fix: cover streamed access errors without alerting on outages

src/core/AnonymizedFile.ts

@@ -1,3 +1,4 @@
+import { notifyOwnerAccessProblem } from "./owner-notifications";
 import { streamResponse } from "./response-stream";
 import { join, basename, dirname } from "path";
 import { Response } from "express";
@@ -355,7 +356,11 @@ export default class AnonymizedFile {
         httpStatus: 413,
       });
     }
-    const content = await this.repository.source?.getFileContent(this);
+    const content = await this.repository.source?.getFileContent(this).catch(error => {
+      void notifyOwnerAccessProblem(this.repository.owner.id, error);
+      throw error;
+    });
+    content.on("error", error => { void notifyOwnerAccessProblem(this.repository.owner.id, error); });
     const cacheWasReset = this.repository.model.isReseted;
     if (cacheWasReset) {
       await this.repository.markCachePresent();
@@ -378,7 +383,7 @@ export default class AnonymizedFile {
     }
 
     // use the streamer service
-    return got.stream(join(config.STREAMER_ENTRYPOINT, "api"), {
+    const content = got.stream(join(config.STREAMER_ENTRYPOINT, "api"), {
       method: "POST",
       headers: requestHeaders(),
       json: {
@@ -395,6 +400,11 @@ export default class AnonymizedFile {
         anonymizerOptions: anonymizer.opt,
       },
     });
+    content.on("error", err => {
+      const { error } = streamerErrorToAnonymous(err, { repoId: this.repository.repoId, filePath: this.anonymizedPath });
+      void notifyOwnerAccessProblem(this.repository.owner.id, error);
+    });
+    return content;
   }
 
   get filePath() {
@@ -472,6 +482,7 @@ export default class AnonymizedFile {
                   filePath: this.anonymizedPath,
                 }
               );
+              void notifyOwnerAccessProblem(this.repository.owner.id, error);
               error.value = this;
               handleError(error, res);
               if (res.headersSent && !res.writableEnded) res.destroy();

src/core/GitHubUtils.ts

@@ -1,3 +1,4 @@
+import { isGitHubRateLimitError, OctokitRequestErrorLike } from "./github-rate-limit";
 import { refreshLegacyToken } from "./legacy-token-refresh";
 import { notifyOwnerAccessProblem } from "./owner-notifications";
 import AnonymizedRepositoryModel from "./model/anonymizedRepositories/anonymizedRepositories.model";
@@ -17,35 +18,7 @@ import { measureStage } from "./request-monitoring";
 
 const logger = createLogger("github");
 
-// Octokit RequestError shape (subset we care about for rate-limit detection).
+export { isGitHubRateLimitError } from "./github-rate-limit";
-interface OctokitRequestErrorLike {
-  status?: number;
-  message?: string;
-  response?: {
-    headers?: Record<string, string | undefined>;
-  };
-}
-
-/**
- * Detect GitHub rate-limit / abuse responses (primary 5k/h or undocumented
- * "secondary" limits) and rewrap them as a translatable AnonymousError so
- * the UI can show a friendly message instead of a raw HttpError stack. The
- * GitHub `x-github-request-id` header is preserved in `detail` so users can
- * cite it if they reach out to GitHub Support.
- */
-export function isGitHubRateLimitError(err: unknown): boolean {
-  if (!err || typeof err !== "object") return false;
-  const e = err as OctokitRequestErrorLike;
-  const msg = (e.message ?? "").toLowerCase();
-  // Primary limits return 403 with "x-ratelimit-remaining: 0"; secondary
-  // limits return 403 (sometimes 429) with "secondary rate limit" in the
-  // body. Match on either signal so we catch both.
-  const status = e.status ?? 0;
-  if (status !== 403 && status !== 429) return false;
-  if (msg.includes("rate limit") || msg.includes("abuse")) return true;
-  const remaining = e.response?.headers?.["x-ratelimit-remaining"];
-  return remaining === "0";
-}
 
 function rateLimitDetail(err: OctokitRequestErrorLike): string {
   const headers = err.response?.headers ?? {};
@@ -299,15 +272,10 @@ export async function checkToken(token: string, ownerId?: string) {
     else await oct.users.getAuthenticated();
     return true;
   } catch (err) {
-    if (
+    // Only a confirmed invalid credential permits fallback. Network failures,
-      err instanceof AnonymousError &&
+    // permission failures and rate limits must retain their original meaning.
-      err.message === "github_rate_limit_exceeded"
+    if ((err as { status?: number })?.status !== 401) throw err;
-    ) {
+    if (ownerId) void notifyOwnerAccessProblem(ownerId, "token_expired");
-      throw err;
-    }
-    if (ownerId && (err as { status?: number }).status === 401) {
-      void notifyOwnerAccessProblem(ownerId, "token_expired");
-    }
     return false;
   }
 }

