ElidedSemicolonAndRightBrace in compile error message

[trancexpress]

Found during experiments for #859.

To reproduce, compile snippet:

public class X {
    public interface S<T> {
        T s(T t);
    }
    public void f() {
    	S<?> s = v -> v ) ;
    }
}

Observe compile error from ecj (4.27):

----------
1. ERROR in .../X.java (at line 6)
        S<?> s = v -> v ) ;
                        ^
Syntax error on token ")", ElidedSemicolonAndRightBrace expected
----------
1 problem (1 error)

Compile error from javac (OpenJDK 17):

X.java:6: error: ';' expected
        S<?> s = v -> v ) ;
                       ^
1 error

This seems to not be expected (unsurprisingly), see ProblemReporter.replaceIfSynthetic(String):

private String replaceIfSynthetic(String token) {
	/* Java 8 grammar changes use some synthetic tokens to make the grammar LALR(1). These tokens should not be exposed in messages
	   as it would make no sense to the programmer whatsoever. Replace such artificial tokens with some "suitable"  alternative. At
	   the moment, there are two synthetic tokens that need such massaging viz : "BeginLambda" and "BeginTypeArguments". There is a
	   third synthetic token "ElidedSemicolonAndRightBrace" that we don't expect to show up in messages since it is manufactured by
	   the parser automatically.
	*/

[mpalat]

@trancexpress Am planning to work with a couple of newcomers (already identified) for this - hence self-assigned for now.

[trancexpress]

@trancexpress Am planning to work with a couple of newcomers (already identified) for this - hence self-assigned for now.

Sounds great! If possible, can you look into #859, #367, and #125 as well? They seems to be very related to this bug. Possibly all can be fixed with the same change. See also the comment from Stephan here: #125 (comment)

There seems to be some complicated workaround for lambdas, that is maybe not necessary anymore. To my understanding, this workaround is the cause for the unexpected error message here.

[srikanth-sankaran]

I don't think this qualifies as a good-first-issue. Nothing to do with DiagnoseParse qualifies as a "good" issue or a "first" issue let alone a "good first" issue (guffaw)

[srikanth-sankaran]

You may want to look at the decidedly low-tech solution adopted for #367

[mpalat]

I don't think this qualifies as a good-first-issue. Nothing to do with DiagnoseParse qualifies as a "good" issue or a "first" issue let alone a "good first" issue (guffaw)

point taken - removed the assignment for now.