Compare commits

...

2 Commits

Author SHA1 Message Date
unkin-agent 2905ec99e4 Add vcs remote type (#162)
ci/woodpecker/tag/docker Pipeline was successful
The generic `github` remote only allowlists URL patterns: every repo needs hand-written mutable patterns for branch archives, and a moved branch re-downloads the whole archive under a ref-named key.

- add `vcs` package type for GitHub or Gitea (picked from `base_url`) tag, commit and release-download paths, cached immutable
- redirect `archive/refs/heads/<branch>` to the commit archive, resolving the branch via the branch API with one shared, per-remote bounded lookup cache reused for `mutable_ttl`
- cache missing branches for at least 60s and map API rate limits to `503` with `Retry-After`
- reject any other path shape, escaping segments and denied redirect targets
- add `docs/vcs/remote.md`, a UI type def and a vcs e2e caching case

Reviewed-on: #162
Co-authored-by: unkin-agent <unkin-agent@unkin.net>
Co-committed-by: unkin-agent <unkin-agent@unkin.net>
2026-10-24 17:21:05 +11:00
unkin-agent 7fa5434dbc 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>
2026-10-24 16:21:33 +11:00
17 changed files with 1602 additions and 14 deletions
+2 -1
View File
@@ -29,7 +29,7 @@ Each type's paths, Terraform examples and local/virtual behaviour are in `docs/<
- `alpine`: [remote](docs/alpine/remote.md), [remote-github](docs/alpine/remote-github.md)
- `ansible`: [remote](docs/ansible/remote.md), [local](docs/ansible/local.md), [virtual](docs/ansible/virtual.md)
- `cargo`: [remote](docs/cargo/remote.md)
- `deb`: [remote](docs/deb/remote.md), [remote-github](docs/deb/remote-github.md)
- `deb`: [remote](docs/deb/remote.md), [remote-github](docs/deb/remote-github.md), [virtual](docs/deb/virtual.md)
- `docker`: [remote](docs/docker/remote.md), [local](docs/docker/local.md)
- `generic`: [remote](docs/generic/remote.md)
- `gems`: [remote](docs/gems/remote.md), [local](docs/gems/local.md), [virtual](docs/gems/virtual.md)
@@ -41,6 +41,7 @@ Each type's paths, Terraform examples and local/virtual behaviour are in `docs/<
- `pypi`: [remote](docs/pypi/remote.md)
- `rpm`: [remote](docs/rpm/remote.md), [remote-github](docs/rpm/remote-github.md)
- `terraform`: [remote](docs/terraform/remote.md), [local](docs/terraform/local.md)
- `vcs`: [remote](docs/vcs/remote.md)
### GitHub authentication
+1 -1
View File
@@ -98,4 +98,4 @@ Each type's page lists its auto-detected mutable and immutable paths:
[docker](../docker/remote.md), [generic](../generic/remote.md), [gems](../gems/remote.md), [goproxy](../goproxy/remote.md),
[helm](../helm/remote.md), [npm](../npm/remote.md), [nuget](../nuget/remote.md),
[puppet](../puppet/remote.md), [pypi](../pypi/remote.md), [rpm](../rpm/remote.md),
[rpm github](../rpm/remote-github.md), [terraform](../terraform/remote.md).
[rpm github](../rpm/remote-github.md), [terraform](../terraform/remote.md), [vcs](../vcs/remote.md).
+32
View File
@@ -0,0 +1,32 @@
# `deb` virtual
A deb virtual serves one apt source over local and remote deb members at
`/api/v1/virtual/<name>/`, in either the flat (`./`) or `dists/<suite>` layout.
- Packages indexes are merged; on a duplicate (Package, Version, Architecture)
the earlier member wins.
- The Release is generated unsigned with `Acquire-By-Hash: yes`; indexes are
served by name and by SHA256 from the current and previous merge.
- `Filename` points at the virtual (`<member>/<path>`), which 302s to the
owning member.
- All members must share one layout; mixing flat and suite members is
unsupported (suite requests fail with 502).
- A member error fails the request with 502. Merges are cached for 60s per
suite, up to 32 suites per virtual.
```hcl
resource "artifactapi_virtual" "deb" {
name = "deb"
package_type = "deb"
members = [artifactapi_local_deb.internal.name, artifactapi_remote_github_deb.tool.name]
}
```
```
# /etc/apt/sources.list.d/artifactapi.list (flat)
deb [trusted=yes] https://artifactapi.example.com/api/v1/virtual/deb/ ./
# suite layout
deb [trusted=yes] https://artifactapi.example.com/api/v1/virtual/debian/ trixie main
```
Upload to the local member directly; the virtual is read-only.
+29
View File
@@ -0,0 +1,29 @@
# `vcs` remote
Shared classification, TTL and eviction rules: [caching](../common/caching.md).
Caches source archives and release assets from GitHub (`base_url = "https://github.com"`) or Gitea (any other `base_url`). GitHub Enterprise is not supported: a non-github.com host is treated as Gitea. Only the paths below are served; anything else is a 404.
| Path | Behaviour |
|---|---|
| `<owner>/<repo>/archive/refs/tags/<tag>.(tar.gz\|zip)` | immutable |
| `<owner>/<repo>/archive/<40-char sha>.(tar.gz\|zip)` | immutable |
| `<owner>/<repo>/releases/download/<tag>/<asset>` | immutable |
| `<owner>/<repo>/releases/latest/download/<asset>` | mutable |
| `<owner>/<repo>/archive/refs/heads/<branch>.(tar.gz\|zip)` | `302` to the archive of the branch's current commit, resolved via the provider's branch API and reused for `mutable_ttl` |
Branch archives redirect, so `curl` needs `-L`; pip and npm follow redirects. The archive is the host's commit archive (GitHub names its top-level directory `<repo>-<sha>`). `patterns` and `blocklist` match both the branch path and the commit path it redirects to, so allowlist repos by prefix (`^cli/cli/`); a branch whose commit path is denied is a `403`. A missing branch is a `404`, cached for `mutable_ttl` (at least 60s). An API rate limit (`429`, or `403` with rate-limit headers) is a `503` with `Retry-After`, unless `stale_on_error` serves the last known commit. Each remote caches up to 4096 branch lookups. GitHub API calls use the [server GitHub credential](../../README.md#github-authentication); Gitea API calls use the remote's `username`/`password`.
```hcl
resource "artifactapi_remote_vcs" "github" {
name = "github-vcs"
base_url = "https://github.com"
mutable_ttl = 600
patterns = ["^cli/cli/", "^psf/requests/"]
}
```
```bash
curl -fsSLO https://artifactapi.example/api/v1/remote/github-vcs/cli/cli/archive/refs/tags/v2.60.0.tar.gz
pip install https://artifactapi.example/api/v1/remote/github-vcs/psf/requests/archive/refs/tags/v2.32.3.tar.gz
```
+7
View File
@@ -0,0 +1,7 @@
//go:build dockere2e
package e2edocker
func init() {
cachingCases = append(cachingCases, cachingCase{"vcs", "octo/repo/archive/refs/tags/v1.0.0.tar.gz", "octo/repo/archive/refs/tags/v1.0.0.tar.gz"})
}
+115
View File
@@ -0,0 +1,115 @@
//go:build dockere2e
package e2edocker
import (
"net/http"
"os"
"os/exec"
"strings"
"testing"
"time"
"git.unkin.net/unkin/artifactapi/internal/testsupport"
)
// TestVirtualDebMerge merges two local deb repos carrying the same package: the
// first member wins the duplicate, and a real apt sees both members' packages
// and installs through the virtual.
func TestVirtualDebMerge(t *testing.T) {
createRepo(t, `{"name":"vdeb-a","package_type":"deb","repo_type":"local"}`)
createRepo(t, `{"name":"vdeb-b","package_type":"deb","repo_type":"local"}`)
defer deleteRepo(t, "vdeb-a")
defer deleteRepo(t, "vdeb-b")
uploadFile(t, "vdeb-a", "vdeb-one_1.0.0_amd64.deb", testsupport.MinimalDeb("vdeb-one", "1.0.0", "amd64"), "application/vnd.debian.binary-package")
uploadFile(t, "vdeb-b", "vdeb-one_1.0.0_amd64.deb", testsupport.MinimalDeb("vdeb-one", "1.0.0", "amd64"), "application/vnd.debian.binary-package")
uploadFile(t, "vdeb-b", "vdeb-two_2.0.0_amd64.deb", testsupport.MinimalDeb("vdeb-two", "2.0.0", "amd64"), "application/vnd.debian.binary-package")
// Local indexes regenerate asynchronously; merge only once vdeb-b lists both.
