The error in the title is thrown only in Google Chrome, according to my tests. I'm base64 encoding a big XML file so that it can be downloaded:

this.loader.src = "data:application/x-forcedownload;base64,"+
                  btoa("<?xml version=\"1.0\" encoding=\"utf-8\"?>"
                  +"<"+this.gamesave.tagName+">"
                  +this.xml.firstChild.innerHTML
                  +"</"+this.gamesave.tagName+">");

this.loader is hidden iframe.

This error is actually quite a change because normally, Google Chrome would crash upon btoa call. Mozilla Firefox has no problems here, so the issue is browser related. I'm not aware of any strange characters in file. Actually I do believe there are no non-ascii characters.

Q: How do I find the problematic characters and replace them so that Chrome stops complaining?

_{I have tried to use Downloadify to initiate the download, but it does not work. It's unreliable and throws no errors to allow debug.}

• javascript
• google-chrome

If you have UTF8, use this (actually works with SVG source), like:

btoa(unescape(encodeURIComponent(str)))

example:

 var imgsrc = 'data:image/svg+xml;base64,' + btoa(unescape(encodeURIComponent(markup)));
 var img = new Image(1, 1); // width, height values are optional params 
 img.src = imgsrc;

If you need to decode that base64, use this:

var str2 = decodeURIComponent(escape(window.atob(b64)));
console.log(str2);

Example:

var str = "äöüÄÖÜçéèñ";
var b64 = window.btoa(unescape(encodeURIComponent(str)))
console.log(b64);

var str2 = decodeURIComponent(escape(window.atob(b64)));
console.log(str2);

Note: if you need to get this to work in mobile-safari, you might need to strip all the white-space from the base64 data...

function b64_to_utf8( str ) {
    str = str.replace(/\s/g, '');    
    return decodeURIComponent(escape(window.atob( str )));
}

2017 Update

This problem has been bugging me again.
The simple truth is, atob doesn't really handle UTF8-strings - it's ASCII only.
Also, I wouldn't use bloatware like js-base64.
But webtoolkit does have a small, nice and very maintainable implementation:

/**
*
*  Base64 encode / decode
*  http://www.webtoolkit.info
*
**/
var Base64 = {

    // private property
    _keyStr: "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/="

    // public method for encoding
    , encode: function (input)
    {
        var output = "";
        var chr1, chr2, chr3, enc1, enc2, enc3, enc4;
        var i = 0;

        input = Base64._utf8_encode(input);

        while (i < input.length)
        {
            chr1 = input.charCodeAt(i++);
            chr2 = input.charCodeAt(i++);
            chr3 = input.charCodeAt(i++);

            enc1 = chr1 >> 2;
            enc2 = ((chr1 & 3) << 4) | (chr2 >> 4);
            enc3 = ((chr2 & 15) << 2) | (chr3 >> 6);
            enc4 = chr3 & 63;

            if (isNaN(chr2))
            {
                enc3 = enc4 = 64;
            }
            else if (isNaN(chr3))
            {
                enc4 = 64;
            }

            output = output +
                this._keyStr.charAt(enc1) + this._keyStr.charAt(enc2) +
                this._keyStr.charAt(enc3) + this._keyStr.charAt(enc4);
        } // Whend 

        return output;
    } // End Function encode 


    // public method for decoding
    ,decode: function (input)
    {
        var output = "";
        var chr1, chr2, chr3;
        var enc1, enc2, enc3, enc4;
        var i = 0;

        input = input.replace(/[^A-Za-z0-9\+\/\=]/g, "");
        while (i < input.length)
        {
            enc1 = this._keyStr.indexOf(input.charAt(i++));
            enc2 = this._keyStr.indexOf(input.charAt(i++));
            enc3 = this._keyStr.indexOf(input.charAt(i++));
            enc4 = this._keyStr.indexOf(input.charAt(i++));

            chr1 = (enc1 << 2) | (enc2 >> 4);
            chr2 = ((enc2 & 15) << 4) | (enc3 >> 2);
            chr3 = ((enc3 & 3) << 6) | enc4;

            output = output + String.fromCharCode(chr1);

            if (enc3 != 64)
            {
                output = output + String.fromCharCode(chr2);
            }

            if (enc4 != 64)
            {
                output = output + String.fromCharCode(chr3);
            }

