Summary

javac will now correctly reject programs that contains character literals containing multi-surrogate characters.

Problem

Certain Unicode characters are not representable by a single char, and are represented as two chars (surrogates). In Java, a character literal can only contain one char value, not two. But javac will accept code like:

   char c = '😊';

javac will only use the first char/surrogate, and ignore the second one, leading to incorrect and surprising behavior.

Solution

javac will produce an error when a character literal contains a Unicode character, that cannot be represented by a single char value.

Specification

The new behavior is in line with the existing specification. Specifically, JLS 3.10.4 states:

Character literals can only represent UTF-16 code units (§3.1), i.e., they are limited to values from \u0000 to \uffff.

I.e. the specification does not allow to use a multi-surrogate character in a char literal.