deadline := time.Now().Add(15 * time.Second)
for {
resp, body := doRequest(t, http.MethodGet, api("/api/v1/local/vdeb-b/Packages"), nil, "")
if resp.StatusCode == http.StatusOK && strings.Contains(string(body), "vdeb-two") || time.Now().After(deadline) {
break
}
time.Sleep(500 * time.Millisecond)
}
createVirtual(t, `{"name":"vdeb","package_type":"deb","members":["vdeb-a","vdeb-b"]}`)
defer deleteVirtual(t, "vdeb")
_, body := doRequest(t, http.MethodGet, api("/api/v1/virtual/vdeb/Packages"), nil, "")
pk := string(body)
if strings.Count(pk, "Package: vdeb-one\n") != 1 || !strings.Contains(pk, "Filename: vdeb-a/pool/vdeb-one_1.0.0_amd64.deb") ||
!strings.Contains(pk, "Filename: vdeb-b/pool/vdeb-two_2.0.0_amd64.deb") {
t.Fatalf("merged Packages should hold vdeb-one from the first member and vdeb-two:\n%s", pk)
}
network := os.Getenv("COMPOSE_NETWORK")
internal := os.Getenv("ARTIFACTAPI_INTERNAL")
if network == "" || internal == "" {
t.Skip("COMPOSE_NETWORK/ARTIFACTAPI_INTERNAL not set; skipping real apt")
}
if _, err := exec.LookPath("docker"); err != nil {
t.Skip("docker not available on the test host")
}
script := "set -euo pipefail; " +
"rm -f /etc/apt/sources.list.d/*; " +
"echo \"deb [trusted=yes] $URL ./\" > /etc/apt/sources.list.d/vdeb.list; " +
"apt-get update; " +
"apt-cache policy vdeb-one vdeb-two | tee /dev/stderr | grep -c virtual/vdeb | grep -qx 2; " +
"apt-get install -y vdeb-one vdeb-two; " +
"dpkg -s vdeb-one vdeb-two"
out, err := exec.Command("docker", "run", "--rm", "--network", network,
"-e", "URL="+strings.TrimRight(internal, "/")+"/api/v1/virtual/vdeb/",
"debian:13", "bash", "-c", script).CombinedOutput()
if err != nil {
t.Fatalf("real apt against deb virtual failed: %v\n%s", err, out)
}
}
// TestVirtualDebSuite merges two proxied deb.debian.org remotes and drives a
// real apt through the virtual's dists/<suite> layout (by-hash indexes).
func TestVirtualDebSuite(t *testing.T) {
network := os.Getenv("COMPOSE_NETWORK")
internal := os.Getenv("ARTIFACTAPI_INTERNAL")
if network == "" || internal == "" {
t.Skip("COMPOSE_NETWORK/ARTIFACTAPI_INTERNAL not set; skipping real apt")
}
if _, err := exec.LookPath("docker"); err != nil {
t.Skip("docker not available on the test host")
}
for _, n := range []string{"vdebs-a", "vdebs-b"} {
createRepo(t, `{"name":"`+n+`","package_type":"deb","repo_type":"remote","base_url":"http://deb.debian.org/debian"}`)
defer deleteRepo(t, n)
}
createVirtual(t, `{"name":"vdebs","package_type":"deb","members":["vdebs-a","vdebs-b"]}`)
defer deleteVirtual(t, "vdebs")
resp, rel := doRequest(t, http.MethodGet, api("/api/v1/virtual/vdebs/dists/trixie-updates/Release"), nil, "")
if resp.StatusCode != http.StatusOK || !strings.Contains(string(rel), "Acquire-By-Hash: yes\n") ||
!strings.Contains(string(rel), " main/binary-amd64/Packages.gz\n") {
t.Fatalf("merged suite Release: status %d:\n%s", resp.StatusCode, rel)
}
script := "set -euo pipefail; " +
"rm -f /etc/apt/sources.list.d/*; " +
"echo \"deb [trusted=yes] $URL trixie-updates main\" > /etc/apt/sources.list.d/vdebs.list; " +
"apt-get update; " +
"idx=/tmp/avail; apt-cache dumpavail > $idx; " +
"grep -q '^Filename: vdebs-a/pool/' $idx; if grep -q '^Filename: vdebs-b/' $idx; then echo 'vdebs-b not deduped'; exit 1; fi; " +
"pkg=$(awk '/^Package:/{p=$2} /^Size:/{if(!m||$2<m){m=$2;b=p}} END{print b}' $idx); " +
"cd /tmp && apt-get download $pkg && ls -l /tmp/*.deb"
out, err := exec.Command("docker", "run", "--rm", "--network", network,
"-e", "URL="+strings.TrimRight(internal, "/")+"/api/v1/virtual/vdebs/",
"debian:13", "bash", "-c", script).CombinedOutput()
if err != nil {
t.Fatalf("real apt against deb virtual suite failed: %v\n%s", err, out)
}
t.Logf("%s", out)
resp, _ = doRequest(t, http.MethodGet, api("/api/v1/remote/vdebs-a/dists/trixie-updates/main/binary-amd64/Packages.xz"), nil, "")
if src := resp.Header.Get("X-Artifact-Source"); resp.StatusCode != http.StatusOK || src != "cache" {
t.Fatalf("member index the virtual merged should be cached: status %d source %q", resp.StatusCode, src)
}
}
+3
View File
@@ -0,0 +1,3 @@
package all
import _ "git.unkin.net/unkin/artifactapi/internal/provider/vcs"
+3 -3
View File
@@ -71,7 +71,7 @@ func (p *Provider) AuthHeaders(ctx context.Context, remote models.Remote) (http.
h.Set("Authorization", "Basic "+base64.StdEncoding.EncodeToString([]byte(remote.Username+":"+remote.Password)))
return h, nil
}
if isGitHubHost(remote.BaseURL) {
if IsGitHubHost(remote.BaseURL) {
if c := githubauth.Server(); c != nil {
tok, err := c.Token(ctx)
if err != nil {
@@ -85,11 +85,11 @@ func (p *Provider) AuthHeaders(ctx context.Context, remote models.Remote) (http.
return h, nil
}
// isGitHubHost reports whether rawURL targets a GitHub API/download host that
// IsGitHubHost reports whether rawURL targets a GitHub API/download host that
// accepts the server credential. objects.githubusercontent.com is deliberately
// excluded: release-asset downloads 302-redirect there with a pre-signed URL
// that must not carry an Authorization header.
func isGitHubHost(rawURL string) bool {
func IsGitHubHost(rawURL string) bool {
u, err := url.Parse(rawURL)
if err != nil {
return false
+309
View File
@@ -0,0 +1,309 @@
// Package vcs caches source archives and release assets from GitHub or Gitea,
// keyed by ref: tags, commits and release assets are immutable, and branch
// archives redirect to the archive of the commit the branch currently points at.
package vcs
import (
"context"
"encoding/json"
"errors"
"fmt"
"io"
"net/http"
"net/url"
"regexp"
"strconv"
"strings"
"sync"
"time"
"golang.org/x/sync/singleflight"
"git.unkin.net/unkin/artifactapi/internal/provider"
"git.unkin.net/unkin/artifactapi/internal/provider/generic"
"git.unkin.net/unkin/artifactapi/internal/proxy"
"git.unkin.net/unkin/artifactapi/pkg/models"
)
func init() {
provider.Register(New())
}
type kind int
const (
tagArchive kind = iota + 1
branchArchive
commitArchive
releaseAsset
latestReleaseAsset
)
// target is a parsed vcs request path.
type target struct {
Kind kind
Owner, Repo string
Name string // tag, branch, commit or asset
Ext string
}
var (
segmentRe = regexp.MustCompile(`^[A-Za-z0-9._+-]+$`)
commitRe = regexp.MustCompile(`^[0-9a-f]{40}$`)
)
var errNotFound = errors.New("ref not found")
const (
maxRefs = 4096
negativeTTL = time.Minute
)
// Provider embeds generic for upstream URLs, content types and GitHub/Basic auth.
type Provider struct {
generic.Provider
client *http.Client
githubAPI string
group singleflight.Group
mu sync.Mutex
refs map[string]map[string]resolved
}
// resolved is a branch lookup; an empty commit records a missing branch.
