimPrefix(path.Clean("/"+reqPath), "/"))
if !ok {
return nil, "", ErrNotFound
}
key := "deb\x00" + virt.Name + "\x00" + base
if name != "Release" {
if body, ok := e.cachedFile(key, name); ok {
return body, debContentType(name), nil
}
}
repo, err := e.merged(ctx, key, func(ctx context.Context) (mergedRepo, error) {
members, err := fetchMembers(virt, func(name string) (*DebMember, error) { return e.debMember(ctx, name, base) })
if err != nil {
return nil, err
}
return MergeDeb(members), nil
})
if err != nil {
return nil, "", err
}
e.evictLRU("deb\x00"+virt.Name+"\x00", maxDebSuites)
if name == "Release" {
return repo.index(), "text/plain", nil
}
if body, ok := repo.files()[name]; ok {
return body, debContentType(name), nil
}
return nil, "", ErrNotFound
}
func debContentType(name string) string {
if strings.HasSuffix(name, ".gz") {
return "application/gzip"
}
return "text/plain"
}
// fetchDebMember reads a member's Release under base and each binary Packages
// index it lists.
func (e *Engine) fetchDebMember(ctx context.Context, name, base string) (*DebMember, error) {
remote, err := e.getRemote(ctx, name)
if err != nil {
return nil, fmt.Errorf("remote %q: %w", name, err)
}
rel, err := e.fetchMemberPath(ctx, *remote, base+"Release")
if err != nil {
return nil, err
}
m := &DebMember{RemoteName: name, Release: rel, Packages: map[string][]byte{}}
for dir, file := range debIndexFiles(parseDebRelease(rel)) {
raw, err := e.fetchMemberPath(ctx, *remote, base+file)
if err != nil {
return nil, err
}
if m.Packages[dir], err = decompress(file, raw); err != nil {
return nil, fmt.Errorf("decompress %s: %w", file, err)
}
}
return m, nil
}
// evictLRU drops the least recently used merges under prefix beyond keep.
func (e *Engine) evictLRU(prefix string, keep int) {
e.mu.Lock()
defer e.mu.Unlock()
var keys []string
for k := range e.gens {
if strings.HasPrefix(k, prefix) {
keys = append(keys, k)
}
}
slices.SortFunc(keys, func(a, b string) int { return e.gens[b].used.Compare(e.gens[a].used) })
for _, k := range keys[min(len(keys), keep):] {
delete(e.gens, k)
}
}
func (e *Engine) cachedFile(key, path string) ([]byte, bool) {
e.mu.Lock()
defer e.mu.Unlock()
@@ -376,6 +473,7 @@ func (e *Engine) cachedFile(key, path string) ([]byte, bool) {
continue
}
if body, ok := r.files()[path]; ok {
g.used = time.Now()
return body, true
}
}