Custom variables don't work for non-English symbols

Thanks for report. I guess this may be caused by the url encoding. What does google analytics statistics interface show with German Sonderzeichen or rusdion chars?

The \u[code] doesn't mean it's wrong. This are just the javascript unicode chars. Are the chars wrong on the ga statistics and if they are, how? Can you debug the code, please?

Normally the modules checks if the value or key is empty. If the token value is empty the custom var is not added to the content. This is the way how it's designed by google and the module. Nothing wrong.

If this is not the way how it works it's a bug.

Per docs the "key" was named "name" in past... I have just followed the google docs. We need to check if this has changed in wording and than we can change to key, too.

This cannot be a correct change and makes no sense to me, too. Google does not allow the string to be longer than 128 chars after url encoding.

Is there any bug in rawurldecode() with Russian letters?

This sounds like an important bug. The idea behind was that we do not log a broken/half hex key of a letter. I think it should just trim the string to the closest hex value if there is one, not remove all. We need to write a test for this, too.

This does not work, too. If the last characters are a broken/half unicode hex value (%C3) the string becomes null by drupal_json_encode().

I have a fix ready based on Drupal API, but just in case we may need any other solution later I'm posting a regex that seems working, too. Found this at php - Remove non-utf8 characters from string - Stack Overflow + two other workarounds that work with iconv() with //IGNORE and mb_convert_encoding() like drupal_convert_to_utf8() does except the //IGNORE in iconv() that seem to be superfluous if error reporting is suppressed with @iconv().

  $custom_var_value = preg_replace('/
    (
      (?: [\x00-\x7F]                 # single-byte sequences   0xxxxxxx
      |   [\xC0-\xDF][\x80-\xBF]      # double-byte sequences   110xxxxx 10xxxxxx
      |   [\xE0-\xEF][\x80-\xBF]{2}   # triple-byte sequences   1110xxxx 10xxxxxx * 2
      |   [\xF0-\xF7][\x80-\xBF]{3}   # quadruple-byte sequence 11110xxx 10xxxxxx * 3
      )+                              # ...one or more times
    )
  | .                                 # anything else
  /x', '$1', $tmp_value);

Heck, the more test cases I'm trying the more issues I'm finding. Any suggestions or ideas?

        if ($name_length + $value_length > 128) {
          // Trim name and value to maximum combined length.
          $tmp_value = substr($tmp_value, 0, 127 - $name_length);
          $tmp_value = urldecode($tmp_value);
          // Silently remove non-utf8 characters created by trim of encoded url.
          $tmp_value = drupal_convert_to_utf8($tmp_value, 'utf-8');

          // FAIL: but we need something similar or we will see e.g. "%C" on end of strings
          // FAIL: Destroys strings that end with UPPER CASE chars.
          $tmp_value = rtrim($tmp_value, '%0..9A..F');

          // DOES NOT WORK PROPERLY IN ALL CASES e.g. "%C"
          $tmp_value = rtrim($tmp_value, '%');
          $tmp_value = rtrim($tmp_value);
          $custom_var_value = $tmp_value;
        }

Additional to this my Eclipse Egit or Git itself does not handle the Russian chars properly. After I commit this as a patch it looks destroyed.

    // Test whether invalid utf-8 tests may fail.
    $custom_vars = array(
      'slots' => array(
        1 => array(
          'slot' => 1,
          'name' => 'UTF8 test abc',
          'value' => 'Роли пользователей äöüß',
          'scope' => 3,
        ),
        2 => array(
          'slot' => 2,
          'name' => 'UTF8 test abcd',
          'value' => 'Роли пользователей äöüß',
          'scope' => 3,
        ),
        3 => array(
          'slot' => 3,
          'name' => 'Test',
          'value' => 'Роли пользователей äöüß',
          'scope' => 3,
        ),
        4 => array(
          'slot' => 4,
          'name' => 'Test',
          'value' => $this->randomName(255),
          'scope' => 3,
        ),
        5 => array(
          'slot' => 5,
          'name' => '',
          'value' => '',
          'scope' => 3,
        ),
      )
    );
    variable_set('googleanalytics_custom_var', $custom_vars);
    $this->verbose('<pre>' . print_r($custom_vars, TRUE) . '</pre>');

    $this->drupalGet('');

    // 'UTF8 test abc' + 'Роли пользователей äöüß' = 'Роли пользователе%C' = 'Роли пользователе'
    // 1. Trim HEX values
    // 2. Trim trailing spaces
    $this->assertRaw("_gaq.push(['_setCustomVar', 1, \"UTF8 test abc\", \"Роли пользователе\", 3]", '[testGoogleAnalyticsCustomVariables]: String trim 1 worked properly.');

    // 'UTF8 test abcd' + 'Роли пользователей äöüß' = 'Роли пользователей %' -> 'Роли пользователей'
    // 1. Trim %
    // 2. Trim spaces
    $this->assertRaw("_gaq.push(['_setCustomVar', 2, \"UTF8 test abcd\", \"Роли пользователей\", 3]", '[testGoogleAnalyticsCustomVariables]: String trim 2 worked properly.');

    // 'Test' + 'Роли пользователей äöüß' = 'Роли пользователей äö�' -> 'Роли пользователей äö'
    // 1. Remove invalid/broken utf-8 character e.g. %3C
    $this->assertRaw("_gaq.push(['_setCustomVar', 3, \"Test\", \"Роли пользователей äö\", 3]", '[testGoogleAnalyticsCustomVariables]: String trim 3 worked properly.');

    // 'Test' + Random string with 255 chars = 4 + 124 chars.
    // 1. Trim combinded name and value string length to max 128 characters.
    $this->assertRaw("_gaq.push(['_setCustomVar', 4, \"Test\", \"" . substr($site_slogan, 0, 123) . "\", 3]", '[testGoogleAnalyticsCustomVariables]: String trim 4 worked properly.');

Maybe we are able to build a regex for this use case and run preg_replace() to cut off the string at 128 chars or earlier.

Fixed last code.

I'm closing this now as the code has been removed in 2.x. I'm not sure if we are allowed to send unlimited stings in UA, but it's at least not documented that there is a limit like in ga.js.