type resolved struct {
commit string
at time.Time
}
func New() *Provider {
return &Provider{
client: &http.Client{Timeout: 30 * time.Second},
githubAPI: "https://api.github.com",
refs: map[string]map[string]resolved{},
}
}
func (p *Provider) Type() models.PackageType { return models.PackageVCS }
func (p *Provider) Classify(path string) provider.Mutability {
if ref, ok := parsePath(path); ok && (ref.Kind == branchArchive || ref.Kind == latestReleaseAsset) {
return provider.Mutable
}
return provider.Immutable
}
// parsePath accepts only the archive and release-download shapes, rejecting
// any segment that could escape owner/repo or a ref.
func parsePath(path string) (target, bool) {
parts := strings.SplitN(strings.TrimPrefix(path, "/"), "/", 3)
if len(parts) != 3 || !validSegment(parts[0]) || !validSegment(parts[1]) {
return target{}, false
}
ref := target{Owner: parts[0], Repo: parts[1]}
rest := parts[2]
switch {
case strings.HasPrefix(rest, "archive/"):
name, ext, ok := archiveName(strings.TrimPrefix(rest, "archive/"))
if !ok {
return target{}, false
}
ref.Ext = ext
switch {
case strings.HasPrefix(name, "refs/tags/"):
ref.Kind, ref.Name = tagArchive, strings.TrimPrefix(name, "refs/tags/")
case strings.HasPrefix(name, "refs/heads/"):
ref.Kind, ref.Name = branchArchive, strings.TrimPrefix(name, "refs/heads/")
case commitRe.MatchString(name):
return target{Kind: commitArchive, Owner: ref.Owner, Repo: ref.Repo, Name: name, Ext: ext}, true
default:
return target{}, false
}
return ref, validRefName(ref.Name)
case strings.HasPrefix(rest, "releases/latest/download/"):
ref.Kind, ref.Name = latestReleaseAsset, strings.TrimPrefix(rest, "releases/latest/download/")
return ref, validSegment(ref.Name)
case strings.HasPrefix(rest, "releases/download/"):
tagAsset := strings.TrimPrefix(rest, "releases/download/")
i := strings.LastIndex(tagAsset, "/")
if i < 0 {
return target{}, false
}
ref.Kind, ref.Name = releaseAsset, tagAsset[i+1:]
return ref, validRefName(tagAsset[:i]) && validSegment(ref.Name)
}
return target{}, false
}
func archiveName(s string) (string, string, bool) {
for _, ext := range []string{".tar.gz", ".zip"} {
if name, ok := strings.CutSuffix(s, ext); ok && name != "" {
return name, ext, true
}
}
return "", "", false
}
func validSegment(s string) bool {
return segmentRe.MatchString(s) && s != "." && s != ".." && !strings.Contains(s, "..")
}
func validRefName(s string) bool {
for _, seg := range strings.Split(s, "/") {
if !validSegment(seg) || strings.HasPrefix(seg, ".") {
return false
}
}
return true
}
// ServeRemote rejects unsupported paths and redirects branch archives to the
// immutable archive of the branch's current commit; everything else is proxied.
func (p *Provider) ServeRemote(w http.ResponseWriter, r *http.Request, remote models.Remote, path, proxyBaseURL string, _ provider.RemoteMetadataStore) bool {
ref, ok := parsePath(path)
if !ok {
http.Error(w, "unsupported vcs path", http.StatusNotFound)
return true
}
if ref.Kind != branchArchive {
return false
}
classifier := proxy.NewClassifier(p)
if classifier.Classify(remote, path) == proxy.ClassDenied {
http.Error(w, "access denied", http.StatusForbidden)
return true
}
commit, err := p.resolve(r.Context(), remote, ref)
var se *provider.UpstreamStatusError
switch {
case errors.Is(err, errNotFound):
http.Error(w, "branch not found", http.StatusNotFound)
return true
case errors.As(err, &se) && (se.Status == http.StatusTooManyRequests || se.Status == http.StatusForbidden && !se.RetryAt.IsZero()):
secs := 60
if !se.RetryAt.IsZero() {
secs = max(1, int(time.Until(se.RetryAt).Seconds()+1))
}
w.Header().Set("Retry-After", strconv.Itoa(secs))
http.Error(w, "upstream rate limited", http.StatusServiceUnavailable)
return true
case err != nil:
http.Error(w, "bad gateway", http.StatusBadGateway)
return true
}
target := ref.Owner + "/" + ref.Repo + "/archive/" + commit + ref.Ext
if classifier.Classify(remote, target) == proxy.ClassDenied {
http.Error(w, "access denied", http.StatusForbidden)
return true
}
http.Redirect(w, r, strings.TrimRight(proxyBaseURL, "/")+provider.RemoteMountPath+remote.Name+"/"+target, http.StatusFound)
return true
}
// resolve maps a branch to its commit, reusing the answer for mutable_ttl (a
// missing branch for at least negativeTTL) and, with stale_on_error, past an
// upstream failure. Concurrent lookups of one branch share a single API call.
func (p *Provider) resolve(ctx context.Context, remote models.Remote, ref target) (string, error) {
key := ref.Owner + "/" + ref.Repo + "\x00" + ref.Name
ttl := time.Duration(remote.MutableTTL) * time.Second
p.mu.Lock()
prev, ok := p.refs[remote.Name][key]
p.mu.Unlock()
if ok && prev.commit != "" && time.Since(prev.at) < ttl {
return prev.commit, nil
}
if ok && prev.commit == "" && time.Since(prev.at) < max(ttl, negativeTTL) {
return "", errNotFound
}
v, err, _ := p.group.Do(remote.Name+"\x00"+key, func() (any, error) {
commit, err := p.lookup(context.WithoutCancel(ctx), remote, ref)
if err == nil || errors.Is(err, errNotFound) {
p.store(remote.Name, key, commit)
}
return commit, err
})
if err != nil {
if ok && prev.commit != "" && remote.StaleOnError && !errors.Is(err, errNotFound) {
return prev.commit, nil
}
return "", err
}
return v.(string), nil
}
// store records a lookup, evicting the remote's oldest entry once it holds maxRefs.
