Invalid code generated with uglify-js 3.4.9

[alippai]

This:

/**
 * Encode javascript string as utf8 byte array
 * @param {string} str
 * @returns {Uint8Array}
 */
function encode(str) {
  if (typeof str !== 'string') throw new Error('Only strings are supported');
  if (str === '') return new Uint8Array([]);

  // the resulting byte count is at least the length of the string
  const length = str.length;
  let i;
  let cp;
  let cursor = 0;
  // First, we count the needed number of bytes, by iterating through once
  for (i = 0; i < length; i += cp > 0xffff ? 2 : 1) {
    cp = codePointAt(str, i);
    // Expecting most common codepoints below 0x80, so checking for that first
    if (cp < 0x80) {
      cursor += 1;
    } else if (cp < 0x800) {
      cursor += 2;
    } else if (cp < 0x10000) {
      cursor += 3;
    } else if (cp < 0x200000) {
      cursor += 4;
    } else {
      encodeError(cp);
    }
  }
  // Allocate the buffer of suitable size
  const bytes = new Uint8Array(cursor);
  // Fill the right sized buffer with utf8 data
  for (i = 0, cursor = 0; i < length; i += cp > 0xffff ? 2 : 1) {
    cp = codePointAt(str, i);
    if (cp < 0x80) {
      bytes[cursor++] = cp;
    } else if (cp < 0x800) {
      bytes[cursor++] = (cp >>> 6) | 0xc0;
      bytes[cursor++] = (cp & 0x3f) | 0x80;
    } else if (cp < 0x10000) {
      bytes[cursor++] = (cp >>> 12) | 0xe0;
      bytes[cursor++] = ((cp >>> 6) & 0x3f) | 0x80;
      bytes[cursor++] = (cp & 0x3f) | 0x80;
    } else {
      bytes[cursor++] = (cp >>> 18) | 0xf0;
      bytes[cursor++] = ((cp >>> 12) & 0x3f) | 0x80;
      bytes[cursor++] = ((cp >>> 6) & 0x3f) | 0x80;
      bytes[cursor++] = (cp & 0x3f) | 0x80;
    }
  }
  return bytes;
}

Is compiled to (v3.4.9):

function e(e) {
    if ('string' != typeof e) throw new Error('Only strings are supported');
    if ('' === e) return new Uint8Array([]);
    var t,
      r,
      n = e.length,
      i = 0;
    for (t = 0; t < n; t += R(r))
      (r = I(e, t)), r < 128 ? (i += 1) : r < 2048 ? (i += 2) : r < 65536 ? (i += 3) : r < 2097152 ? (i += 4) : C(r);
    var a = new Uint8Array(i);
    for (t = 0, i = 0; t < n; t += 65535 < r ? 2 : 1)
      (r = I(e, t)),
        (a[i++] =
          r < 128
            ? r
            : ((a[i++] =
                r < 2048
                  ? (r >>> 6) | 192
                  : ((a[i++] = r < 65536 ? (r >>> 12) | 224 : ((a[i++] = (r >>> 18) | 240), ((r >>> 12) & 63) | 128)),
                    ((r >>> 6) & 63) | 128)),
              (63 & r) | 128));
    return a;
  }

Instead of this (v3.4.8):

function e(e) {
    if ('string' != typeof e) throw new Error('Only strings are supported');
    if ('' === e) return new Uint8Array([]);
    var t,
      r,
      n = e.length,
      i = 0;
    for (t = 0; t < n; t += R(r))
      (r = I(e, t)), r < 128 ? (i += 1) : r < 2048 ? (i += 2) : r < 65536 ? (i += 3) : r < 2097152 ? (i += 4) : C(r);
    var a = new Uint8Array(i);
    for (t = 0, i = 0; t < n; t += 65535 < r ? 2 : 1)
      (r = I(e, t)),
        r < 128
          ? (a[i++] = r)
          : (r < 2048
              ? (a[i++] = (r >>> 6) | 192)
              : (r < 65536
                  ? (a[i++] = (r >>> 12) | 224)
                  : ((a[i++] = (r >>> 18) | 240), (a[i++] = ((r >>> 12) & 63) | 128)),
                (a[i++] = ((r >>> 6) & 63) | 128)),
            (a[i++] = (63 & r) | 128));
    return a;
  }

[alippai]

Which encodes sequence c2 a0 into a0 c2 (flips the order somewhat)

[alippai]

I can try to narrow down the test case later next week, but until that maybe this bugreport helps you to find the bug in the latest release.

[kzc]

Duplicate of #3245.

Curious that it's a different utf8 encoder.

[alippai]

It's our internal lib, definitely not protobufjs, but it seems to be similar :)) Nice coincidence (maybe they origin from the same C++/wiki/other pseudo code)