        } // Whend 

        output = Base64._utf8_decode(output);

        return output;
    } // End Function decode 


    // private method for UTF-8 encoding
    ,_utf8_encode: function (string)
    {
        var utftext = "";
        string = string.replace(/\r\n/g, "\n");

        for (var n = 0; n < string.length; n++)
        {
            var c = string.charCodeAt(n);

            if (c < 128)
            {
                utftext += String.fromCharCode(c);
            }
            else if ((c > 127) && (c < 2048))
            {
                utftext += String.fromCharCode((c >> 6) | 192);
                utftext += String.fromCharCode((c & 63) | 128);
            }
            else
            {
                utftext += String.fromCharCode((c >> 12) | 224);
                utftext += String.fromCharCode(((c >> 6) & 63) | 128);
                utftext += String.fromCharCode((c & 63) | 128);
            }

        } // Next n 

        return utftext;
    } // End Function _utf8_encode 

    // private method for UTF-8 decoding
    ,_utf8_decode: function (utftext)
    {
        var string = "";
        var i = 0;
        var c, c1, c2, c3;
        c = c1 = c2 = 0;

        while (i < utftext.length)
        {
            c = utftext.charCodeAt(i);

            if (c < 128)
            {
                string += String.fromCharCode(c);
                i++;
            }
            else if ((c > 191) && (c < 224))
            {
                c2 = utftext.charCodeAt(i + 1);
                string += String.fromCharCode(((c & 31) << 6) | (c2 & 63));
                i += 2;
            }
            else
            {
                c2 = utftext.charCodeAt(i + 1);
                c3 = utftext.charCodeAt(i + 2);
                string += String.fromCharCode(((c & 15) << 12) | ((c2 & 63) << 6) | (c3 & 63));
                i += 3;
            }

        } // Whend 

        return string;
    } // End Function _utf8_decode 

}

Using btoa with unescape and encodeURIComponent didn't work for me. Replacing all the special caracteres by XML/HTML entities and then converting to the base64 representation was the only way to solve this issue for me. Some code:

base64 = btoa(str.replace(/[\u00A0-\u2666]/g, function(c) {
    return '&#' + c.charCodeAt(0) + ';';
}));

I just thought I should share how I actually solved the problem and why I think this is the right solution (provided you don't optimize for old browser).

Converting data to dataURL (data: ...)

var blob = new Blob(
              // I'm using page innerHTML as data
              // note that you can use the array
              // to concatenate many long strings EFFICIENTLY
              [document.body.innerHTML],
              // Mime type is important for data url
              {type : 'text/html'}
); 
// This FileReader works asynchronously, so it doesn't lag
// the web application
var a = new FileReader();
a.onload = function(e) {
     // Capture result here
     console.log(e.target.result);
};
a.readAsDataURL(blob);

Allowing user to save data

Apart from obvious solution - opening new window with your dataURL as URL you can do two other things.

1. Use fileSaver.js

File saver can create actual fileSave dialog with predefined filename. It can also fallback to normal dataURL approach.

2. Use (experimental) URL.createObjectURL

This is great for reusing base64 encoded data. It creates a short URL for your dataURL:

console.log(URL.createObjectURL(blob));
//Prints: blob:http://stackoverflow.com/7c18953f-f5f8-41d2-abf5-e9cbced9bc42

Don't forget to use the URL including the leading blob prefix. I used document.body again:

You can use this short URL as AJAX target, <script> source or <a> href location. You're responsible for destroying the URL though:

URL.revokeObjectURL('blob:http://stackoverflow.com/7c18953f-f5f8-41d2-abf5-e9cbced9bc42')

Use a library instead

We don't have to reinvent the wheel. Just use a library to save the time and headache.

js-base64

https://github.com/dankogai/js-base64 is good and I confirm it supports unicode very well.

