Weird error reporting: extraneous input '<EOF>'

We found a weird case where the default error recovery strategy reports the EOF token as being an extraneous token.

Grammar:

grammar Test;

program
    : declarations? expression EOF;

declarations
    : declaration+;

declaration
    : 'let' declarator (',' declarator)* ';';

declarator
    : ID '=' expression ;

expression
    : ID
    | NUMBER
    | BOOLEAN
;

NUMBER: [0-9]+;
BOOLEAN: 'true' | 'false';
ID: [a-zA-Z_][a-zA-Z_0-9]*;
WS: (' ' | '\n' | '\r' | '\t')+ -> channel(HIDDEN);

Scenario:

public class Test {
    public static void main(String[] args) throws Exception {
        CharStream input = CharStreams.fromString("let a = 1, b = ;\ntrue");
        TestLexer lex = new TestLexer(input);
        CommonTokenStream tokens = new CommonTokenStream(lex);
        TestParser parser = new TestParser(tokens);
        parser.program();
    }
}

Console output:

line 1:15 extraneous input ';' expecting {NUMBER, BOOLEAN, ID}
line 2:4 extraneous input '<EOF>' expecting {',', ';'}

I tracked down the method of the error recovery strategy that reports the EOF token as being extraneous. I believe the correct here would be to report a token mismatch rather than an extraneous token.

I’d be glad to provide a PR if welcome, but I need some guidance on how to proceed.