fix(review): inspect complete file lists for large pull requests

CHANGELOG.md

@@ -18,6 +18,7 @@ checkpoint, and status-only commits are intentionally omitted.
 
 ### Changed
 
+- Large OpenClaw pull requests now reuse each paginated file-list page for bounded context and deterministic classification, accepting the complete sidecar only when filenames are unique and the PR base/head snapshot remains stable.
 - Event-review artifact publication completes as a superseded no-op when the reviewed branch vanished upstream (force-push or deletion) instead of failing the run.
 - Worker record requests now retry transient blank/invalid 2xx bodies from the edge within the bounded budget instead of failing hydration on the first occurrence.
 - Exact reviews of items that closed after enqueue now complete as superseded no-ops, and GitHub-throttled reservations defer as held retries — neither spends the item's review-failure budget.

src/clawsweeper-types.ts

@@ -8,6 +8,7 @@ import type { SchedulerDueCandidate } from "./scheduler-policy.js";
 
 /** Shared ClawSweeper domain, review, scheduling, and dashboard shapes. */
 export const completeActivityContextSymbol = Symbol("completeActivityContext");
+export const completePullFilesSymbol = Symbol("completePullFiles");
 
 export type ItemKind = "issue" | "pull_request";
 export type ApplyKind = ItemKind | "all";
@@ -535,6 +536,7 @@ export interface CompleteActivityContext {
 
 export interface ItemContext {
   [completeActivityContextSymbol]?: CompleteActivityContext;
+  [completePullFilesSymbol]?: readonly unknown[];
   issue: unknown;
   comments: unknown[];
   timeline: unknown[];

src/clawsweeper.ts

@@ -234,7 +234,7 @@ import {
   themedRatingName,
 } from "./clawsweeper-rating.js";
 
-import { completeActivityContextSymbol } from "./clawsweeper-types.js";
+import { completeActivityContextSymbol, completePullFilesSymbol } from "./clawsweeper-types.js";
 import type {
   AcquiredReviewStartLease,
   Action,
@@ -4473,6 +4473,163 @@ function compactSemanticPullFile(value: unknown): unknown {
   };
 }
 