Base64.encode('dankogai');  // ZGFua29nYWk=
Base64.encode('小飼弾');    // 5bCP6aO85by+
Base64.encodeURI('小飼弾'); // 5bCP6aO85by-

Base64.decode('ZGFua29nYWk=');  // dankogai
Base64.decode('5bCP6aO85by+');  // 小飼弾
// note .decodeURI() is unnecessary since it accepts both flavors
Base64.decode('5bCP6aO85by-');  // 小飼弾

I just ran into this problem myself.

First, modify your code slightly:

var download = "<?xml version=\"1.0\" encoding=\"utf-8\"?>"
                  +"<"+this.gamesave.tagName+">"
                  +this.xml.firstChild.innerHTML
                  +"</"+this.gamesave.tagName+">";

this.loader.src = "data:application/x-forcedownload;base64,"+
                  btoa(download);

Then use your favorite web inspector, put a breakpoint on the line of code that assigns this.loader.src, then execute this code:

for (var i = 0; i < download.length; i++) {
  if (download[i].charCodeAt(0) > 255) {
    console.warn('found character ' + download[i].charCodeAt(0) + ' "' + download[i] + '" at position ' + i);
  }
}

Depending on your application, replacing the characters that are out of range may or may not work, since you'll be modifying the data. See the note on MDN about unicode characters with the btoa method:

https://developer.mozilla.org/en-US/docs/Web/API/window.btoa

As an complement to Stefan Steiger answer: (as it doesn't look nice as a comment)

Extending String prototype:

String.prototype.b64encode = function() { 
    return btoa(unescape(encodeURIComponent(this))); 
};
String.prototype.b64decode = function() { 
    return decodeURIComponent(escape(atob(this))); 
};

Usage:

var str = "äöüÄÖÜçéèñ";
var encoded = str.b64encode();
console.log( encoded.b64decode() );

NOTE:

As stated in the comments, using unescape is not recommended as it may be removed in the future:

Warning: Although unescape() is not strictly deprecated (as in "removed from the Web standards"), it is defined in Annex B of the ECMA-262 standard, whose introduction states: … All of the language features and behaviours specified in this annex have one or more undesirable characteristics and in the absence of legacy usage would be removed from this specification.

Note: Do not use unescape to decode URIs, use decodeURI or decodeURIComponent instead.

btoa() only support characters from String.fromCodePoint(0) up to String.fromCodePoint(255). For Base64 characters with a code point 256 or higher you need to encode/decode these before and after.

And in this point it becomes tricky...

Every possible sign are arranged in a Unicode-Table. The Unicode-Table is divided in different planes (languages, math symbols, and so on...). Every sign in a plane has a unique code point number. Theoretically, the number can become arbitrarily large.

A computer stores the data in bytes (8 bit, hexadecimal 0x00 - 0xff, binary 00000000 - 11111111, decimal 0 - 255). This range normally use to save basic characters (Latin1 range).

For characters with higher codepoint then 255 exist different encodings. JavaScript use 16 bits per sign (UTF-16), the string called DOMString. Unicode can handle code points up to 0x10fffff. That means, that a method must be exist to store several bits over several cells away.

String.fromCodePoint(0x10000).length == 2

UTF-16 use surrogate pairs to store 20bits in two 16bit cells. The first higher surrogate begins with 110110xxxxxxxxxx, the lower second one with 110111xxxxxxxxxx. Unicode reserved own planes for this: https://unicode-table.com/de/#high-surrogates

To store characters in bytes (Latin1 range) standardized procedures use UTF-8.

Sorry to say that, but I think there is no other way to implement this function self.

function stringToUTF8(str)
{
    let bytes = [];

    for(let character of str)
    {
        let code = character.codePointAt(0);

        if(code <= 127)
        {
            let byte1 = code;

            bytes.push(byte1);
        }
        else if(code <= 2047)
        {
            let byte1 = 0xC0 | (code >> 6);
            let byte2 = 0x80 | (code & 0x3F);

            bytes.push(byte1, byte2);
        }
        else if(code <= 65535)
        {
            let byte1 = 0xE0 | (code >> 12);
            let byte2 = 0x80 | ((code >> 6) & 0x3F);
            let byte3 = 0x80 | (code & 0x3F);

