Fix balena push "Segmentation fault" on Windows (replace 'mmmagic' wi…

…th 'isBinaryFile')

Connects-to: #1611
Change-type: patch

automation/build-bin.ts

@@ -99,7 +99,6 @@ async function buildPkg() {
 		// [platform, [source path], [destination path]]
 		['*', ['open', 'xdg-open'], ['xdg-open']],
 		['darwin', ['denymount', 'bin', 'denymount'], ['denymount']],
-		['win32', ['mmmagic', 'magic', 'magic.mgc'], ['mmmagic', 'magic.mgc']],
 	];
 	await Bluebird.map(paths, ([platform, source, dest]) => {
 		if (platform === '*' || platform === process.platform) {

lib/utils/eol-conversion.ts

@@ -15,35 +15,19 @@
  * limitations under the License.
  */
 
+import { fs } from 'mz';
+import Logger = require('./logger');
+
+const globalLogger = Logger.getLogger();
+
 // Define file size threshold (bytes) over which analysis/conversion is not performed.
 const LARGE_FILE_THRESHOLD = 10 * 1000 * 1000;
 
-// The list of encodings to convert is intentionally conservative for now
+// Note that `convertEolInPlace()` only works with UTF-8 or single-byte encodings
 const CONVERTIBLE_ENCODINGS = ['ascii', 'utf-8'];
 
-/**
- * Attempt to detect the encoding of a data buffer
- * @param data
- */
-async function detectEncoding(data: Buffer): Promise<string> {
-	const mmmagic = await import('mmmagic');
-	// Instantiate mmmagic for mime encoding analysis
-	const magic = new mmmagic.Magic(mmmagic.MAGIC_MIME_ENCODING);
-
-	// Promisify magic.detect
-	// For some reason, got 'Illegal Invocation' when using:
-	//   const detectEncoding = promisify(magic.detect);
-	return new Promise((resolve, reject) => {
-		magic.detect(data, (err, encoding) => {
-			if (err) {
-				return reject(err);
-			}
-			// mmmagic reports ascii as 'us-ascii', but node Buffer uses 'ascii'
-			encoding = encoding === 'us-ascii' ? 'ascii' : encoding;
-			return resolve(encoding);
-		});
-	});
-}
+// Maximum number of bytes to consider when detecting the file encoding
+const DETECT_MAX_BYTES = 1024;
 
 /**
  * Convert EOL (CRLF → LF) in place, i.e. modifying the input buffer.
@@ -90,10 +74,7 @@ export async function readFileWithEolConversion(
 	filepath: string,
 	convertEol: boolean,
 ): Promise<Buffer> {
-	const { fs } = await import('mz');
 	const fileBuffer = await fs.readFile(filepath);
-	const Logger = await import('./logger');
-	const globalLogger = Logger.getLogger();
 
 	// Skip processing of very large files
 	const fileStats = await fs.stat(filepath);
@@ -111,7 +92,7 @@ export async function readFileWithEolConversion(
 	}
 
 	// Skip further processing of files that don't contain CRLF
-	if (!fileBuffer.includes('\r\n', 0, encoding)) {
+	if (!fileBuffer.includes('\r\n')) {
 		return fileBuffer;
 	}
 