+type PullFileSnapshot = {
+  baseSha: string;
+  headSha: string;
+};
+
+const GITHUB_PULL_FILES_MAX = 3000;
+
+function pullFileSnapshot(value: unknown): PullFileSnapshot | null {
+  const pull = asRecord(value);
+  const baseSha = stringOrUndefined(asRecord(pull.base).sha)?.toLowerCase() ?? "";
+  const headSha = stringOrUndefined(asRecord(pull.head).sha)?.toLowerCase() ?? "";
+  return /^[0-9a-f]{40}$/.test(baseSha) && /^[0-9a-f]{40}$/.test(headSha)
+    ? { baseSha, headSha }
+    : null;
+}
+
+function completePullFilesForDeterministicReview(options: {
+  initialPullRequest: unknown;
+  repo: string;
+  total: number;
+  truncated: boolean;
+  fetchAll: () => unknown[];
+  fetchPullRequest: () => unknown;
+}): unknown[] | null {
+  if (
+    options.repo !== "openclaw/openclaw" ||
+    !options.truncated ||
+    options.total > GITHUB_PULL_FILES_MAX
+  ) {
+    return null;
+  }
+  try {
+    const initialSnapshot = pullFileSnapshot(options.initialPullRequest);
+    if (!initialSnapshot) return null;
+    const files = options.fetchAll();
+    if (files.length !== options.total) return null;
+    const filenames = files.map((file) => stringOrUndefined(asRecord(file).filename)?.trim() ?? "");
+    if (filenames.some((filename) => !filename) || new Set(filenames).size !== filenames.length) {
+      return null;
+    }
+    const refreshedSnapshot = pullFileSnapshot(options.fetchPullRequest());
+    if (
+      !refreshedSnapshot ||
+      refreshedSnapshot.baseSha !== initialSnapshot.baseSha ||
+      refreshedSnapshot.headSha !== initialSnapshot.headSha
+    ) {
+      return null;
+    }
+    return files.map(compactSemanticPullFile);
+  } catch {
+    // Keep the bounded window usable; deterministic classifiers retain the unknown marker.
+    return null;
+  }
+}
+
+function pullFileReviewHydration(options: {
+  repo: string;
+  path: string;
+  total: unknown;
+  pullRequest: unknown;
+  fetchPage: (path: string, page: number) => unknown[];
+  fetchPullRequest: () => unknown;
+}): { window: ContextHydration<unknown>; completeFiles: unknown[] | null } {
+  const pages = new Map<number, unknown[]>();
+  const readPage = (path: string, page: number): unknown[] => {
+    const cached = pages.get(page);
+    if (cached) return cached;
+    const files = options.fetchPage(path, page);
+    pages.set(page, files);
+    return files;
+  };
+  const window = ghPagedContextWindow<unknown>(options.path, options.total, 80, {
+    page: readPage,
+  });
+  const completeFiles = completePullFilesForDeterministicReview({
+    initialPullRequest: options.pullRequest,
+    repo: options.repo,
+    total: window.total,
+    truncated: window.truncated,
+    fetchAll: () => {
+      const files: unknown[] = [];
+      const pageCount = Math.ceil(window.total / 100);
+      for (let page = 1; page <= pageCount; page += 1) {
+        const pageFiles = readPage(options.path, page);
+        files.push(...pageFiles);
+        if (pageFiles.length < 100) break;
+      }
+      return files;
+    },
+    fetchPullRequest: options.fetchPullRequest,
+  });
+  return { window, completeFiles };
+}
+
+export function completePullFilesForDeterministicReviewForTest(options: {
+  initialBaseSha?: string;
+  initialHeadSha?: string;
+  refreshedBaseSha?: string;
+  refreshedHeadSha?: string;
+  repo?: string;
+  total: number;
+  truncated: boolean;
+  files: unknown[];
+  fetchError?: Error;
+}): unknown[] | null {
+  const initialBaseSha = options.initialBaseSha ?? "a".repeat(40);
+  const initialHeadSha = options.initialHeadSha ?? "b".repeat(40);
+  return completePullFilesForDeterministicReview({
+    initialPullRequest: {
+      base: { sha: initialBaseSha },
+      head: { sha: initialHeadSha },
+    },
+    repo: options.repo ?? "openclaw/openclaw",
+    total: options.total,
+    truncated: options.truncated,
+    fetchAll: () => {
+      if (options.fetchError) throw options.fetchError;
+      return options.files;
+    },
+    fetchPullRequest: () => ({
+      base: { sha: options.refreshedBaseSha ?? initialBaseSha },
+      head: { sha: options.refreshedHeadSha ?? initialHeadSha },
+    }),
+  });
+}
+
+export function pullFileReviewHydrationForTest(options: {
+  total: number;
+  pages: readonly unknown[][];
+}): {
+  completeFiles: unknown[] | null;
+  hydrated: number;
+  pageCalls: number[];
+} {
+  const pageCalls: number[] = [];
+  const pullRequest = {
+    base: { sha: "a".repeat(40) },
+    head: { sha: "b".repeat(40) },
+  };
+  const result = pullFileReviewHydration({
+    repo: "openclaw/openclaw",
+    path: "repos/openclaw/openclaw/pulls/123/files",
+    total: options.total,
+    pullRequest,
+    fetchPage: (_path, page) => {
+      pageCalls.push(page);
+      return [...(options.pages[page - 1] ?? [])];
+    },
+    fetchPullRequest: () => pullRequest,
+  });
+  return {
+    completeFiles: result.completeFiles,
+    hydrated: result.window.hydrated,
+    pageCalls,
+  };
+}
+
 function normalizedPullFileStatus(value: unknown): string {
   const status = typeof value === "string" ? value.trim().toLowerCase() : "";
   if (status === "m" || status === "modified" || status === "changed") return "modified";
@@ -6563,12 +6720,19 @@ function collectItemContext(
   if (item.kind === "pull_request") {
     pullRequest = ghJson<unknown>(["api", `repos/${targetRepo()}/pulls/${item.number}`]);
     const pullRecord = asRecord(pullRequest);
-    const pullFilesWindow = ghPagedContextWindow<unknown>(
-      `repos/${targetRepo()}/pulls/${item.number}/files`,
-      pullRecord.changed_files,
-      80,
-    );
+    const pullFilesPath = `repos/${targetRepo()}/pulls/${item.number}/files`;
+    const pullFileHydration = pullFileReviewHydration({
+      repo: targetRepo(),
+      path: pullFilesPath,
+      total: pullRecord.changed_files,
+      pullRequest,
+      fetchPage: ghPage,
+      fetchPullRequest: () =>
+        ghJson<unknown>(["api", `repos/${targetRepo()}/pulls/${item.number}`]),
+    });
+    const pullFilesWindow = pullFileHydration.window;
     const pullFiles = pullFilesWindow.items;
+    const completePullFiles = pullFileHydration.completeFiles;
     const pullCommitsWindow = ghPagedContextWindow<unknown>(
       `repos/${targetRepo()}/pulls/${item.number}/commits`,
       pullRecord.commits,
@@ -6599,6 +6763,12 @@ function collectItemContext(
       fullPullReviewComments.length >= pullReviewCommentsWindow.total;
     context.pullRequest = compactPullRequest(pullRequest);
     context.pullFiles = compactMappedWindow(pullFiles, pullFilesWindow.total, 80, compactPullFile);
+    if (completePullFiles) {
+      Object.defineProperty(context, completePullFilesSymbol, {
+        value: completePullFiles,
+        enumerable: false,
+      });
+    }
     context.semanticPullFiles =
       options.reviewCacheDigest &&
       options.reviewCacheGitDir &&
@@ -10724,13 +10894,28 @@ function pullRequestFilePathsFromReport(markdown: string): string[] {
   return frontMatterStringArray(markdown, "pull_files");
 }
 