            bytes.push(byte1, byte2, byte3);
        }
        else if(code <= 2097151)
        {
            let byte1 = 0xF0 | (code >> 18);
            let byte2 = 0x80 | ((code >> 12) & 0x3F);
            let byte3 = 0x80 | ((code >> 6) & 0x3F);
            let byte4 = 0x80 | (code & 0x3F);

            bytes.push(byte1, byte2, byte3, byte4);
        }
    }

    return bytes;
}

function utf8ToString(bytes, fallback)
{
    let valid = undefined;
    let codePoint = undefined;
    let codeBlocks = [0, 0, 0, 0];

    let result = "";

    for(let offset = 0; offset < bytes.length; offset++)
    {
        let byte = bytes[offset];

        if((byte & 0x80) == 0x00)
        {
            codeBlocks[0] = byte & 0x7F;

            codePoint = codeBlocks[0];
        }
        else if((byte & 0xE0) == 0xC0)
        {
            codeBlocks[0] = byte & 0x1F;

            byte = bytes[++offset];
            if(offset >= bytes.length || (byte & 0xC0) != 0x80) { valid = false; break; }

            codeBlocks[1] = byte & 0x3F;

            codePoint = (codeBlocks[0] << 6) + codeBlocks[1];
        }
        else if((byte & 0xF0) == 0xE0)
        {
            codeBlocks[0] = byte & 0xF;

            for(let blockIndex = 1; blockIndex <= 2; blockIndex++)
            {
                byte = bytes[++offset];
                if(offset >= bytes.length || (byte & 0xC0) != 0x80) { valid = false; break; }

                codeBlocks[blockIndex] = byte & 0x3F;
            }
            if(valid === false) { break; }

            codePoint = (codeBlocks[0] << 12) + (codeBlocks[1] << 6) + codeBlocks[2];
        }
        else if((byte & 0xF8) == 0xF0)
        {
            codeBlocks[0] = byte & 0x7;

            for(let blockIndex = 1; blockIndex <= 3; blockIndex++)
            {
                byte = bytes[++offset];
                if(offset >= bytes.length || (byte & 0xC0) != 0x80) { valid = false; break; }

                codeBlocks[blockIndex] = byte & 0x3F;
            }
            if(valid === false) { break; }

            codePoint = (codeBlocks[0] << 18) + (codeBlocks[1] << 12) + (codeBlocks[2] << 6) + (codeBlocks[3]);
        }
        else
        {
            valid = false; break;
        }

        result += String.fromCodePoint(codePoint);
    }

    if(valid === false)
    {
        if(!fallback)
        {
            throw new TypeError("Malformed utf-8 encoding.");
        }

        result = "";

        for(let offset = 0; offset != bytes.length; offset++)
        {
            result += String.fromCharCode(bytes[offset] & 0xFF);
        }
    }

    return result;
}

function decodeBase64(text, binary)
{
    if(/[^0-9a-zA-Z\+\/\=]/.test(text)) { throw new TypeError("The string to be decoded contains characters outside of the valid base64 range."); }

    let codePointA = 'A'.codePointAt(0);
    let codePointZ = 'Z'.codePointAt(0);
    let codePointa = 'a'.codePointAt(0);
    let codePointz = 'z'.codePointAt(0);
    let codePointZero = '0'.codePointAt(0);
    let codePointNine = '9'.codePointAt(0);
    let codePointPlus = '+'.codePointAt(0);
    let codePointSlash = '/'.codePointAt(0);

    function getCodeFromKey(key)
    {
        let keyCode = key.codePointAt(0);

        if(keyCode >= codePointA && keyCode <= codePointZ)
        {
            return keyCode - codePointA;
        }
        else if(keyCode >= codePointa && keyCode <= codePointz)
        {
            return keyCode + 26 - codePointa;
        }
        else if(keyCode >= codePointZero && keyCode <= codePointNine)
        {
            return keyCode + 52 - codePointZero;
        }
        else if(keyCode == codePointPlus)
        {
            return 62;
        }
        else if(keyCode == codePointSlash)
        {
            return 63;
        }

        return undefined;
    }

    let codes = Array.from(text).map(character => getCodeFromKey(character));