@@ -136,3 +117,123 @@ export async function readFileWithEolConversion(
 		return fileBuffer;
 	}
 }
+
+/**
+ * Attempt to detect the encoding of a data buffer.
+ * Code copied and modified from the npm package 'isbinaryfile' (MIT licence)
+ *   https://github.com/gjtorikian/isBinaryFile/blob/master/src/index.ts
+ *
+ * @returns one of the possible values: '' (empty file), 'utf-8', 'utf-16',
+ * 'utf-32', 'gb-18030', 'pdf', and 'binary'.
+ *
+ * Note: pure ASCII data is identified as 'utf-8' (ASCII is indeed a subset
+ * of UTF-8).
+ *
+ * @param fileBuffer File contents whose encoding should be detected
+ * @param bytesRead Optional "file size" if smaller than the buffer size
+ */
+export async function detectEncoding(
+	fileBuffer: Buffer,
+	bytesRead?: number,
+): Promise<string> {
+	bytesRead = bytesRead ?? fileBuffer.length;
+	// empty file
+	if (bytesRead === 0) {
+		return '';
+	}
+
+	const totalBytes = Math.min(bytesRead, DETECT_MAX_BYTES);
+
+	// UTF-8 BOM
+	if (
+		bytesRead >= 3 &&
+		fileBuffer[0] === 0xef &&
+		fileBuffer[1] === 0xbb &&
+		fileBuffer[2] === 0xbf
+	) {
+		return 'utf-8';
+	}
+
+	// UTF-32 BOM
+	if (
+		bytesRead >= 4 &&
+		fileBuffer[0] === 0x00 &&
+		fileBuffer[1] === 0x00 &&
+		fileBuffer[2] === 0xfe &&
+		fileBuffer[3] === 0xff
+	) {
+		return 'utf-32';
+	}
+
+	// UTF-32 LE BOM
+	if (
+		bytesRead >= 4 &&
+		fileBuffer[0] === 0xff &&
+		fileBuffer[1] === 0xfe &&
+		fileBuffer[2] === 0x00 &&
+		fileBuffer[3] === 0x00
+	) {
+		return 'utf-32';
+	}
+
+	// GB BOM (https://en.wikipedia.org/wiki/GB_18030)
+	if (
+		bytesRead >= 4 &&
+		fileBuffer[0] === 0x84 &&
+		fileBuffer[1] === 0x31 &&
+		fileBuffer[2] === 0x95 &&
+		fileBuffer[3] === 0x33
+	) {
+		return 'gb-18030';
+	}
+
+	if (totalBytes >= 5 && fileBuffer.slice(0, 5).toString() === '%PDF-') {
+		/* PDF. This is binary. */
+		return 'pdf';
+	}
+
+	// UTF-16 BE BOM
+	if (bytesRead >= 2 && fileBuffer[0] === 0xfe && fileBuffer[1] === 0xff) {
+		return 'utf-16';
+	}
+
+	// UTF-16 LE BOM
+	if (bytesRead >= 2 && fileBuffer[0] === 0xff && fileBuffer[1] === 0xfe) {
+		return 'utf-16';
+	}
+
+	for (let i = 0; i < totalBytes; i++) {
+		let c = fileBuffer[i];
+		if (c === 0) {
+			// NULL byte
+			return 'binary';
+		} else if (c === 27) {
+			// ESC character used in ANSI escape sequences for text color (log files)
+			continue;
+		} else if ((c < 7 || c > 14) && (c < 32 || c > 127)) {
+			// UTF-8 detection
+			if (c > 193 && c < 224 && i + 1 < totalBytes) {
+				i++;
+				c = fileBuffer[i];
+				if (c > 127 && c < 192) {
+					continue;
+				}
+			} else if (c > 223 && c < 240 && i + 2 < totalBytes) {
+				i++;
+				c = fileBuffer[i];
+				if (
+					c > 127 &&
+					c < 192 &&
+					fileBuffer[i + 1] > 127 &&
+					fileBuffer[i + 1] < 192
+				) {
+					i++;
+					continue;
+				}
+			}
+			return 'binary';
+		}
+	}
+
+	return 'utf-8';
+}

package.json

@@ -110,7 +110,6 @@
     "@types/lodash": "4.14.112",
     "@types/mixpanel": "2.14.0",
     "@types/mkdirp": "0.5.2",
-    "@types/mmmagic": "0.4.16-alpha",
     "@types/mocha": "^5.2.7",
     "@types/mz": "0.0.32",
     "@types/net-keepalive": "^0.4.0",
@@ -205,7 +204,6 @@
     "minimatch": "^3.0.4",
     "mixpanel": "^0.10.3",
     "mkdirp": "^0.5.1",
-    "mmmagic": "^0.5.3",
     "moment": "^2.24.0",
     "moment-duration-format": "^2.3.2",
     "mz": "^2.7.0",

tests/utils/eol-conversion.spec.ts

@@ -16,8 +16,13 @@
  */
 
 import { expect } from 'chai';
+import { fs } from 'mz';
+import * as path from 'path';
 
-import { convertEolInPlace } from '../../build/utils/eol-conversion';
+import {
+	convertEolInPlace,
+	detectEncoding,
+} from '../../build/utils/eol-conversion';
 
 describe('convertEolInPlace() function', function() {
 	it('should return expected values', () => {
@@ -53,3 +58,43 @@ describe('convertEolInPlace() function', function() {
 		}
 	});
 });
+
+describe('detectEncoding() function', function() {
+	it('should correctly detect the encoding of a few selected files', async () => {
+		const sampleBinary = [
+			'lzma-native/binding-v4.0.6-node-v64-darwin-x64/lzma_native.node',
+			'net-keepalive/node_modules/ffi-napi/build/Release/ffi_bindings.node',
+			'net-keepalive/node_modules/ffi-napi/build/Release/nothing.node',
+			'ext2fs/build/Release/bindings.node',
+			'drivelist/build/Release/drivelist.node',
+			'resin-cli-visuals/node_modules/drivelist/build/Release/drivelist.node',
+			'@balena.io/usb/build/Release/usb_bindings.node',
+			'xxhash/build/Release/hash.node',
+			'mountutils/build/Release/MountUtils.node',
+			'fsevents/build/Release/fse.node',
+		];
+		const sampleText = [
+			'node_modules/.bin/etcher-image-write',
+			'node_modules/.bin/mocha',
+			'node_modules/.bin/rimraf',
+			'node_modules/.bin/gulp',
+			'node_modules/.bin/prettier',
+			'node_modules/.bin/coffeelint',
+			'node_modules/.bin/tsc',
+			'node_modules/.bin/resin-lint',
+			'node_modules/.bin/balena-preload',
+			'node_modules/.bin/catch-uncommitted',
+		];
+
+		for (const fname of sampleBinary) {
+			const buf = await fs.readFile(path.join('node_modules', fname));
+			const encoding = await detectEncoding(buf);
+			expect(encoding).to.equal('binary');
+		}
+		for (const fname of sampleText) {
+			const buf = await fs.readFile(fname);
+			const encoding = await detectEncoding(buf);
+			expect(encoding).to.equal('utf-8');
+		}
+	});
+});