func (p *Provider) store(remote, key, commit string) {
p.mu.Lock()
defer p.mu.Unlock()
m := p.refs[remote]
if m == nil {
m = map[string]resolved{}
p.refs[remote] = m
}
if _, ok := m[key]; !ok && len(m) >= maxRefs {
var oldest string
for k, e := range m {
if oldest == "" || e.at.Before(m[oldest].at) {
oldest = k
}
}
delete(m, oldest)
}
m[key] = resolved{commit: commit, at: time.Now()}
}
func (p *Provider) lookup(ctx context.Context, remote models.Remote, ref target) (string, error) {
segs := strings.Split(ref.Name, "/")
for i, s := range segs {
segs[i] = url.PathEscape(s)
}
branch := strings.Join(segs, "/")
github := generic.IsGitHubHost(remote.BaseURL)
api := strings.TrimRight(remote.BaseURL, "/") + "/api/v1/repos/" + ref.Owner + "/" + ref.Repo + "/branches/" + branch
if github {
api = p.githubAPI + "/repos/" + ref.Owner + "/" + ref.Repo + "/git/ref/heads/" + branch
}
req, err := http.NewRequestWithContext(ctx, http.MethodGet, api, nil)
if err != nil {
return "", err
}
h, err := p.AuthHeaders(ctx, remote)
if err != nil {
return "", err
}
req.Header = h
if github {
req.Header.Set("Accept", "application/vnd.github+json")
}
resp, err := p.client.Do(req)
if err != nil {
return "", err
}
defer func() { _ = resp.Body.Close() }()
if resp.StatusCode == http.StatusNotFound || resp.StatusCode == http.StatusUnprocessableEntity {
return "", errNotFound
}
if resp.StatusCode != http.StatusOK {
return "", provider.NewUpstreamStatusError(api, resp)
}
var b struct {
Commit struct {
ID string `json:"id"`
} `json:"commit"`
Object struct {
SHA string `json:"sha"`
} `json:"object"`
}
if err := json.NewDecoder(io.LimitReader(resp.Body, 1<<20)).Decode(&b); err != nil {
return "", err
}
commit := b.Commit.ID
if github {
commit = b.Object.SHA
}
if !commitRe.MatchString(commit) {
return "", fmt.Errorf("branch lookup %s: unexpected commit %q", api, commit)
}
return commit, nil
}
+378
View File
@@ -0,0 +1,378 @@
package vcs
import (
"context"
"fmt"
"net/http"
"net/http/httptest"
"strconv"
"strings"
"sync"
"sync/atomic"
"testing"
"time"
"git.unkin.net/unkin/artifactapi/internal/githubauth"
"git.unkin.net/unkin/artifactapi/internal/provider"
"git.unkin.net/unkin/artifactapi/pkg/models"
)
const (
shaA = "4f2cfe92e3499f8581bf39db77e0a1431fa7b55c"
shaB = "7fa5434dbc083c853eef435d1915020345219d2f"
)
func TestParsePath(t *testing.T) {
tests := []struct {
path string
ok bool
kind kind
name string
}{
{"cli/cli/archive/refs/tags/v2.60.0.tar.gz", true, tagArchive, "v2.60.0"},
{"/cli/cli/archive/refs/tags/v2.60.0.zip", true, tagArchive, "v2.60.0"},
{"o/r/archive/refs/tags/release/1.0.tar.gz", true, tagArchive, "release/1.0"},
{"o/r/archive/refs/heads/feature/x.tar.gz", true, branchArchive, "feature/x"},
{"o/r/archive/" + shaA + ".tar.gz", true, commitArchive, shaA},
{"o/r/releases/download/v1.0/tool_linux_amd64.tar.gz", true, releaseAsset, "tool_linux_amd64.tar.gz"},
{"o/r/releases/latest/download/tool.zip", true, latestReleaseAsset, "tool.zip"},
{"o/r/archive/main.tar.gz", false, 0, ""},
{"o/r/archive/" + shaA[:7] + ".tar.gz", false, 0, ""},
{"o/r/archive/refs/tags/v1.tar.bz2", false, 0, ""},
{"o/r/archive/refs/tags/.tar.gz", false, 0, ""},
{"../r/archive/refs/tags/v1.tar.gz", false, 0, ""},
{"o/../archive/refs/tags/v1.tar.gz", false, 0, ""},
{"o/r/archive/refs/tags/../../../etc/passwd.tar.gz", false, 0, ""},
{"o/r/archive/refs/heads/a//b.tar.gz", false, 0, ""},
{"o/r/archive/refs/heads/.hidden.tar.gz", false, 0, ""},
{"o/r/archive/refs/heads/a..b.tar.gz", false, 0, ""},
{"o/r/archive/refs/tags/v1%2f.tar.gz", false, 0, ""},
{"o/r/releases/download/v1/../x", false, 0, ""},
{"o/r/releases/download/asset", false, 0, ""},
{"o/r/raw/main/README.md", false, 0, ""},
{"o/r", false, 0, ""},
}
for _, tt := range tests {
ref, ok := parsePath(tt.path)
if ok != tt.ok || (ok && (ref.Kind != tt.kind || ref.Name != tt.name)) {
t.Errorf("parsePath(%q) = %+v, %v; want kind %v name %q ok %v", tt.path, ref, ok, tt.kind, tt.name, tt.ok)
}
}
}
func TestClassify(t *testing.T) {
p := New()
for path, want := range map[string]provider.Mutability{
"o/r/archive/refs/tags/v1.tar.gz": provider.Immutable,
"o/r/archive/" + shaA + ".zip": provider.Immutable,
"o/r/releases/download/v1/a.bin": provider.Immutable,
"o/r/archive/refs/heads/main.tar.gz": provider.Mutable,
"o/r/releases/latest/download/webui.zip": provider.Mutable,
} {
if got := p.Classify(path); got != want {
t.Errorf("Classify(%q) = %v, want %v", path, got, want)
}
}
}
type upstream struct {
srv *httptest.Server
hits atomic.Int32
commit atomic.Value
status atomic.Int32
header atomic.Value
accept atomic.Value
auth atomic.Value
path atomic.Value
gate chan struct{}
}
func newUpstream(t *testing.T, gitea bool) *upstream {
u := &upstream{}
u.commit.Store(shaA)
u.status.Store(http.StatusOK)
u.header.Store(http.Header{})
u.srv = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
u.hits.Add(1)
if u.gate != nil {
<-u.gate
}
u.accept.Store(r.Header.Get("Accept"))
u.auth.Store(r.Header.Get("Authorization"))
u.path.Store(r.URL.EscapedPath())
for k, v := range u.header.Load().(http.Header) {
w.Header()[k] = v
}
if s := int(u.status.Load()); s != http.StatusOK {
w.WriteHeader(s)
return
}
if gitea {
_, _ = w.Write([]byte(`{"name":"main","commit":{"id":"` + u.commit.Load().(string) + `"}}`))
return
}
_, _ = w.Write([]byte(`{"ref":"refs/heads/main","object":{"sha":"` + u.commit.Load().(string) + `","type":"commit"}}`))
}))
t.Cleanup(u.srv.Close)
return u
}
func serve(p *Provider, remote models.Remote, path string) (*httptest.ResponseRecorder, bool) {
w := httptest.NewRecorder()
r := httptest.NewRequest(http.MethodGet, "/api/v1/remote/"+remote.Name+"/"+path, nil)
return w, p.ServeRemote(w, r, remote, path, "http://proxy", nil)
}
// age backdates every cached lookup of a remote by d.
func age(p *Provider, remote string, d time.Duration) {
p.mu.Lock()
defer p.mu.Unlock()
for k, e := range p.refs[remote] {
e.at = e.at.Add(-d)
p.refs[remote][k] = e
}
}
type staticCred string
func (s staticCred) Token(context.Context) (string, error) { return string(s), nil }
func TestServeRemoteGiteaBranchRedirectsToCommit(t *testing.T) {
u := newUpstream(t, true)
p := New()
remote := models.Remote{Name: "gitea", BaseURL: u.srv.URL, MutableTTL: 3600, Username: "bot", Password: "pw"}
w, handled := serve(p, remote, "unkin/app/archive/refs/heads/feature/x.zip")
if !handled || w.Code != http.StatusFound {
t.Fatalf("handled=%v code=%d", handled, w.Code)
}
if got, want := w.Header().Get("Location"), "http://proxy/api/v1/remote/gitea/unkin/app/archive/"+shaA+".zip"; got != want {
t.Errorf("Location = %q, want %q", got, want)
}
if got := u.path.Load(); got != "/api/v1/repos/unkin/app/branches/feature/x" {
t.Errorf("api path = %q", got)
}
if got := u.auth.Load(); got != "Basic Ym90OnB3" {
t.Errorf("Authorization = %q, want remote basic auth", got)
}
}
func TestServeRemoteGitHubBranchUsesAPI(t *testing.T) {
githubauth.SetServer(staticCred("ghs_server"))
t.Cleanup(func() { githubauth.SetServer(nil) })
u := newUpstream(t, false)
p := New()
p.githubAPI = u.srv.URL
for _, base := range []string{"https://github.com", "https://www.github.com/", "https://api.github.com"} {
remote := models.Remote{Name: base, BaseURL: base, MutableTTL: 3600}
w, _ := serve(p, remote, "cli/cli/archive/refs/heads/feature/x.tar.gz")
if !strings.HasSuffix(w.Header().Get("Location"), "/cli/cli/archive/"+shaA+".tar.gz") {
