From 51df9e0d8c360fb5ceea8ada43ada891cfb2b382 Mon Sep 17 00:00:00 2001
From: Brian Lalor <blalor@wandb.com>
Date: Thu, 20 Aug 2026 13:28:50 -0400
Subject: [PATCH] fix: dedupe GitHub App reports

---
 .changeset/fresh-app-deduplication.md |  5 +++++
 src/Github.test.ts                    | 18 ++++++++++++++++--
 src/Github.ts                         | 10 +++++-----
 test/github.ts                        |  2 ++
 4 files changed, 28 insertions(+), 7 deletions(-)
 create mode 100644 .changeset/fresh-app-deduplication.md

diff --git a/.changeset/fresh-app-deduplication.md b/.changeset/fresh-app-deduplication.md
new file mode 100644
index 0000000..0036f97
--- /dev/null
+++ b/.changeset/fresh-app-deduplication.md
@@ -0,0 +1,5 @@
+---
+'frog': patch
+---
+
+Matched labelled reports for GitHub App tokens and checked recent issues first during unlabelled fallback scans.
diff --git a/src/Github.test.ts b/src/Github.test.ts
index e4b060a..8a2c137 100644
--- a/src/Github.test.ts
+++ b/src/Github.test.ts
@@ -543,7 +543,7 @@ describe('find', () => {
     ).toBeUndefined()
   })
 
-  test('behavior: caps pagination at 50 pages', async () => {
+  test('behavior: scans the newest issues before reaching the 50-page cap', async () => {
     const seeded = Array.from({ length: 5_001 }, (_, index) => ({
       labels: [],
       title: index === 5_000 ? title : `Unrelated ${index}`,
@@ -552,7 +552,7 @@ describe('find', () => {
 
     expect(
       await Github.find(client(instance.url), { hash: Github.hash(title), repo, title }),
-    ).toBeUndefined()
+    ).toMatchObject({ number: 5_001 })
     expect(
       instance.requests.filter(
         (request) => request.method === 'GET' && request.path === '/repos/wevm/viem/issues',
@@ -629,6 +629,20 @@ describe('index', () => {
 })
 
 describe('matcher', () => {
+  test('behavior: indexes labelled issues without reported push access', async () => {
+    const instance = await github({ [repo]: [{ title }] }, { pushAccess: [] })
+
+    const matcher = await Github.matcher(client(instance.url), { label: 'friction', repo })
+
+    expect(matcher.labelled).toBe(false)
+    expect(
+      instance.requests.filter(
+        (request) => request.method === 'GET' && request.path === `/repos/${repo}/issues`,
+      ),
+    ).toHaveLength(1)
+    await expect(matcher.match(title)).resolves.toMatchObject({ number: 1 })
+  })
+
   test('behavior: matches a report before a changed title', async () => {
     const report = 'acme/app:entry-a'
     const instance = await github(
diff --git a/src/Github.ts b/src/Github.ts
index c128c89..e7da797 100644
--- a/src/Github.ts
+++ b/src/Github.ts
@@ -1010,7 +1010,7 @@ async function listAll(client: Client, options: { repo: string }): Promise<reado
   for (let page = 1; page <= 50; page++) {
     const response = await client.issues.listForRepo({
       ...split(options.repo),
-      direction: 'asc',
+      direction: 'desc',
       page,
       per_page: 100,
       sort: 'created',
@@ -1064,9 +1064,9 @@ export type Matcher = {
 /**
  * Prepares dedupe for a repository, choosing a strategy the token can actually use.
  *
- * With push access, issues are indexed by label first. A miss falls back to one unfiltered listing,
- * which also covers labels removed by users or dropped during creation. Without push access, that
- * fallback is the primary index.
+ * Issues are indexed by label first. A miss falls back to one unfiltered listing, which also covers
+ * labels removed by users or dropped during creation. Push access only determines whether labels are
+ * expected to stick on newly created issues.
  *
  * @param client - Authenticated client for the repository.
  */
@@ -1075,7 +1075,7 @@ export async function matcher(client: Client, options: matcher.Options): Promise
   const accepts = (issue: Issue) =>
     !options.exclude?.(issue) &&
     (!options.expectedAuthor || issue.author === options.expectedAuthor)
-  const labelledIssues = push ? (await list(client, options)).filter(accepts) : []
+  const labelledIssues = (await list(client, options)).filter(accepts)
   const titleOverrides = new Map<number, string>()
   const updateIndex = (index: Map<string, Issue>, issues: readonly Issue[]) => {
     index.clear()
diff --git a/test/github.ts b/test/github.ts
index cbac478..42a22ac 100644
--- a/test/github.ts
+++ b/test/github.ts
@@ -370,6 +370,7 @@ export async function github(seed: Seed = {}, options: Options = {}): Promise<In
         const label = url.searchParams.get('labels')
         const creator = url.searchParams.get('creator')
         const state = url.searchParams.get('state') ?? 'open'
+        const direction = url.searchParams.get('direction') ?? 'desc'
         const page = Number(url.searchParams.get('page') ?? '1')
         const perPage = Number(url.searchParams.get('per_page') ?? '30')
 
@@ -381,6 +382,7 @@ export async function github(seed: Seed = {}, options: Options = {}): Promise<In
             ...issue,
             ...(pulls.has(`${repo}#${issue.number}`) ? { pull_request: { url: '' } } : {}),
           }))
+        if (direction === 'desc') matching.reverse()
 
         return json(response, 200, matching.slice((page - 1) * perPage, page * perPage))
       }