fix: additional comments

Signed-off-by: Umberto Sgueglia <usgueglia@contractor.linuxfoundation.org>

Author: ulemons

services/apps/packages_worker/src/maven/__tests__/normalize.test.ts

@@ -119,6 +119,22 @@ describe('parseRepoUrl', () => {
     const result = parseRepoUrl('https://github.com/')
     expect(result).toEqual({ host: 'github', owner: null, name: null })
   })
+
+  // The canonical OpenDaylight link keeps repo identity in ?p=, not the (fixed)
+  // pathname — a generic parts[0]/parts[1] split would collapse every OpenDaylight
+  // repo to the same owner='gerrit', name='gitweb'.
+  it('derives owner=null and name from the ?p= query for OpenDaylight Gerrit links', () => {
+    expect(parseRepoUrl('https://git.opendaylight.org/gerrit/gitweb?p=netconf.git')).toEqual({
+      host: 'gerrit',
+      owner: null,
+      name: 'netconf',
+    })
+    expect(parseRepoUrl('https://git.opendaylight.org/gerrit/gitweb?p=aaa.git')).toEqual({
+      host: 'gerrit',
+      owner: null,
+      name: 'aaa',
+    })
+  })
 })
 
 describe('normalizeScmUrl', () => {
@@ -206,6 +222,17 @@ describe('normalizeScmUrl', () => {
     expect(normalizeScmUrl('https://svn.apache.org/repos/asf')).toBeNull()
   })
 
+  // java.net requires an explicit SVN marker because the bare host is shared with
+  // non-SVN content (forums, wikis, JIRA) — this must not accidentally widen to
+  // treat every java.net URL as SVN.
+  it('treats java.net as SVN only with an explicit scm:svn:/svn:// marker', () => {
+    expect(normalizeScmUrl('https://java.net/projects/foo/sources/svn')).toBeNull()
+    expect(normalizeScmUrl('scm:svn:https://java.net/svn/foo/trunk')).toBe(
+      'https://java.net/svn/foo',
+    )
+    expect(normalizeScmUrl('svn://java.net/svn/foo/trunk')).toBe('https://java.net/svn/foo')
+  })
+
   it('does not map dead services that incidentally match an "svn" text filter', () => {
     expect(normalizeScmUrl('http://specs.googlecode.com/svn/trunk')).toBeNull()
   })
@@ -408,6 +435,21 @@ describe('normalizeScmUrl', () => {
     expect(normalizeScmUrl('https://android.googlesource.com/')).toBeNull()
   })
 
+  // cs.android.com is the Android Code Search UI, backed by the same Gitiles
+  // repos as android.googlesource.com — same identity passthrough applies.
+  it('passes through cs.android.com urls unchanged, mirroring the googlesource identity form', () => {
+    expect(normalizeScmUrl('https://cs.android.com/androidx/platform/frameworks/support')).toBe(
+      'https://cs.android.com/androidx/platform/frameworks/support',
+    )
+  })
+
+  it('returns null for a bare OpenDaylight Gerrit root with no repo path', () => {
+    // /gerrit/ alone (as opposed to /gerrit/gitweb?p=...) has nothing to extract —
+    // same "no derivable identity" case as the bare gerrit.onosproject.org root
+    // tested above, confirmed against real declared_repository_url values.
+    expect(normalizeScmUrl('https://git.opendaylight.org/gerrit/')).toBeNull()
+  })
+
   it('recovers GitHub Pages project-page urls (owner.github.io/repo)', () => {
     expect(normalizeScmUrl('https://silentbalanceyh.github.io/vertx-zero/')).toBe(
       'https://github.com/silentbalanceyh/vertx-zero',
@@ -594,9 +636,7 @@ describe('normalizeScmUrl', () => {
 
   it('returns null for non-SCM OpenDaylight URLs instead of mistaking them for a repo', () => {
     // A `p` query on some unrelated path is not a gitweb browse link.
-    expect(
-      normalizeScmUrl('https://git.opendaylight.org/somewhere?p=netconf.git'),
-    ).toBeNull()
+    expect(normalizeScmUrl('https://git.opendaylight.org/somewhere?p=netconf.git')).toBeNull()
     // Single path segment with no .git suffix — e.g. a static asset, not a clone URL.
     expect(normalizeScmUrl('https://git.opendaylight.org/favicon.ico')).toBeNull()
   })