t.Fatalf("%s: Location = %q", base, w.Header().Get("Location"))
}
if u.path.Load() != "/repos/cli/cli/git/ref/heads/feature/x" || u.accept.Load() != "application/vnd.github+json" {
t.Errorf("%s: api path %q accept %q", base, u.path.Load(), u.accept.Load())
}
if got := u.auth.Load(); got != "Bearer ghs_server" {
t.Errorf("%s: Authorization = %q, want server credential", base, got)
}
}
}
func TestServeRemoteBranchResolutionHonoursMutableTTL(t *testing.T) {
u := newUpstream(t, true)
p := New()
remote := models.Remote{Name: "g", BaseURL: u.srv.URL, MutableTTL: 600}
path := "o/r/archive/refs/heads/main.tar.gz"
serve(p, remote, path)
u.commit.Store(shaB)
age(p, remote.Name, 599*time.Second)
w, _ := serve(p, remote, path)
if u.hits.Load() != 1 || !strings.Contains(w.Header().Get("Location"), shaA) {
t.Fatalf("within ttl: hits=%d location=%q", u.hits.Load(), w.Header().Get("Location"))
}
age(p, remote.Name, 2*time.Second)
w, _ = serve(p, remote, path)
if u.hits.Load() != 2 || !strings.Contains(w.Header().Get("Location"), shaB) {
t.Fatalf("past ttl: hits=%d location=%q", u.hits.Load(), w.Header().Get("Location"))
}
}
func TestServeRemoteBranchUpstreamFailure(t *testing.T) {
path := "o/r/archive/refs/heads/main.tar.gz"
for _, tc := range []struct {
name string
stale bool
want int
}{
{"stale", true, http.StatusFound},
{"strict", false, http.StatusBadGateway},
} {
t.Run(tc.name, func(t *testing.T) {
u := newUpstream(t, true)
p := New()
remote := models.Remote{Name: tc.name, BaseURL: u.srv.URL, StaleOnError: tc.stale}
serve(p, remote, path)
u.status.Store(http.StatusInternalServerError)
if w, _ := serve(p, remote, path); w.Code != tc.want {
t.Fatalf("code = %d, want %d", w.Code, tc.want)
}
})
}
}
func TestServeRemoteBranchRateLimited(t *testing.T) {
reset := strconv.FormatInt(time.Now().Add(120*time.Second).Unix(), 10)
for _, tc := range []struct {
name string
status int
header http.Header
want int
retry [2]int
}{
{"primary 403", http.StatusForbidden, http.Header{"X-Ratelimit-Remaining": {"0"}, "X-Ratelimit-Reset": {reset}}, http.StatusServiceUnavailable, [2]int{100, 122}},
{"secondary 403", http.StatusForbidden, http.Header{"Retry-After": {"30"}}, http.StatusServiceUnavailable, [2]int{30, 31}},
{"429", http.StatusTooManyRequests, http.Header{"Retry-After": {"5"}}, http.StatusServiceUnavailable, [2]int{5, 6}},
{"429 no hint", http.StatusTooManyRequests, http.Header{}, http.StatusServiceUnavailable, [2]int{60, 60}},
{"plain 403", http.StatusForbidden, http.Header{}, http.StatusBadGateway, [2]int{}},
} {
t.Run(tc.name, func(t *testing.T) {
u := newUpstream(t, false)
u.status.Store(int32(tc.status))
u.header.Store(tc.header)
p := New()
p.githubAPI = u.srv.URL
w, _ := serve(p, models.Remote{Name: "gh", BaseURL: "https://github.com"}, "o/r/archive/refs/heads/main.tar.gz")
if w.Code != tc.want {
t.Fatalf("code = %d, want %d", w.Code, tc.want)
}
ra := w.Header().Get("Retry-After")
if tc.want != http.StatusServiceUnavailable {
if ra != "" {
t.Fatalf("Retry-After = %q on %d", ra, w.Code)
}
return
}
if n, err := strconv.Atoi(ra); err != nil || n < tc.retry[0] || n > tc.retry[1] {
t.Fatalf("Retry-After = %q, want %d..%d", ra, tc.retry[0], tc.retry[1])
}
})
}
}
func TestServeRemoteBranchNotFoundIsCached(t *testing.T) {
u := newUpstream(t, false)
u.status.Store(http.StatusNotFound)
p := New()
p.githubAPI = u.srv.URL
remote := models.Remote{Name: "gh", BaseURL: "https://github.com"}
path := "o/r/archive/refs/heads/nope.tar.gz"
for range 3 {
if w, _ := serve(p, remote, path); w.Code != http.StatusNotFound {
t.Fatalf("code = %d, want 404", w.Code)
}
}
if u.hits.Load() != 1 {
t.Fatalf("hits = %d, want 1 within negative ttl", u.hits.Load())
}
u.status.Store(http.StatusOK)
age(p, remote.Name, negativeTTL)
if w, _ := serve(p, remote, path); w.Code != http.StatusFound || u.hits.Load() != 2 {
t.Fatalf("after negative ttl: code=%d hits=%d", w.Code, u.hits.Load())
}
}
func TestServeRemoteConcurrentLookupsShareOneCall(t *testing.T) {
u := newUpstream(t, true)
u.gate = make(chan struct{})
p := New()
remote := models.Remote{Name: "g", BaseURL: u.srv.URL, MutableTTL: 600}
var wg sync.WaitGroup
codes := make([]int, 20)
for i := range codes {
wg.Add(1)
go func() {
defer wg.Done()
w, _ := serve(p, remote, "o/r/archive/refs/heads/main.tar.gz")
codes[i] = w.Code
}()
}
for u.hits.Load() == 0 {
time.Sleep(time.Millisecond)
}
time.Sleep(50 * time.Millisecond)
close(u.gate)
wg.Wait()
if u.hits.Load() != 1 {
t.Fatalf("hits = %d, want 1", u.hits.Load())
}
for i, c := range codes {
if c != http.StatusFound {
t.Fatalf("request %d: code %d", i, c)
}
}
}
func TestStoreBoundsEntriesPerRemote(t *testing.T) {
p := New()
for i := range maxRefs + 10 {
p.store("a", fmt.Sprint(i), "")
}
p.store("b", "x", shaA)
if n := len(p.refs["a"]); n != maxRefs {
t.Fatalf("remote a holds %d entries, want %d", n, maxRefs)
}
if len(p.refs["b"]) != 1 {
t.Fatalf("remote b must keep its own budget")
}
if _, ok := p.refs["a"][fmt.Sprint(maxRefs+9)]; !ok {
t.Fatalf("newest entry evicted")
}
}
func TestServeRemoteRejectsBadCommit(t *testing.T) {
u := newUpstream(t, true)
u.commit.Store("../../etc")
w, _ := serve(New(), models.Remote{Name: "g", BaseURL: u.srv.URL}, "o/r/archive/refs/heads/main.tar.gz")
if w.Code != http.StatusBadGateway {
t.Fatalf("code = %d, want 502", w.Code)
}
}
func TestServeRemoteAllowlist(t *testing.T) {
u := newUpstream(t, true)
remote := models.Remote{Name: "g", BaseURL: u.srv.URL, Patterns: []string{"^allowed/repo/"}}
if w, _ := serve(New(), remote, "other/repo/archive/refs/heads/main.tar.gz"); w.Code != http.StatusForbidden {
t.Fatalf("code = %d, want 403", w.Code)
}
if u.hits.Load() != 0 {
t.Fatalf("denied branch must not hit the API")
}
}
func TestServeRemoteBlockedRedirectTarget(t *testing.T) {
u := newUpstream(t, true)
remote := models.Remote{Name: "g", BaseURL: u.srv.URL, Blocklist: []string{"/archive/" + shaA}}
w, _ := serve(New(), remote, "o/r/archive/refs/heads/main.tar.gz")
if w.Code != http.StatusForbidden || w.Header().Get("Location") != "" {
t.Fatalf("code = %d location %q, want 403 without redirect", w.Code, w.Header().Get("Location"))
}
}
func TestServeRemotePassesThroughAndRejects(t *testing.T) {
p := New()
remote := models.Remote{Name: "g", BaseURL: "https://github.com"}
if _, handled := serve(p, remote, "o/r/archive/refs/tags/v1.tar.gz"); handled {
t.Error("tag archives must go through the proxy engine")
}
if _, handled := serve(p, remote, "o/r/releases/download/v1/a.bin"); handled {
t.Error("release assets must go through the proxy engine")
}
if w, handled := serve(p, remote, "o/r/raw/main/x"); !handled || w.Code != http.StatusNotFound {
t.Errorf("unsupported path: handled=%v code=%d", handled, w.Code)
}
}
+245
View File
@@ -0,0 +1,245 @@
package virtual
import (
"bytes"
"crypto/md5"
"encoding/hex"
"fmt"
"path"
"regexp"
"slices"
"strings"
"time"
)
// debIndexRe matches the binary Packages indexes a virtual merges: the flat
// repo's root Packages and a suite's <component>/binary-<arch>/Packages.
var debIndexRe = regexp.MustCompile(`^([^/]+/binary-[^/]+/)?Packages(\.[a-z0-9]+)?$`)
// debByHashRe matches an index requested by its SHA256 (Acquire-By-Hash).
var debByHashRe = regexp.MustCompile(`^([^/]+/binary-[^/]+/)?by-hash/SHA256/[0-9a-f]{64}$`)
// debCompressions is the order a member's index variants are fetched in.
