fix(miner): bound self-review manifest reads

Author: JSONbored

packages/gittensory-miner/lib/self-review-context.js

@@ -1,4 +1,4 @@
-import { buildCollisionReport, parseFocusManifestContent } from "@jsonbored/gittensory-engine";
+import { buildCollisionReport, MAX_FOCUS_MANIFEST_BYTES, parseFocusManifestContent } from "@jsonbored/gittensory-engine";
 
 // Real SelfReviewContext fetcher (#5145, Wave 3.5). Builds the context object the miner's self-review pass
 // (packages/gittensory-engine/src/miner/self-review-adapter.ts) needs, at the SAME fidelity the live gate's
@@ -210,15 +210,49 @@ async function fetchOpenPullRequestRecords(target, resolved) {
   return payloads.map((pr) => toPullRequestRecord(`${target.owner}/${target.repo}`, pr));
 }
 
-// Mirrors src/signals/focus-manifest-loader.ts's fetchRepoFocusManifestFile exactly: unauthenticated GET
-// against raw.githubusercontent.com, first candidate path that resolves wins.
+// Mirrors src/signals/focus-manifest-loader.ts's raw-content lookup order and bounded body read:
+// first candidate path that resolves wins, but hostile manifests never exceed the parser byte cap in memory.
+async function readBoundedManifestResponseText(response) {
+  const contentLength = response.headers?.get?.("content-length") ?? null;
+  if (contentLength !== null) {
+    const parsedLength = Number.parseInt(contentLength, 10);
+    if (Number.isFinite(parsedLength) && parsedLength > MAX_FOCUS_MANIFEST_BYTES) return null;
+  }
+  if (!response.body?.getReader) {
+    const text = await response.text();
+    if (typeof text !== "string") return null;
+    return new TextEncoder().encode(text).byteLength > MAX_FOCUS_MANIFEST_BYTES ? null : text;
+  }
+
+  const reader = response.body.getReader();
+  const decoder = new TextDecoder();
+  let totalBytes = 0;
+  let text = "";
+  try {
+    while (true) {
+      const { done, value } = await reader.read();
+      if (done) break;
+      totalBytes += value.byteLength;
+      if (totalBytes > MAX_FOCUS_MANIFEST_BYTES) {
+        await reader.cancel();
+        return null;
+      }
+      text += decoder.decode(value, { stream: true });
+    }
+    text += decoder.decode();
+    return text;
+  } finally {
+    reader.releaseLock();
+  }
+}
+
 async function fetchManifestContent(target, resolved) {
   for (const path of MANIFEST_FILE_CANDIDATES) {
     const url = `${resolved.rawContentBaseUrl}/${encodeURIComponent(target.owner)}/${encodeURIComponent(target.repo)}/HEAD/${path}`;
     try {
       const response = await resolved.fetchImpl(url, { method: "GET", headers: { accept: "application/json", "user-agent": "gittensory-miner" } });
       if (response.ok) {
-        const text = await response.text();
+        const text = await readBoundedManifestResponseText(response);
         if (typeof text === "string") return text;
       }
     } catch {

test/unit/miner-self-review-context.test.ts

@@ -4,6 +4,7 @@ vi.mock("@jsonbored/gittensory-engine", async () => {
   return import("../../packages/gittensory-engine/src/index");
 });
 
+import { MAX_FOCUS_MANIFEST_BYTES } from "../../packages/gittensory-engine/src/index";
 import { fetchSelfReviewContext } from "../../packages/gittensory-miner/lib/self-review-context.js";
 
 function jsonResponse(body: unknown, status = 200) {
@@ -24,6 +25,40 @@ function textResponse(text: string, status = 200) {
   };
 }
 
+function oversizedContentLengthResponse() {
+  return {
+    ok: true,
+    status: 200,
+    headers: new Headers({ "content-length": String(MAX_FOCUS_MANIFEST_BYTES + 1) }),
+    json: async () => null,
+    text: async () => {
+      throw new Error("oversized manifest should not be materialized");
+    },
+  };
+}
+
+function chunkedManifestResponse(chunks: Uint8Array[], onCancel: () => void) {
+  let index = 0;
+  return {
+    ok: true,
+    status: 200,
+    headers: new Headers(),
+    body: {
+      getReader: () => ({
+        read: async () => (index < chunks.length ? { done: false, value: chunks[index++] } : { done: true, value: undefined }),
+        cancel: async () => {
+          onCancel();
+        },
+        releaseLock: () => {},
+      }),
+    },
+    json: async () => null,
+    text: async () => {
+      throw new Error("streaming manifest should not use response.text()");
+    },
+  };
+}
+
 const REPO_PAYLOAD = {
   name: "widgets",
   full_name: "acme/widgets",
@@ -283,6 +318,43 @@ describe("fetchSelfReviewContext (#5145)", () => {
     expect(result.manifest.gate).toBeDefined();
   });
 
+  it("rejects oversized manifest responses from content-length before reading the body", async () => {
+    let manifestRequests = 0;
+    const fetchImpl = async (url: string) => {
+      if (url.includes("raw.githubusercontent.com")) {
+        manifestRequests += 1;
+        return oversizedContentLengthResponse();
+      }
+      if (url.includes("/repos/acme/widgets/issues")) return jsonResponse([]);
+      if (url.includes("/repos/acme/widgets/pulls")) return jsonResponse([]);
+      if (url.includes("/repos/acme/widgets")) return jsonResponse(REPO_PAYLOAD);
+      if (url.includes("api.gittensor.io/miners")) return jsonResponse([]);
+      return jsonResponse(null, 404);
+    };
+
+    const result = await fetchSelfReviewContext("acme/widgets", { fetchImpl: fetchImpl as never });
+    expect(manifestRequests).toBe(4);
+    expect(result.manifest.present).toBe(false);
+    expect(result.manifest.warnings).toEqual([]);
+  });
+
+  it("cancels chunked manifest responses once the byte cap is exceeded", async () => {
+    const chunks = [new Uint8Array(MAX_FOCUS_MANIFEST_BYTES), new Uint8Array([123])];
+    let cancelCount = 0;
+    const fetchImpl = async (url: string) => {
+      if (url.includes("raw.githubusercontent.com")) return chunkedManifestResponse(chunks, () => { cancelCount += 1; });
+      if (url.includes("/repos/acme/widgets/issues")) return jsonResponse([]);
+      if (url.includes("/repos/acme/widgets/pulls")) return jsonResponse([]);
+      if (url.includes("/repos/acme/widgets")) return jsonResponse(REPO_PAYLOAD);
+      if (url.includes("api.gittensor.io/miners")) return jsonResponse([]);
+      return jsonResponse(null, 404);
+    };
+
+    const result = await fetchSelfReviewContext("acme/widgets", { fetchImpl: fetchImpl as never });
+    expect(cancelCount).toBe(4);
+    expect(result.manifest.present).toBe(false);
+  });
+
   it("stops at the first manifest candidate that resolves, skipping the rest", async () => {
     let requestedPaths: string[] = [];
     const fetchImpl = async (url: string) => {