[23] ECJ accepts both boolean values and a default case in a switch

[jarthana]

Consider this code:

public class X {
	public void foo(Boolean b) {
		final boolean TRUE = true;
		final boolean FALSE = false;
		switch (b) {
			case TRUE -> { break;}
			case FALSE -> { break;}
			default -> { break;}
		}
	}
}

ECJ accepts this but Javac rejects this with error:
X.java:8: error: switch has both boolean values and a default label

I haven't checked with the spec on this, though.

[stephan-herrmann]

Thanks, @jarthana this is a good one. Keep'em coming! :)

[jarthana]

FWIW, when I remove any of the three case statements, Javac crashes. I have a raised a defect with Oracle for this.

[stephan-herrmann]

Fixed by #2919

[stephan-herrmann]

FWIW, when I remove any of the three case statements, Javac crashes. I have a raised a defect with Oracle for this.

This one: https://bugs.openjdk.org/browse/JDK-8339602 ? Wow that was fast.

Turned out my tests tip-toed narrowly around that case, otherwise run.javac mode would bark.

Should I add that case to our suite anyway?

[jarthana]

FWIW, when I remove any of the three case statements, Javac crashes. I have a raised a defect with Oracle for this.

This one: https://bugs.openjdk.org/browse/JDK-8339602 ? Wow that was fast.

The JDK issue has been closed as "not an issue". Don't know what to make of it :)

[srikanth-sankaran]

FWIW, when I remove any of the three case statements, Javac crashes. I have a raised a defect with Oracle for this.

This one: https://bugs.openjdk.org/browse/JDK-8339602 ? Wow that was fast.

The JDK issue has been closed as "not an issue". Don't know what to make of it :)

Not an issue could mean - it got fixed otherwise somehow and is not reproducible on their tip. It would be good to check their recent build. I don't think they would close a compiler crash otherwise.