var debCompressions = []string{".xz", ".gz", ".zst", ".bz2", ""}
// DebMember is one member's Release plus its decompressed Packages indexes,
// keyed by index directory relative to the Release ("" for a flat repo).
type DebMember struct {
RemoteName string
Release []byte
Packages map[string][]byte
}
// DebRepo is a merged apt repo: Release plus the indexes it lists, keyed by
// their path relative to the Release.
type DebRepo struct {
Release []byte
Files map[string][]byte
}
type debRelease struct {
fields map[string]string
files []string
}
// parseDebRelease reads a Release file's single-line fields and the names
// listed in its checksum sections.
func parseDebRelease(body []byte) debRelease {
rel := debRelease{fields: map[string]string{}}
inSums := false
seen := map[string]bool{}
for line := range strings.SplitSeq(string(body), "\n") {
line = strings.TrimRight(line, "\r")
if line == "" {
continue
}
if line[0] == ' ' || line[0] == '\t' {
if f := strings.Fields(line); inSums && len(f) == 3 && !seen[f[2]] {
seen[f[2]] = true
rel.files = append(rel.files, f[2])
}
continue
}
k, v, _ := strings.Cut(line, ":")
v = strings.TrimSpace(v)
inSums = v == "" && (k == "MD5Sum" || strings.HasPrefix(k, "SHA"))
if _, ok := rel.fields[k]; !ok {
rel.fields[k] = v
}
}
return rel
}
// debIndexFiles returns, per index directory, the variant of its Packages file
// to fetch from a member's Release listing.
func debIndexFiles(rel debRelease) map[string]string {
listed := map[string]bool{}
for _, f := range rel.files {
listed[f] = true
}
out := map[string]string{}
for _, f := range rel.files {
m := debIndexRe.FindStringSubmatch(f)
if m == nil || out[m[1]] != "" {
continue
}
for _, c := range debCompressions {
if listed[m[1]+"Packages"+c] {
out[m[1]] = m[1] + "Packages" + c
break
}
}
}
return out
}
func splitStanzas(body []byte) []string {
var out []string
for s := range strings.SplitSeq(strings.ReplaceAll(string(body), "\r\n", "\n"), "\n\n") {
if s = strings.Trim(s, "\n"); s != "" {
out = append(out, s)
}
}
return out
}
func stanzaField(stanza, key string) string {
for line := range strings.SplitSeq(stanza, "\n") {
if k, v, ok := strings.Cut(line, ":"); ok && k == key {
return strings.TrimSpace(v)
}
}
return ""
}
func unionFields(dst []string, field string) []string {
for f := range strings.FieldsSeq(field) {
if !slices.Contains(dst, f) {
dst = append(dst, f)
}
}
return dst
}
func parseDebDate(s string) (time.Time, bool) {
for _, layout := range []string{time.RFC1123Z, time.RFC1123} {
if t, err := time.Parse(layout, s); err == nil {
return t, true
}
}
return time.Time{}, false
}
// MergeDeb merges members into one apt repo. Members are in priority order: the
// first member to carry a (Package, Version, Architecture) in an index wins;
// stanzas missing any of those are always kept. Filename is prefixed with the
// owning member's name so the virtual can route downloads back to it.
func MergeDeb(members []DebMember) *DebRepo {
var suite, codename string
var arches, comps, dirs []string
date := time.Unix(0, 0)
for _, m := range members {
rel := parseDebRelease(m.Release)
if suite == "" {
suite, codename = rel.fields["Suite"], rel.fields["Codename"]
}
arches = unionFields(arches, rel.fields["Architectures"])
comps = unionFields(comps, rel.fields["Components"])
if t, ok := parseDebDate(rel.fields["Date"]); ok && t.After(date) {
date = t
}
for d := range m.Packages {
if !slices.Contains(dirs, d) {
dirs = append(dirs, d)
}
}
}
slices.Sort(dirs)
repo := &DebRepo{Files: map[string][]byte{}}
for _, d := range dirs {
seen := map[string]bool{}
var b bytes.Buffer
for _, m := range members {
for _, s := range splitStanzas(m.Packages[d]) {
pkg, ver, arch := stanzaField(s, "Package"), stanzaField(s, "Version"), stanzaField(s, "Architecture")
if pkg != "" && ver != "" && arch != "" {
key := pkg + "\x00" + ver + "\x00" + arch
if seen[key] {
continue
}
seen[key] = true
}
if fn := stanzaField(s, "Filename"); fn != "" {
s = rewriteFilename(s, m.RemoteName+"/"+strings.TrimPrefix(path.Clean("/"+fn), "/"))
}
b.WriteString(s)
b.WriteString("\n\n")
}
}
gz := gzipDeterministic(b.Bytes())
repo.Files[d+"Packages"] = b.Bytes()
repo.Files[d+"Packages.gz"] = gz
repo.Files[d+"by-hash/SHA256/"+sha256Hex(b.Bytes())] = b.Bytes()
repo.Files[d+"by-hash/SHA256/"+sha256Hex(gz)] = gz
}
var r bytes.Buffer
if suite != "" {
fmt.Fprintf(&r, "Suite: %s\n", suite)
}
if codename != "" {
fmt.Fprintf(&r, "Codename: %s\n", codename)
}
fmt.Fprintf(&r, "Date: %s\n", date.UTC().Format(time.RFC1123Z))
fmt.Fprintf(&r, "Architectures: %s\n", strings.Join(arches, " "))
if len(comps) > 0 {
fmt.Fprintf(&r, "Components: %s\n", strings.Join(comps, " "))
}
r.WriteString("Acquire-By-Hash: yes\n")
for _, sec := range []struct {
name string
sum func([]byte) string
}{{"MD5Sum", md5Hex}, {"SHA256", sha256Hex}} {
r.WriteString(sec.name + ":\n")
for _, d := range dirs {
for _, n := range []string{d + "Packages", d + "Packages.gz"} {
fmt.Fprintf(&r, " %s %d %s\n", sec.sum(repo.Files[n]), len(repo.Files[n]), n)
}
}
}
repo.Release = r.Bytes()
return repo
}
func rewriteFilename(stanza, fn string) string {
lines := strings.Split(stanza, "\n")
for i, l := range lines {
if strings.HasPrefix(l, "Filename:") {
lines[i] = "Filename: " + fn
break
}
}
return strings.Join(lines, "\n")
}
// debIndexPath splits an index request into its Release directory and the file
// under it. ok is false for anything a deb virtual does not merge.