    let bytesLength = Math.ceil(codes.length / 4) * 3;

    if(codes[codes.length - 2] == undefined) { bytesLength = bytesLength - 2; } else if(codes[codes.length - 1] == undefined) { bytesLength--; }

    let bytes = new Uint8Array(bytesLength);

    for(let offset = 0, index = 0; offset < bytes.length;)
    {
        let code1 = codes[index++];
        let code2 = codes[index++];
        let code3 = codes[index++];
        let code4 = codes[index++];

        let byte1 = (code1 << 2) | (code2 >> 4);
        let byte2 = ((code2 & 0xf) << 4) | (code3 >> 2);
        let byte3 = ((code3 & 0x3) << 6) | code4;

        bytes[offset++] = byte1;
        bytes[offset++] = byte2;
        bytes[offset++] = byte3;
    }

    if(binary) { return bytes; }

    return utf8ToString(bytes, true);
}

function encodeBase64(bytes) {
    if (bytes === undefined || bytes === null) {
        return '';
    }
    if (bytes instanceof Array) {
        bytes = bytes.filter(item => {
            return Number.isFinite(item) && item >= 0 && item <= 255;
        });
    }

    if (
        !(
            bytes instanceof Uint8Array ||
            bytes instanceof Uint8ClampedArray ||
            bytes instanceof Array
        )
    ) {
        if (typeof bytes === 'string') {
            const str = bytes;
            bytes = Array.from(unescape(encodeURIComponent(str))).map(ch =>
                ch.codePointAt(0)
            );
        } else {
            throw new TypeError('bytes must be of type Uint8Array or String.');
        }
    }

    const keys = [
        'A',
        'B',
        'C',
        'D',
        'E',
        'F',
        'G',
        'H',
        'I',
        'J',
        'K',
        'L',
        'M',
        'N',
        'O',
        'P',
        'Q',
        'R',
        'S',
        'T',
        'U',
        'V',
        'W',
        'X',
        'Y',
        'Z',
        'a',
        'b',
        'c',
        'd',
        'e',
        'f',
        'g',
        'h',
        'i',
        'j',
        'k',
        'l',
        'm',
        'n',
        'o',
        'p',
        'q',
        'r',
        's',
        't',
        'u',
        'v',
        'w',
        'x',
        'y',
        'z',
        '0',
        '1',
        '2',
        '3',
        '4',
        '5',
        '6',
        '7',
        '8',
        '9',
        '+',
        '/'
    ];
    const fillKey = '=';

    let byte1;
    let byte2;
    let byte3;
    let sign1 = ' ';
    let sign2 = ' ';
    let sign3 = ' ';
    let sign4 = ' ';

    let result = '';

    for (let index = 0; index < bytes.length; ) {
        let fillUpAt = 0;

        // tslint:disable:no-increment-decrement
        byte1 = bytes[index++];
        byte2 = bytes[index++];
        byte3 = bytes[index++];

        if (byte2 === undefined) {
            byte2 = 0;
            fillUpAt = 2;
        }

        if (byte3 === undefined) {
            byte3 = 0;
            if (!fillUpAt) {
                fillUpAt = 3;
            }
        }

        // tslint:disable:no-bitwise
        sign1 = keys[byte1 >> 2];
        sign2 = keys[((byte1 & 0x3) << 4) + (byte2 >> 4)];
        sign3 = keys[((byte2 & 0xf) << 2) + (byte3 >> 6)];
        sign4 = keys[byte3 & 0x3f];

        if (fillUpAt > 0) {
            if (fillUpAt <= 2) {
                sign3 = fillKey;
            }
            if (fillUpAt <= 3) {
                sign4 = fillKey;
            }
        }

        result += sign1 + sign2 + sign3 + sign4;

        if (fillUpAt) {
            break;
        }
    }

    return result;
}

let base64 = encodeBase64("\u{1F604}"); // unicode code point escapes for smiley
let str = decodeBase64(base64);

console.log("base64", base64);
console.log("str", str);

document.body.innerText = str;

how to use it: decodeBase64(encodeBase64("\u{1F604}"))

demo: https://jsfiddle.net/qrLadeb8/