fix(error-tracking): search debug IDs progressively

Author: jhssilva

packages/plugins/error-tracking/src/sourcemaps/debugId.test.ts

@@ -3,16 +3,18 @@
 // Copyright 2019-Present Datadog, Inc.
 
 import { outputFileSync, rmSync } from '@dd/core/helpers/fs';
+import fsp from 'fs/promises';
 import os from 'os';
 import path from 'path';
 
-import { extractDebugId } from './debugId';
+import { DEBUG_ID_SEARCH_CHUNK_BYTES, extractDebugId } from './debugId';
 
 describe('extractDebugId', () => {
     const debugId = '93fd4850-7b77-4f2e-9aa2-ba013e1a5027';
     const tempDir = path.join(os.tmpdir(), 'dd-build-plugins-debug-id-test');
 
     afterEach(() => {
+        jest.restoreAllMocks();
         rmSync(tempDir);
     });
 
@@ -36,6 +38,20 @@ describe('extractDebugId', () => {
         await expect(extractDebugId(filePath)).resolves.toBe(debugId);
     });
 
+    test('Should stop reading after finding the debug ID in the first chunk', async () => {
+        const literal = `ddDebugId:"${debugId}"`;
+        const read = jest.fn(async (buffer: Buffer) => {
+            buffer.write(literal);
+            return { bytesRead: literal.length, buffer };
+        });
+        const close = jest.fn(async () => undefined);
+        jest.spyOn(fsp, 'open').mockResolvedValue({ read, close } as never);
+
+        await expect(extractDebugId('first-chunk.min.js')).resolves.toBe(debugId);
+        expect(read).toHaveBeenCalledTimes(1);
+        expect(close).toHaveBeenCalledTimes(1);
+    });
+
     test('Should return undefined when there is no debug ID in the content', async () => {
         const filePath = path.join(tempDir, 'no-debug-id.min.js');
         outputFileSync(
@@ -46,6 +62,35 @@ describe('extractDebugId', () => {
         await expect(extractDebugId(filePath)).resolves.toBeUndefined();
     });
 
+    test('Should progressively find a debug ID after the first chunk', async () => {
+        const filePath = path.join(tempDir, 'later-debug-id.min.js');
+        outputFileSync(
+            filePath,
+            `${'x'.repeat(DEBUG_ID_SEARCH_CHUNK_BYTES + 100)}ddDebugId:"${debugId}"`,
+        );
+
+        await expect(extractDebugId(filePath)).resolves.toBe(debugId);
+    });
+
+    test('Should find a debug ID split across two chunks', async () => {
+        const filePath = path.join(tempDir, 'split-debug-id.min.js');
+        const literal = `ddDebugId:"${debugId}"`;
+        const literalPrefixBytes = 20;
+        outputFileSync(
+            filePath,
+            `${'x'.repeat(DEBUG_ID_SEARCH_CHUNK_BYTES - literalPrefixBytes)}${literal}`,
+        );
+
+        await expect(extractDebugId(filePath)).resolves.toBe(debugId);
+    });
+
+    test('Should scan to EOF and return undefined when a large file has no debug ID', async () => {
+        const filePath = path.join(tempDir, 'large-no-debug-id.min.js');
+        outputFileSync(filePath, 'x'.repeat(DEBUG_ID_SEARCH_CHUNK_BYTES * 3 + 100));
+
+        await expect(extractDebugId(filePath)).resolves.toBeUndefined();
+    });
+
     test('Should return undefined when the file cannot be read', async () => {
         const filePath = path.join(tempDir, 'missing.min.js');
 

packages/plugins/error-tracking/src/sourcemaps/debugId.ts

@@ -13,26 +13,42 @@ import fsp from 'fs/promises';
 // bundler renaming step.
 const DEBUG_ID_RX = /"?ddDebugId"?:"([0-9a-fA-F-]{36})"/;
 
-// The RUM plugin injects its snippet as a BEFORE-position banner (packages/plugins/rum/src/index.ts),
-// so the ddDebugId literal always lands within the file's first couple hundred bytes, regardless
-// of the file's total size — no need to read the whole (potentially large) minified bundle to
-// find it. Measured against ~2.8k real built chunks (mixed bundlers/minifiers), the match always
-// ended by byte 242; this leaves ~4x headroom for longer service/version strings.
-export const DEBUG_ID_SEARCH_PREFIX_BYTES = 1024;
+// Read progressively so the common case only needs the first KiB, while still supporting
+// bundlers or transforms that place the injected snippet later in the artifact.
+export const DEBUG_ID_SEARCH_CHUNK_BYTES = 1024;
+
+// Keep enough content from the previous chunk to match a debug ID literal split across a read
+// boundary. The longest supported literal is shorter than this overlap.
+const DEBUG_ID_SEARCH_OVERLAP_CHARACTERS = 64;
 
 const matchDebugId = (fileContent: string): string | undefined => {
     return DEBUG_ID_RX.exec(fileContent)?.[1];
 };
 
-// Read only the first DEBUG_ID_SEARCH_PREFIX_BYTES bytes of the file, since that's all
-// we need to find the ddDebugId literal and reading the whole (potentially large)
-// minified bundle into memory would be wasteful.
-const readFilePrefix = async (filePath: string): Promise<string> => {
+// Search in fixed-size reads and stop as soon as the debug ID is found. Only a small overlap is
+// retained between reads, so even the worst case (scanning to EOF) uses bounded memory.
+const readDebugId = async (filePath: string): Promise<string | undefined> => {
     const fd = await fsp.open(filePath, 'r');
     try {
-        const buffer = Buffer.alloc(DEBUG_ID_SEARCH_PREFIX_BYTES);
-        const { bytesRead } = await fd.read(buffer, 0, DEBUG_ID_SEARCH_PREFIX_BYTES, 0);
-        return buffer.toString('utf-8', 0, bytesRead);
+        const buffer = Buffer.alloc(DEBUG_ID_SEARCH_CHUNK_BYTES);
+        let overlap = '';
+        let position = 0;
+
+        while (true) {
+            const { bytesRead } = await fd.read(buffer, 0, DEBUG_ID_SEARCH_CHUNK_BYTES, position);
+            if (bytesRead === 0) {
+                return undefined;
+            }
+
+            const searchableContent = overlap + buffer.toString('utf-8', 0, bytesRead);
+            const debugId = matchDebugId(searchableContent);
+            if (debugId) {
+                return debugId;
+            }
+
+            overlap = searchableContent.slice(-DEBUG_ID_SEARCH_OVERLAP_CHARACTERS);
+            position += bytesRead;
+        }
     } finally {
         await fd.close();
     }
@@ -43,6 +59,5 @@ const readFilePrefix = async (filePath: string): Promise<string> => {
 // rename the file after injection (e.g. webpack/rspack's realContentHash), but the content,
 // and the debug_id embedded in it, is unaffected.
 export const extractDebugId = async (filePath: string): Promise<string | undefined> => {
-    const fileContent = await readFilePrefix(filePath).catch(() => undefined);
-    return fileContent ? matchDebugId(fileContent) : undefined;
+    return readDebugId(filePath).catch(() => undefined);
 };