PHP json_decode throws error with well-formed json

[agaripian]

This might be a PHP json_decode issue but wanted to share our production bug here after upgrading to Node 12.

var a = '𝘥𝘦𝘴𝘪𝘨𝘯𝘦𝘳 𝘢𝘯𝘥 𝘪𝘭𝘭𝘶𝘴𝘵𝘳𝘢𝘵𝘰𝘳';
JSON.stringify(a.slice(0, 15));

// Node 10 output: 
'"𝘥𝘦𝘴𝘪𝘨𝘯𝘦�"';

// Node 12 output:
'"𝘥𝘦𝘴𝘪𝘨𝘯𝘦\\ud835"'

This response is then sent to a PHP server as JSON and decoded. Which is where the error occurs. Node10's output used to work fine with PHP json_decode but it no longer works with Node12's output.

I simplified the NODE->PHP example see below.

<?php
$string = '{"string": "𝘥𝘦𝘴𝘪𝘨𝘯𝘦\\ud835"}';
var_dump(json_decode($string, false, 512, JSON_THROW_ON_ERROR | JSON_INVALID_UTF8_IGNORE | JSON_INVALID_UTF8_SUBSTITUTE));

// Output:
Fatal error: Uncaught JsonException: Single unpaired UTF-16 surrogate in unicode escape in phptest.php:36
Stack trace:
#0 phptest.php(36): json_decode('{"string": "\xF0\x9D\x98...', false, 512, 7340032)
#1 {main}
  thrown in phptest.php on line 36


//Node 10's output works fine.
object(stdClass)#1 (1) {
  ["string"]=>
  string(31) "𝘥𝘦𝘴𝘪𝘨𝘯𝘦�"
}

[bakkot]

See #13.

As in that thread,

It seems weird to guard against lone surrogates in escaped form but not in their unescaped form.

Those two representations should be equivalent. You might consider filing a bug against PHP.