src/core/github-rate-limit.ts

@@ -0,0 +1,26 @@
+// Octokit RequestError shape (subset we care about for rate-limit detection).
+export interface OctokitRequestErrorLike {
+  status?: number;
+  httpStatus?: number;
+  message?: string;
+  response?: {
+    statusCode?: number;
+    headers?: Record<string, string | undefined>;
+  };
+}
+
+/** Detect GitHub primary and secondary rate limits in API and stream errors. */
+export function isGitHubRateLimitError(err: unknown): boolean {
+  if (!err || typeof err !== "object") return false;
+  const e = err as OctokitRequestErrorLike;
+  const msg = (e.message ?? "").toLowerCase();
+  // Primary limits return 403 with "x-ratelimit-remaining: 0"; secondary
+  // limits return 403 (sometimes 429) with "secondary rate limit" in the
+  // body. Match on either signal so we catch both.
+  const status = e.status ?? e.response?.statusCode ?? e.httpStatus ?? 0;
+  if (status !== 403 && status !== 429) return false;
+  if (msg.includes("rate limit") || msg.includes("abuse")) return true;
+  const remaining = e.response?.headers?.["x-ratelimit-remaining"];
+  return remaining === "0";
+}
+

src/core/owner-notifications.ts

@@ -1,3 +1,4 @@
+import { isGitHubRateLimitError } from "./github-rate-limit";
 import config from "../config";
 import UserModel from "./model/users/users.model";
 import { createLogger } from "./logger";
@@ -7,7 +8,7 @@ const DAY = 24 * 60 * 60 * 1000;
 const accessErrors = new Set([
   "token_expired", "github_oauth_required", "github_app_reconnect_required",
   "github_app_access_required", "repo_not_found", "repository_not_found",
-  "pull_request_not_found", "gist_not_found", "repo_not_accessible", "repo_access_limited", "repo_saml_enforcement",
+  "file_not_accessible", "file_not_found", "pull_request_not_found", "gist_not_found", "repo_not_accessible", "repo_access_limited", "repo_saml_enforcement",
 ]);
 
 export function normalizeEmail(value: unknown): string | null {
@@ -26,14 +27,16 @@ export function notificationEmail(emails?: { email: string; default: boolean }[]
 }
 
 export function isAccessFailure(error: unknown): boolean {
+  if (isGitHubRateLimitError(error)) return false;
   const code = typeof error === "string" ? error : error instanceof Error ? error.message : "";
   const rawStatus = error instanceof Error ? (error as Error & { status?: number }).status : undefined;
-  if (!accessErrors.has(code) && ![401, 404].includes(rawStatus || 0)) return false;
+  if (!accessErrors.has(code) && ![401, 403, 404].includes(rawStatus || 0)) return false;
   if (error instanceof Error) {
-    const upstream = (error as Error & { cause?: { status?: number; httpStatus?: number } }).cause;
+    const upstream = (error as Error & { cause?: { status?: number; httpStatus?: number; response?: { statusCode?: number } } }).cause;
     // Some legacy paths wrap timeouts or 5xx responses as repo_not_found.
     // Only a confirmed access failure should prompt the owner to reconnect.
-    if (upstream && ![401, 403, 404].includes(upstream.status || upstream.httpStatus || 0)) return false;
+    if (isGitHubRateLimitError(upstream)) return false;
+    if (upstream && ![401, 403, 404].includes(upstream.status || upstream.response?.statusCode || upstream.httpStatus || 0)) return false;
     const status = (error as Error & { httpStatus?: number }).httpStatus;
     if (status && (status === 429 || status >= 500)) return false;
   }

test/owner-notifications.test.js

@@ -94,10 +94,25 @@ describe("owner access notifications", () => {
   });
 
   it("recognizes raw authentication and missing-resource errors without alerting for outages", () => {
-    for (const status of [401, 404]) expect(notifications.isAccessFailure(Object.assign(new Error("GitHub rejected access"), { status }))).to.equal(true);
+    for (const status of [401, 403, 404]) expect(notifications.isAccessFailure(Object.assign(new Error("GitHub rejected access"), { status }))).to.equal(true);
     for (const status of [429, 500, 503]) expect(notifications.isAccessFailure(Object.assign(new Error("upstream error"), { status }))).to.equal(false);
   });
 
