feat: use wasm content hash implementation (microsoft#288)

The hash speed was bugging me, particularly since we're soon going to
need to hash files twice (as of microsoft#277) so last night I wanted to see if a
webassembly version would fare better. Turns out it does. Hashing 1KB:

```
              7,520 ops/sec > current (1x)
             72,500 ops/sec > wasm (9.64x)

  Benches: 2
  Fastest: wasm
  Elapsed: 10.9s
```

The current JS algorithm only is hasing 7.5MB/s, which may sound like
a decent amount, but remember some node_modules can be massive. This
brings up a bit.

![](https://i.redd.it/tfugj4n3l6ez.png)

gulpfile.js

@@ -225,7 +225,9 @@ gulp.task('package:copy-extension-files', () =>
         base: buildDir,
       },
     ),
-    gulp.src('node_modules/source-map/lib/*.wasm').pipe(rename({ dirname: 'src' })),
+    gulp
+      .src(['node_modules/source-map/lib/*.wasm', 'node_modules/@c4312/chromehash/pkg/*.wasm'])
+      .pipe(rename({ dirname: 'src' })),
     gulp.src(`${buildDir}/src/**/*.sh`).pipe(rename({ dirname: 'src' })),
   ).pipe(gulp.dest(distDir)),
 );

package.json

@@ -49,11 +49,11 @@
     "intTest": "mocha --exit --timeout 20000 -s 3500 -u tdd --colors --reporter out/src/int-chrome/testSupport/loggingReporter.js \"./out/src/int-chrome/**/*.test.js\""
   },
   "dependencies": {
+    "@c4312/chromehash": "^0.2.0",
     "color": "^3.1.2",
     "glob-stream": "^6.1.0",
     "js-beautify": "^1.10.0",
     "jsonc-parser": "^2.2.0",
-    "long": "^4.0.0",
     "micromatch": "^4.0.2",
     "source-map": "^0.7.3",
     "split2": "^3.1.1",

src/common/hash/hash.ts

@@ -1,110 +1,9 @@
 /*---------------------------------------------------------
  * Copyright (C) Microsoft Corporation. All rights reserved.
  *--------------------------------------------------------*/
-import Long from 'long';
+import { hash } from '@c4312/chromehash';
 import { readFileRaw } from '../fsUtils';
 
-/**
- * An implementation of the Chrome content hashing algorithm used to verify
- * whether files on disk are the same as those in the debug session.
- */
-function calculateHash(input: Buffer): string {
-  const prime = [
-    new Long(0x3fb75161, 0, true),
-    new Long(0xab1f4e4f, 0, true),
-    new Long(0x82675bc5, 0, true),
-    new Long(0xcd924d35, 0, true),
-    new Long(0x81abe279, 0, true),
-  ];
-  const random = [
-    new Long(0x67452301, 0, true),
-    new Long(0xefcdab89, 0, true),
-    new Long(0x98badcfe, 0, true),
-    new Long(0x10325476, 0, true),
-    new Long(0xc3d2e1f0, 0, true),
-  ];
-  const randomOdd = [
-    new Long(0xb4663807, 0, true),
-    new Long(0xcc322bf5, 0, true),
-    new Long(0xd4f91bbd, 0, true),
-    new Long(0xa7bea11d, 0, true),
-    new Long(0x8f462907, 0, true),
-  ];
-
-  const hashes = [
-    new Long(0, 0, true),
-    new Long(0, 0, true),
-    new Long(0, 0, true),
-    new Long(0, 0, true),
-    new Long(0, 0, true),
-  ];
-  const zi = [
-    new Long(1, 0, true),
-    new Long(1, 0, true),
-    new Long(1, 0, true),
-    new Long(1, 0, true),
-    new Long(1, 0, true),
-  ];
-  const k0x7FFFFFFF = new Long(0x7fffffff);
-
-  const buffer = normalize(input);
-  const inc = 4;
-
-  // First pass reads 4 bytes at a time
-  let current = 0;
-  for (let i = 0; i < buffer.byteLength - (buffer.byteLength % inc); i += inc) {
-    const d = buffer.readUInt32LE(i);
-    const v = d;
-
-    const xi = new Long(v).mul(randomOdd[current]).and(k0x7FFFFFFF);
-    hashes[current] = hashes[current].add(zi[current].mul(xi)).mod(prime[current]);
-    zi[current] = zi[current].mul(random[current]).mod(prime[current]);
-    current = current === hashes.length - 1 ? 0 : current + 1;
-  }
-
-  // If we have an odd number of bytes, calculate the rest of the hash
-  if (buffer.byteLength % inc) {
-    let v = 0;
-    for (let i = buffer.byteLength - (buffer.byteLength % inc); i < buffer.byteLength; ++i) {
-      v <<= 8;
-      v |= buffer.readUInt8(i);
-    }
-    const xi = new Long(v).mul(randomOdd[current]).and(k0x7FFFFFFF);
-    hashes[current] = hashes[current].add(zi[current].mul(xi)).mod(prime[current]);
-    zi[current] = zi[current].mul(random[current]).mod(prime[current]);
-    current = current === hashes.length - 1 ? 0 : current + 1;
-  }
-
-  for (let i = 0; i < hashes.length; ++i) {
-    hashes[i] = hashes[i].add(zi[i].mul(prime[i].sub(1))).mod(prime[i]);
-  }
-
-  let hash = '';
-  for (let i = 0; i < hashes.length; ++i) {
-    hash += hashes[i].toString(16).padStart(8, '0');
-  }
-  return hash;
-}
-
-const hasUTF8BOM = (buffer: Buffer) =>
-  buffer.byteLength >= 3 && buffer[0] === 0xef && buffer[1] === 0xbb && buffer[2] === 0xbf;
-const hasUtf16LEBOM = (buffer: Buffer) =>
-  buffer.byteLength >= 2 && buffer[0] === 0xff && buffer[1] === 0xfe;
-const hasUtf16BEBOM = (buffer: Buffer) =>
-  buffer.byteLength >= 2 && buffer[0] === 0xfe && buffer[1] === 0xff;
-
-function normalize(buffer: Buffer): Buffer {
-  if (hasUTF8BOM(buffer)) return normalize(buffer.slice(3));
-  if (hasUtf16LEBOM(buffer)) return buffer.slice(2);
-  if (hasUtf16BEBOM(buffer)) return buffer.slice(2).swap16();
-  // if no byte order mark, assume it's utf8
-  return utf8ToUtf16(buffer);
-}
-
-function utf8ToUtf16(buffer: Buffer) {
-  return Buffer.from(buffer.toString('utf8'), 'utf16le');
-}
-
 /**
  * Message sent to the hash worker.
  */
@@ -120,12 +19,12 @@ function startWorker(send: (message: HashResponse) => void) {
     if ('file' in msg) {
       const file = msg.file;
       readFileRaw(file)
-        .then(data => send({ id: msg.id, hash: calculateHash(data) }))
+        .then(data => send({ id: msg.id, hash: hash(data) }))
         .catch(() => send({ id: msg.id }));
     } else if ('data' in msg) {
       send({
         id: msg.id,
-        hash: calculateHash(msg.data instanceof Buffer ? msg.data : Buffer.from(msg.data, 'utf-8')),
+        hash: hash(msg.data instanceof Buffer ? msg.data : Buffer.from(msg.data, 'utf-8')),
       });
     }
   });