func debIndexPath(p string) (base, name string, ok bool) {
if rest, found := strings.CutPrefix(p, "dists/"); found {
suite, n, found := strings.Cut(rest, "/")
if !found || suite == "" {
return "", "", false
}
base, p = "dists/"+suite+"/", n
}
if p != "Release" && !debIndexRe.MatchString(p) && !debByHashRe.MatchString(p) {
return "", "", false
}
return base, p, true
}
func md5Hex(data []byte) string {
h := md5.Sum(data)
return hex.EncodeToString(h[:])
}
+179
View File
@@ -0,0 +1,179 @@
package virtual
import (
"bytes"
"crypto/md5"
"encoding/hex"
"fmt"
"maps"
"path"
"slices"
"strings"
"testing"
)
func debStanza(pkg, ver, arch, filename string) string {
return fmt.Sprintf("Package: %s\nVersion: %s\nArchitecture: %s\nDescription: test\n multi-line body\nFilename: %s\nSize: 1\n", pkg, ver, arch, filename)
}
func flatDebMember(name, date string, stanzas ...string) DebMember {
rel := "Date: " + date + "\nArchitectures: amd64\nAcquire-By-Hash: no\nSHA256:\n x 1 Packages\n x 1 Packages.gz\n"
return DebMember{RemoteName: name, Release: []byte(rel), Packages: map[string][]byte{"": []byte(strings.Join(stanzas, "\n"))}}
}
func TestMergeDebDedupesByMemberOrder(t *testing.T) {
repo := MergeDeb([]DebMember{
flatDebMember("a", "Mon, 05 Oct 2026 10:00:00 +0000",
debStanza("foo", "1.0", "amd64", "pool/foo_1.0_amd64.deb"),
debStanza("bar", "2.0", "amd64", "/pool/bar.deb"),
debStanza("dot", "1.0", "amd64", "./dot.deb"),
"Package: noversion\nFilename: nv1.deb\n",
"Package: noversion\nFilename: nv2.deb\n"),
flatDebMember("b", "Tue, 06 Oct 2026 10:00:00 UTC",
debStanza("foo", "1.0", "amd64", "pool/foo-from-b.deb"),
debStanza("foo", "1.0", "arm64", "pool/foo_arm64.deb"),
debStanza("foo", "1.1", "amd64", "o/r/releases/download/v1/foo.deb")),
})
pk := string(repo.Files["Packages"])
stanzas := splitStanzas(repo.Files["Packages"])
if len(stanzas) != 7 {
t.Fatalf("want 7 stanzas (one duplicate dropped), got %d:\n%s", len(stanzas), pk)
}
for _, want := range []string{
"Filename: a/pool/foo_1.0_amd64.deb",
"Filename: a/pool/bar.deb",
"Filename: a/dot.deb",
"Filename: a/nv1.deb",
"Filename: a/nv2.deb",
"Filename: b/pool/foo_arm64.deb",
"Filename: b/o/r/releases/download/v1/foo.deb",
" multi-line body",
} {
if !strings.Contains(pk, want) {
t.Errorf("merged Packages missing %q:\n%s", want, pk)
}
}
if strings.Contains(pk, "foo-from-b") {
t.Fatalf("second member's duplicate must lose:\n%s", pk)
}
if !bytes.Equal(gunzip(t, repo.Files["Packages.gz"]), repo.Files["Packages"]) {
t.Fatal("Packages.gz does not match Packages")
}
if !strings.Contains(string(repo.Release), "Date: Tue, 06 Oct 2026 10:00:00 +0000\n") {
t.Fatalf("Date should be the newest member date:\n%s", repo.Release)
}
}
func TestMergeDebReleaseHashesFiles(t *testing.T) {
repo := MergeDeb([]DebMember{flatDebMember("a", "", debStanza("foo", "1", "amd64", "pool/foo.deb"))})
rel := parseDebRelease(repo.Release)
if got := slices.Sorted(slices.Values(rel.files)); !slices.Equal(got, []string{"Packages", "Packages.gz"}) {
t.Fatalf("Release lists %v", got)
}
inSHA := false
checked := 0
for line := range strings.SplitSeq(string(repo.Release), "\n") {
if !strings.HasPrefix(line, " ") {
inSHA = line == "SHA256:"
continue
}
var sum, name string
var size int
if _, err := fmt.Sscan(line, &sum, &size, &name); err != nil {
t.Fatal(err)
}
body := repo.Files[name]
want := sha256Hex(body)
if !inSHA {
h := md5.Sum(body)
want = hex.EncodeToString(h[:])
}
if sum != want || size != len(body) {
t.Errorf("%s: Release has %s %d, file is %s %d", name, sum, size, want, len(body))
}
checked++
}
if checked != 4 {
t.Fatalf("checked %d Release entries, want 4:\n%s", checked, repo.Release)
}
if !bytes.Equal(MergeDeb([]DebMember{flatDebMember("a", "", debStanza("foo", "1", "amd64", "pool/foo.deb"))}).Release, repo.Release) {
t.Fatal("merge must be deterministic")
}
}
func TestMergeDebSuiteLayout(t *testing.T) {
rel := func(comps string) []byte {
return []byte("Suite: stable\nCodename: trixie\nArchitectures: amd64 arm64\nComponents: " + comps + "\nSHA256:\n x 1 main/binary-amd64/Packages.xz\n")
}
repo := MergeDeb([]DebMember{
{RemoteName: "deb1", Release: rel("main"), Packages: map[string][]byte{"main/binary-amd64/": []byte(debStanza("foo", "1", "amd64", "pool/main/f/foo.deb"))}},
{RemoteName: "deb2", Release: rel("main contrib"), Packages: map[string][]byte{
"main/binary-amd64/": []byte(debStanza("bar", "1", "amd64", "pool/main/b/bar.deb")),
"contrib/binary-amd64/": []byte(debStanza("baz", "1", "amd64", "pool/contrib/b/baz.deb")),
}},
})
for _, want := range []string{"Suite: stable\n", "Codename: trixie\n", "Components: main contrib\n", "Architectures: amd64 arm64\n"} {
if !strings.Contains(string(repo.Release), want) {
t.Errorf("Release missing %q:\n%s", want, repo.Release)
}
}
want := []string{"contrib/binary-amd64/Packages", "contrib/binary-amd64/Packages.gz", "main/binary-amd64/Packages", "main/binary-amd64/Packages.gz"}
if got := slices.Sorted(slices.Values(parseDebRelease(repo.Release).files)); !slices.Equal(got, want) {
t.Fatalf("Release lists %v, want %v", got, want)
}
if !strings.Contains(string(repo.Release), "Acquire-By-Hash: yes\n") {
t.Fatalf("Release must advertise by-hash:\n%s", repo.Release)
}
for _, n := range want {
byHash := path.Dir(n) + "/by-hash/SHA256/" + sha256Hex(repo.Files[n])
if !bytes.Equal(repo.Files[byHash], repo.Files[n]) {
t.Errorf("%s not served at %s", n, byHash)
}
}
if len(repo.Files) != 2*len(want) {
t.Fatalf("files %v, want each index plus its by-hash copy", slices.Sorted(maps.Keys(repo.Files)))
}
if n := len(splitStanzas(repo.Files["main/binary-amd64/Packages"])); n != 2 {
t.Fatalf("main/binary-amd64 should merge both members, got %d stanzas", n)
}
}
func TestDebIndexFiles(t *testing.T) {
rel := parseDebRelease([]byte("Origin: Debian\nMD5Sum:\n a 1 main/binary-amd64/Packages\n a 1 main/binary-amd64/Packages.gz\nSHA256:\n" +
" a 1 main/binary-amd64/Packages\n a 1 main/binary-amd64/Packages.xz\n a 1 main/binary-amd64/Packages.gz\n" +
" a 1 main/debian-installer/binary-amd64/Packages.xz\n a 1 main/i18n/Translation-en.xz\n a 1 contrib/binary-all/Packages\n"))
got := debIndexFiles(rel)
want := map[string]string{"main/binary-amd64/": "main/binary-amd64/Packages.xz", "contrib/binary-all/": "contrib/binary-all/Packages"}
if !maps.Equal(got, want) {
t.Fatalf("got %v, want %v", got, want)
}
if flat := debIndexFiles(parseDebRelease([]byte("SHA256:\n a 1 Packages\n a 1 Packages.gz\n"))); flat[""] != "Packages.gz" {
t.Fatalf("flat repo should fetch Packages.gz, got %v", flat)
}
}
func TestDebIndexPath(t *testing.T) {
for _, tc := range []struct {
in, base, name string
ok bool
}{
{"Release", "", "Release", true},
{"Packages.gz", "", "Packages.gz", true},
{"dists/trixie/Release", "dists/trixie/", "Release", true},
{"dists/trixie/main/binary-amd64/Packages", "dists/trixie/", "main/binary-amd64/Packages", true},
{"by-hash/SHA256/abababababababababababababababababababababababababababababababab", "", "by-hash/SHA256/abababababababababababababababababababababababababababababababab", true},
{"dists/trixie/main/binary-amd64/by-hash/SHA256/abababababababababababababababababababababababababababababababab", "dists/trixie/", "main/binary-amd64/by-hash/SHA256/abababababababababababababababababababababababababababababababab", true},
{"dists/trixie/main/binary-amd64/by-hash/MD5Sum/abababababababababababababababab", "", "", false},
{"InRelease", "", "", false},
{"Release.gpg", "", "", false},