+  it("suppresses primary and secondary rate limits including wrapped streaming errors", () => {
+    const AnonymousError = require("../src/core/AnonymousError").default;
+    for (const error of [
+      Object.assign(new Error("API rate limit exceeded"), { status: 403 }),
+      Object.assign(new Error("abuse detection"), { status: 403 }),
+      Object.assign(new Error("Forbidden"), { status: 403, response: { headers: { "x-ratelimit-remaining": "0" } } }),
+      Object.assign(new Error("secondary rate limit"), { response: { statusCode: 403 } }),
+    ]) {
+      expect(notifications.isAccessFailure(error)).to.equal(false);
+      expect(notifications.isAccessFailure(new AnonymousError("file_not_accessible", { httpStatus: 403, cause: error }))).to.equal(false);
+    }
+    const forbidden = Object.assign(new Error("SAML enforcement"), { response: { statusCode: 403 } });
+    expect(notifications.isAccessFailure(new AnonymousError("file_not_accessible", { httpStatus: 403, cause: forbidden }))).to.equal(true);
+  });
+
   it("does not let database failures escape", async () => {
     stub(User, "findOneAndUpdate", () => { throw new Error("database down"); });
     await notifyOwnerAccessProblem("owner-id", "token_expired");
@@ -275,3 +290,63 @@ describe("gist and pull-request notification hooks", () => {
     });
   }
 });
+
+describe("streamed file access alerts", () => {
+  const File = require("../src/core/AnonymizedFile").default;
+  const AnonymousError = require("../src/core/AnonymousError").default;
+  const { PassThrough } = require("stream");
+  const got = require("got");
+  const restores = [];
+  const stub = (object, key, value) => { const old = object[key]; restores.push(() => object[key] = old); object[key] = value; };
+  let sent, repository, file;
+  beforeEach(() => {
+    sent = [];
+    stub(notifications, "notifyOwnerAccessProblem", async (owner, error) => { if (notifications.isAccessFailure(error)) sent.push({ owner, error }); });
+    repository = { owner: { id: "saved-owner" }, options: { terms: [] }, model: { source: {} }, repoId: "anon-id", status: "ready", source: {},
+      getToken: async () => "token", generateAnonymizeTransformer: () => new PassThrough() };
+    file = new File({ repository, anonymizedPath: "README.md" });
+  });
+  afterEach(() => { while (restores.length) restores.pop()(); });
+  for (const status of [403, 404, 503]) it(`handles local file download failure ${status}`, async () => {
+    const cause = Object.assign(new Error("upstream"), { response: { statusCode: status } });
+    const error = new AnonymousError(status === 403 ? "file_not_accessible" : status === 404 ? "file_not_found" : "upstream_error", { httpStatus: status, cause });
+    repository.source.getFileContent = async () => { throw error; };
+    let caught;
+    try { await file.content(); } catch (err) { caught = err; }
+    expect(caught).to.equal(error);
+    expect(sent).to.have.length(status === 503 ? 0 : 1);
+    if (sent.length) expect(sent[0].owner).to.equal("saved-owner");
+  });
+  for (const status of [403, 404, 503]) it(`handles remote streamer error ${status}`, async () => {
+    stub(config, "STREAMER_ENTRYPOINT", "http://streamer/");
+    const stream = new PassThrough();
+    stub(got, "stream", () => stream);
+    file.originalPath = async () => "README.md";
+    file.sha = async () => "sha";
+    file.size = async () => 10;
+    const returned = await file.anonymizedContent();
+    expect(returned).to.equal(stream);
+    const error = Object.assign(new Error("streamer failure"), { response: { statusCode: status, body: JSON.stringify({ error: status === 503 ? "upstream_error" : "file_not_accessible" }) } });
+    stream.emit("error", error);
+    expect(sent).to.have.length(status === 503 ? 0 : 1);
+    stream.destroy();
+  });
+});
+
+describe("token validation preserves upstream failures", () => {
+  const tokens = require("../src/core/GitHubUtils");
+  const { registerGitHubToken } = require("../src/core/github-token-context");
+  for (const status of [undefined, 403, 429, 500, 503]) it(`propagates token-check failure ${status}`, async () => {
+    const error = Object.assign(new Error("upstream check failed"), { status });
+    const token = `test-token-check-${status}`;
+    registerGitHubToken(token, { quotaKey: token, publicRepository: "owner/repo", renew: async () => { throw error; } });
+    let caught;
+    try { await tokens.checkToken(token); } catch (err) { caught = err; }
+    expect(caught).to.equal(error);
+  });
+  it("returns false only for a confirmed invalid token", async () => {
+    const token = "test-token-check-invalid";
+    registerGitHubToken(token, { quotaKey: token, publicRepository: "owner/repo", renew: async () => { throw Object.assign(new Error("Bad credentials"), { status: 401 }); } });
+    expect(await tokens.checkToken(token)).to.equal(false);
+  });
+});