services/apps/packages_worker/src/maven/extract.ts

@@ -487,13 +487,18 @@ export async function extractArtifact(groupId: string, artifactId: string, versi
  * declared_repository_url that need more than a flat owner/repo allowlist are handled
  * by their own dedicated normalizers below rather than here:
  *   - git.eclipse.org: cgit `/c/<owner>/<repo>` paths → normalizeEclipseCgitUrl.
- *   - android.googlesource.com: Gitiles identity paths → normalizeGooglesourceUrl.
+ *   - android.googlesource.com / cs.android.com: Gitiles identity paths (the Android
+ *     Code Search UI mirrors the same repo naming as the Gitiles source itself) →
+ *     normalizeGooglesourceUrl.
  *   - ec.europa.eu: Bitbucket-Server `/projects/x/repos/y` → normalizeBitbucketServerUrl.
  *   - gitbox.apache.org / git.apache.org / git-wip-us.apache.org: gitweb
  *     `/repos/asf/<repo>` or `?p=<repo>` → normalizeApacheGitwebUrl.
  *
  * Still NOT handled (need more than a host allowlist, or are unreachable):
  *   - eclipse.gerrithub.io: Gerrit admin-UI paths (`/admin/repos/...`), not a clone shape.
+ *   - gerrit.onosproject.org: every occurrence observed is the bare Gerrit root
+ *     (`http://gerrit.onosproject.org/`) with no path at all — nothing to extract,
+ *     unlike git.opendaylight.org which has resolvable gitweb/SSH forms.
  *
  * Internal-only hosts (git.corp.adobe.com, fit.corp.adobe.com, gitlab.alibaba-inc.com)
  * are included below even though they're unreachable for external consumers — per
@@ -764,8 +769,12 @@ function normalizeAliyunCodeupUrl(host: string, pathname: string): string | null
  * Android's Gerrit/Gitiles mirror already uses its full path as the canonical,
  * resolvable repo name (e.g. /platform/tools/base) — nested repo naming is normal
  * here, so this is an identity passthrough rather than an owner/repo split.
+ *
+ * cs.android.com (Android Code Search) is included too: it's a browsing UI over
+ * the same Gitiles-backed repos and mirrors the identical path naming (e.g.
+ * /androidx/platform/frameworks/support) — same passthrough applies.
  */
-const GOOGLESOURCE_IDENTITY_HOSTS = new Set(['android.googlesource.com'])
+const GOOGLESOURCE_IDENTITY_HOSTS = new Set(['android.googlesource.com', 'cs.android.com'])
 
 function normalizeGooglesourceUrl(host: string, pathname: string): string | null {
   const path = pathname.replace(/\/+$/, '').replace(/\.git$/, '')
@@ -782,12 +791,14 @@ function normalizeGooglesourceUrl(host: string, pathname: string): string | null
  * Both are unified to the gitweb browse form so the stored link is always
  * resolvable in a browser, not just via `git clone`.
  */
-const OPENDALIGHT_GERRIT_HOST = 'git.opendaylight.org'
+export const OPENDALIGHT_GERRIT_HOST = 'git.opendaylight.org'
+export const OPENDALIGHT_GERRIT_GITWEB_PATH = '/gerrit/gitweb'
 
 function normalizeOpendaylightGerritUrl(pathname: string, search: string): string | null {
   // Query extraction only applies to the actual gitweb browse script — a `p` query
   // on some unrelated path isn't a repo reference.
-  const queryRepo = pathname === '/gerrit/gitweb' ? new URLSearchParams(search).get('p') : null
+  const queryRepo =
+    pathname === OPENDALIGHT_GERRIT_GITWEB_PATH ? new URLSearchParams(search).get('p') : null
   const segments = pathname.split('/').filter(Boolean)
   // The one-segment SSH-clone form must end in .git, otherwise any arbitrary
   // single-segment path (e.g. /favicon.ico) would be mistaken for a repo.

services/apps/packages_worker/src/maven/normalize.ts

@@ -1,4 +1,4 @@
-import { isSvnHost } from './extract'
+import { OPENDALIGHT_GERRIT_GITWEB_PATH, OPENDALIGHT_GERRIT_HOST, isSvnHost } from './extract'
 
 export function isPrerelease(version: string): boolean {
   return /-(SNAPSHOT|alpha|beta|rc|cr|m\d+|dev)/i.test(version)
@@ -18,6 +18,16 @@ export function parseRepoUrl(
     // name='asf'). Keep the whole path as `name` instead.
     if (isSvnHost(h)) return { host: 'svn', owner: null, name: parts.join('/') || null }
 
+    // OpenDaylight's canonical link keeps the actual repo identity only in the
+    // ?p= query — the pathname is always the fixed /gerrit/gitweb script, so a
+    // generic parts[0]/parts[1] split would store owner='gerrit', name='gitweb'
+    // for every single OpenDaylight repo, indistinguishable from one another.
+    if (h === OPENDALIGHT_GERRIT_HOST && parsed.pathname === OPENDALIGHT_GERRIT_GITWEB_PATH) {
+      const p = parsed.searchParams.get('p')
+      const name = p ? p.replace(/\.git$/, '') : null
+      return { host: 'gerrit', owner: null, name }
+    }
+
     let host: string
     if (h === 'github.com' || h.endsWith('.github.com')) host = 'github'
     else if (h === 'gitlab.com' || h.includes('gitlab')) host = 'gitlab'