{"dists/trixie/InRelease", "", "", false},
{"dists/trixie/main/i18n/Translation-en", "", "", false},
{"dists/Release", "", "", false},
{"pool/foo.deb", "", "", false},
} {
base, name, ok := debIndexPath(tc.in)
if base != tc.base || name != tc.name || ok != tc.ok {
t.Errorf("%s: got (%q, %q, %v), want (%q, %q, %v)", tc.in, base, name, ok, tc.base, tc.name, tc.ok)
}
}
}
+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
}
}
+161
View File
@@ -5,6 +5,7 @@ import (
"errors"
"fmt"
"regexp"
"strings"
"sync"
"sync/atomic"
"testing"
@@ -157,6 +158,97 @@ func TestMemberRedirect(t *testing.T) {
}
}
func debEngine(members map[string]*DebMember) *Engine {
e := fakeEngine(nil)
e.debMember = func(_ context.Context, name, base string) (*DebMember, error) {
if m := members[name]; m != nil && base == "" {
return m, nil
}
return nil, errors.New("upstream down")
}
return e
}
func debVirt(members ...string) models.Virtual {
return models.Virtual{Name: "v", PackageType: models.PackageDeb, Members: members}
}
func TestFetchDebFailsClosed(t *testing.T) {
a := flatDebMember("a", "", debStanza("foo", "1", "amd64", "pool/foo.deb"))
e := debEngine(map[string]*DebMember{"a": &a})
for _, p := range []string{"Release", "./Packages.gz", "dists/trixie/Release"} {
if _, _, err := e.Fetch(context.Background(), debVirt("a", "b"), p, ""); err == nil || errors.Is(err, ErrNotFound) {
t.Errorf("%s: want upstream error with a member down, got %v", p, err)
}
}
for _, p := range []string{"InRelease", "./Release.gpg", "Packages.xz"} {
if _, _, err := e.Fetch(context.Background(), debVirt("a"), p, ""); !errors.Is(err, ErrNotFound) {
t.Errorf("%s: want not found, got %v", p, err)
}
}
}
func TestFetchDebRemergesAfterTTL(t *testing.T) {
a := flatDebMember("a", "", debStanza("foo", "1", "amd64", "pool/foo_1.deb"))
e := debEngine(map[string]*DebMember{"a": &a})
virt := debVirt("a")
rel, _, err := e.Fetch(context.Background(), virt, "./Release", "")
if err != nil {
t.Fatal(err)
}
a.Packages[""] = []byte(debStanza("foo", "2", "amd64", "pool/foo_2.deb"))
e.mergeTTL = 0
newRel, _, err := e.Fetch(context.Background(), virt, "Release", "")
if err != nil || string(newRel) == string(rel) {
t.Fatalf("Release should reflect the member change (err %v)", err)
}
pk, ct, err := e.Fetch(context.Background(), virt, "./Packages", "")
if err != nil || ct != "text/plain" || !strings.Contains(string(pk), "Version: 2") {
t.Fatalf("Packages should be the current merge (ct %q, err %v):\n%s", ct, err, pk)
}
}
func TestFetchDebMergeIsSharedPerSuite(t *testing.T) {
var calls atomic.Int32
e := fakeEngine(nil)
e.mergeTTL = time.Minute
e.debMember = func(_ context.Context, name, base string) (*DebMember, error) {
calls.Add(1)
m := flatDebMember(name, "", debStanza("foo", "1", "amd64", "pool/foo.deb"))
return &m, nil
}
for _, p := range []string{"Release", "Packages", "Packages.gz", "./Release", "dists/a/Release", "dists/b/Release"} {
if _, _, err := e.Fetch(context.Background(), debVirt("a"), p, ""); err != nil && !errors.Is(err, ErrNotFound) {
t.Fatal(err)
}
}
if n := calls.Load(); n != 3 {
t.Fatalf("member fetched %d times, want once per Release dir (3)", n)
}
}
func TestMemberRedirectDeb(t *testing.T) {
e := fakeEngine(nil)
virt := debVirt("gh", "loc")
for _, tc := range []struct {
path, want string
ok bool
err error
}{
{path: "gh/o/r/releases/download/v1/foo_1.0_amd64.deb", ok: true, want: "http://h/api/v1/remote/gh/o/r/releases/download/v1/foo_1.0_amd64.deb"},
{path: "./gh/pool/a+b.deb", ok: true, want: "http://h/api/v1/remote/gh/pool/a+b.deb"},
{path: "gh/../loc/x.deb", err: ErrBadPath},
{path: "dists/trixie/Release"},
{path: "Packages"},
} {
got, ok, err := e.MemberRedirect(context.Background(), virt, tc.path, "", "http://h/")
if got != tc.want || ok != tc.ok || !errors.Is(err, tc.err) {
t.Errorf("%s: got (%q, %v, %v), want (%q, %v, %v)", tc.path, got, ok, err, tc.want, tc.ok, tc.err)
}
}
}
type fakeRepo struct {
idx string
fs map[string][]byte
@@ -206,3 +298,72 @@ func TestMergedReusesWithinTTLAndIsolatesKeys(t *testing.T) {
t.Error("unknown key served a file")
}
}
func suiteDebMember(name, version string) *DebMember {
return &DebMember{RemoteName: name, Release: []byte("Suite: s\nSHA256:\n x 1 main/binary-amd64/Packages.gz\n"),
Packages: map[string][]byte{"main/binary-amd64/": []byte(debStanza("foo", version, "amd64", "pool/foo_"+version+".deb"))}}
}
func TestFetchDebByHashSurvivesRemerge(t *testing.T) {
version := "1"
e := fakeEngine(nil)
e.debMember = func(_ context.Context, name, _ string) (*DebMember, error) { return suiteDebMember(name, version), nil }
virt := debVirt("a")
fetch := func(p string) []byte {
body, _, err := e.Fetch(context.Background(), virt, p, "")
if err != nil {
t.Fatalf("%s: %v", p, err)
}
return body
}
byHash := func(rel []byte) string {
m := regexp.MustCompile(`(?m)^SHA256:\n(?: .*\n)*? ([0-9a-f]{64}) \d+ main/binary-amd64/Packages.gz$`).FindSubmatch(rel)
if m == nil {
t.Fatalf("no Packages.gz SHA256 in Release:\n%s", rel)
}
return "dists/trixie/main/binary-amd64/by-hash/SHA256/" + string(m[1])
}
oldRel := fetch("dists/trixie/Release")
version, e.mergeTTL = "2", 0
if newRel := fetch("dists/trixie/Release"); string(newRel) == string(oldRel) {
t.Fatal("Release should reflect the member change")
}
if got := string(gunzip(t, fetch(byHash(oldRel)))); !strings.Contains(got, "Version: 1\n") {
t.Fatalf("by-hash from the previous Release served the wrong index:\n%s", got)
}
if got := string(fetch("dists/trixie/main/binary-amd64/Packages")); !strings.Contains(got, "Version: 2\n") {
t.Fatalf("fixed-name Packages should be the current merge:\n%s", got)
}
}
func TestFetchDebBoundsSuitesPerVirtual(t *testing.T) {
e := fakeEngine(nil)
e.mergeTTL = time.Minute
e.debMember = func(_ context.Context, name, _ string) (*DebMember, error) { return suiteDebMember(name, "1"), nil }
other := models.Virtual{Name: "other", PackageType: models.PackageDeb, Members: []string{"a"}}
if _, _, err := e.Fetch(context.Background(), other, "dists/keep/Release", ""); err != nil {
t.Fatal(err)
}
for i := range maxDebSuites + 5 {
if _, _, err := e.Fetch(context.Background(), debVirt("a"), fmt.Sprintf("dists/s%d/Release", i), ""); err != nil {
t.Fatal(err)
}
time.Sleep(time.Microsecond)
}
n := 0
for k := range e.gens {
if strings.HasPrefix(k, "deb\x00v\x00") {
n++
}
}
if n != maxDebSuites {
t.Fatalf("virtual v caches %d suites, want %d", n, maxDebSuites)
}
if e.gens["deb\x00v\x00dists/s0/"] != nil || e.gens["deb\x00v\x00dists/s4/"] != nil {
t.Fatal("least recently used suites should be evicted first")
}
if e.gens["deb\x00other\x00dists/keep/"] == nil {
t.Fatal("another virtual's suites must not be evicted")
}
}
+5
View File
@@ -0,0 +1,5 @@
package models
const PackageVCS PackageType = "vcs"
func init() { registerPackageType(PackageVCS) }
+26
View File
@@ -0,0 +1,26 @@
import type { PackageTypeUI } from '../types';
const def: PackageTypeUI = {
color: 'default',
snippets: ({ proxy }) => {
return [
{
title: 'Download a tag archive',
language: 'bash',
code: `curl -fsSLO ${proxy}/<owner>/<repo>/archive/refs/tags/<tag>.tar.gz`,
},
{
title: 'Download a branch archive (redirects to its current commit)',
language: 'bash',
code: `curl -fsSL -o src.tar.gz ${proxy}/<owner>/<repo>/archive/refs/heads/<branch>.tar.gz`,
},
{
title: 'Download a release asset',
language: 'bash',
code: `curl -fsSLO ${proxy}/<owner>/<repo>/releases/download/<tag>/<asset>`,
},
];
},
};
export default def;