+function deterministicPullFilesFromContext(context: ItemContext): {
+  complete: boolean;
+  files: readonly unknown[];
+} {
+  const completePullFiles = context[completePullFilesSymbol];
+  if (context.counts?.pullFilesTruncated && completePullFiles) {
+    return { complete: true, files: completePullFiles };
+  }
+  return {
+    complete: context.counts?.pullFilesTruncated !== true,
+    files: context.pullFiles ?? [],
+  };
+}
+
 function configSurfaceChangeFromContext(repo: string, context: ItemContext): ConfigSurfaceChange {
   if (repo !== "openclaw/openclaw") {
     return { change: false, keys: [] };
   }
 
   const keys = new Set<string>();
-  for (const entry of context.pullFiles ?? []) {
+  const deterministicPullFiles = deterministicPullFilesFromContext(context);
+  for (const entry of deterministicPullFiles.files) {
     const file = asRecord(entry);
     const path = typeof file.filename === "string" ? file.filename.trim() : "";
     const previousPath =
@@ -10760,14 +10945,15 @@ function configSurfaceChangeFromContext(repo: string, context: ItemContext): Con
     }
   }
 
-  if (context.counts?.pullFilesTruncated) {
+  if (!deterministicPullFiles.complete) {
     keys.add("unknown-truncated-pull-files");
   }
 
   return { change: keys.size > 0, keys: [...keys].sort() };
 }
 
 export function configSurfaceChangeFromPullFilesForTest(options: {
+  completePullFiles?: readonly unknown[];
   repo?: string;
   pullFiles?: unknown[];
   pullFilesTruncated?: boolean;
@@ -10782,6 +10968,8 @@ export function configSurfaceChangeFromPullFilesForTest(options: {
     counts,
   };
   if (options.pullFiles !== undefined) context.pullFiles = options.pullFiles;
+  if (options.completePullFiles !== undefined)
+    context[completePullFilesSymbol] = options.completePullFiles;
   return configSurfaceChangeFromContext(options.repo ?? "openclaw/openclaw", context);
 }
 
@@ -10791,7 +10979,8 @@ function dataModelChangeFromContext(repo: string, context: ItemContext): DataMod
   }
 
   const surfaces = new Set<string>();
-  for (const entry of context.pullFiles ?? []) {
+  const deterministicPullFiles = deterministicPullFilesFromContext(context);
+  for (const entry of deterministicPullFiles.files) {
     const file = asRecord(entry);
     const path = typeof file.filename === "string" ? file.filename.trim() : "";
     const previousPath =
@@ -10823,14 +11012,15 @@ function dataModelChangeFromContext(repo: string, context: ItemContext): DataMod
     }
   }
 
-  if (context.counts?.pullFilesTruncated) {
+  if (!deterministicPullFiles.complete) {
     surfaces.add("unknown-truncated-pull-files");
   }
 
   return { change: surfaces.size > 0, surfaces: [...surfaces].sort() };
 }
 
 export function dataModelChangeFromPullFilesForTest(options: {
+  completePullFiles?: readonly unknown[];
   repo?: string;
   pullFiles?: unknown[];
   pullFilesTruncated?: boolean;
@@ -10845,6 +11035,8 @@ export function dataModelChangeFromPullFilesForTest(options: {
     counts,
   };
   if (options.pullFiles !== undefined) context.pullFiles = options.pullFiles;
+  if (options.completePullFiles !== undefined)
+    context[completePullFilesSymbol] = options.completePullFiles;
   return dataModelChangeFromContext(options.repo ?? "openclaw/openclaw", context);
 }