PCRE returns incorrect ranges when matching against strings with emojis

[fwrs]

Steps to reproduce the problem

1. Enter the following regex: /test/g
2. Enter the following test string: test 🥶 test

Expected result

You see each instance of the word test being highlighted, twice in total

Actual result

You see two different captured strings test, and <space>tes (the second capture is shifted by 1 letter to the left)

Screenshot

[Attacktive]

The length of 🥶 in JavaScript is 2 not 1. 🥶 On the other hand, 🥶 seems to be treated as a single character in preg_match and preg_match_all. 🥶

I can't think of an easy way to handle the difference. 🥶