fix: keep originCandidate on raw origin remote for fork detection

resolvePrWorkItemSource resolved the raw origin candidate via getOwnerRepo,
which is now upstream-first. For fork repos that collapsed originCandidate
onto upstreamCandidate, hiding the issue-source selector in the renderer.
Resolve it via getOriginOwnerRepo so the raw origin remote is preserved, and
refresh two comments that still claimed PR lookups target origin.

Co-Authored-By: Claude <noreply@anthropic.com>

Co-authored-by: Orca <help@stably.ai>

src/main/github/client-issue-source.test.ts

@@ -36,6 +36,10 @@ vi.mock('./gh-utils', async () => {
     execFileAsync: execFileAsyncMock,
     ghExecFileAsync: ghExecFileAsyncMock,
     getOwnerRepo: getOwnerRepoMock,
+    // Why: resolvePrWorkItemSource resolves the raw origin candidate via
+    // getOriginOwnerRepo; map it to the same mock these tests use for the
+    // origin PR repo.
+    getOriginOwnerRepo: getOwnerRepoMock,
     getIssueOwnerRepo: getIssueOwnerRepoMock,
     getOwnerRepoForRemote: getOwnerRepoForRemoteMock,
     resolveIssueSource: resolveIssueSourceMock,

src/main/github/client-work-items.test.ts

@@ -50,6 +50,9 @@ vi.mock('./gh-utils', () => ({
       ? {}
       : { cwd: context.repoPath, ...(context.wslDistro ? { wslDistro: context.wslDistro } : {}) },
   getOwnerRepo: getOwnerRepoMock,
+  // Why: resolvePrWorkItemSource resolves the raw origin candidate via
+  // getOriginOwnerRepo; map it to the same mock these tests use for origin.
+  getOriginOwnerRepo: getOwnerRepoMock,
   getIssueOwnerRepo: getIssueOwnerRepoMock,
   getOwnerRepoForRemote: getOwnerRepoForRemoteMock,
   resolveIssueSource: resolveIssueSourceMock,

src/main/github/client.ts

@@ -993,8 +993,12 @@ async function resolvePrWorkItemSource(
   connectionId?: string | null,
   localGitOptions: LocalGitExecOptions = {}
 ): Promise<ResolvedPrWorkItemSource> {
+  // Why: originCandidate must stay the raw `origin` remote (not upstream-first),
+  // so the renderer can compare it against upstream to decide whether to show the
+  // issue-source selector. getOwnerRepo is now upstream-first, so use the
+  // origin-specific resolver here.
   const [originCandidate, upstreamCandidate] = await Promise.all([
-    getOwnerRepo(repoPath, connectionId, localGitOptions),
+    getOriginOwnerRepo(repoPath, connectionId, localGitOptions),
     getOwnerRepoForRemote(repoPath, 'upstream', connectionId, localGitOptions)
   ])
   const source =
@@ -1898,12 +1902,13 @@ export async function getWorkItem(
         return issue
       }
     } catch (err) {
-      // Why: the issue lookup now targets `upstream` while the PR lookup targets `origin`,
-      // so a transient upstream failure (5xx, rate limit, network flake) on issue #N would
-      // silently fall through to origin's PR #N — potentially a completely unrelated item.
-      // Only fall through when the issue genuinely doesn't exist (404); re-throw everything
-      // else so the outer catch returns null and the caller sees a real failure instead of
-      // a wrong item. classifyGhError centralizes the 404/"not found" pattern-matching.
+      // Why: issue and PR lookups both resolve upstream-first, so a transient
+      // upstream failure (5xx, rate limit, network flake) on issue #N would
+      // silently fall through to PR #N — a different item that happens to share
+      // the number. Only fall through when the issue genuinely doesn't exist
+      // (404); re-throw everything else so the outer catch returns null and the
+      // caller sees a real failure instead of a wrong item. classifyGhError
+      // centralizes the 404/"not found" pattern-matching.
       const stderr = err instanceof Error ? err.message : String(err)
       if (classifyGhError(stderr).type !== 'not_found') {
         throw err

src/main/github/work-item-details.ts

@@ -801,9 +801,9 @@ async function getWorkItemParticipants(
   connectionId?: string | null,
   localGitOptions: LocalGitExecOptions = {}
 ): Promise<GitHubAssignableUser[]> {
-  // Why: issues in a fork live on the upstream remote, so participants must be
-  // resolved via getIssueOwnerRepo to stay consistent with getIssueBodyAndComments.
-  // PRs remain tied to origin via getOwnerRepo.
+  // Why: both issues and PRs in a fork live on the upstream remote, so
+  // participants are resolved via getIssueOwnerRepo/getOwnerRepo, which both
+  // resolve upstream-first, staying consistent with getIssueBodyAndComments.
   const ownerRepo =
     item.type === 'issue'
       ? await getIssueOwnerRepo(repoPath, connectionId, ...localGitOptionArgs(localGitOptions))