Fixed stack overflow with StringUtils.

Because we were using Function.apply to create strings, large arrays
were causing problems with browser argument count limits.  This
now splits the arrays up before calling.

Closes #335

Change-Id: Ic94a950997e2f17563ecba8fb628f62c0ed18fc2

lib/util/string_utils.js

@@ -36,7 +36,7 @@ goog.require('shaka.util.Error');
 shaka.util.StringUtils.fromUTF8 = function(data) {
   if (!data) return '';
   // http://stackoverflow.com/a/13691499
-  var utf8 = String.fromCharCode.apply(null, new Uint8Array(data));
+  var utf8 = shaka.util.StringUtils.fromCharCode_(new Uint8Array(data));
   // This converts each character in the string to an escape sequence.  If the
   // character is in the ASCII range, it is not converted; otherwise it is
   // converted to a URI escape sequence.
@@ -85,13 +85,13 @@ shaka.util.StringUtils.fromUTF16 = function(data, littleEndian) {
   }
 
   // Use a DataView to ensure correct endianness.
-  var arr = [];
   var length = data.byteLength / 2;
+  var arr = new Uint16Array(length);
   var dataView = new DataView(buffer);
   for (var i = 0; i < length; i++) {
     arr[i] = dataView.getUint16(i * 2, littleEndian);
   }
-  return String.fromCharCode.apply(null, arr);
+  return shaka.util.StringUtils.fromCharCode_(arr);
 };
 
 
@@ -159,3 +159,22 @@ shaka.util.StringUtils.toUTF8 = function(str) {
   }
   return result.buffer;
 };
+
+
+/**
+ * Creates a new string from the given array of char codes.
+ *
+ * @param {!ITypedArray} args
+ * @return {string}
+ * @private
+ */
+shaka.util.StringUtils.fromCharCode_ = function(args) {
+  var max = 16000;
+  var ret = '';
+  for (var i = 0; i < args.length; i += max) {
+    var subArray = args.subarray(i, i + max);
+    ret += String.fromCharCode.apply(null, subArray);
+  }
+
+  return ret;
+};

test/string_utils_unit.js

@@ -33,17 +33,17 @@ describe('StringUtils', function() {
   it('parses fromUTF16 big-endian', function() {
     // This is big-endian pairs of 16-bit numbers.  This translates into 3
     // Unicode characters where the last is split into a surrogate pair.
-    var arr = [0x00, 0x46, 0x38, 0x01, 0xdc, 0x37, 0x5d, 0x00];
+    var arr = [0x00, 0x46, 0x38, 0x01, 0xd8, 0x01, 0xdc, 0x37];
     var buffer = new Uint8Array(arr).buffer;
-    expect(StringUtils.fromUTF16(buffer, false)).toBe('F\u3801\udc37\u5d00');
+    expect(StringUtils.fromUTF16(buffer, false)).toBe('F\u3801\ud801\udc37');
   });
 
   it('parses fromUTF16 little-endian', function() {
     // This is little-endian pairs of 16-bit numbers.  This translates into 3
     // Unicode characters where the last is split into a surrogate pair.
-    var arr = [0x46, 0x00, 0x01, 0x38, 0x37, 0xdc, 0x00, 0x5d];
+    var arr = [0x46, 0x00, 0x01, 0x38, 0x01, 0xd8, 0x37, 0xdc];
     var buffer = new Uint8Array(arr).buffer;
-    expect(StringUtils.fromUTF16(buffer, true)).toBe('F\u3801\udc37\u5d00');
+    expect(StringUtils.fromUTF16(buffer, true)).toBe('F\u3801\ud801\udc37');
   });
 
   describe('fromBytesAutoDetect', function() {
@@ -102,4 +102,11 @@ describe('StringUtils', function() {
     var buffer = StringUtils.toUTF8(str);
     expect(new Uint8Array(buffer)).toEqual(new Uint8Array(arr));
   });
+
+  it('does not cause stack overflow, #335', function() {
+    var buffer = new Uint8Array(8e5).buffer;  // Well above arg count limit.
+    expect(StringUtils.fromUTF8(buffer).length).toBe(buffer.byteLength);
+    expect(StringUtils.fromUTF16(buffer, true).length)
+        .toBe(buffer.byteLength / 2);
+  });
 });