Is all well with org.eclipse.jdt.internal.compiler.ast.SwitchStatement.statementGenerateCode(BlockScope, CodeStream, Statement) ?

[srikanth-sankaran]

(1) This method swaps the Scope involved at the call site and changes it to this.scope - I am not sure if this intentional or inadvertent.

If I eliminate call to here from org.eclipse.jdt.internal.compiler.ast.SwitchStatement.generateCodeForStringSwitch(BlockScope, CodeStream)

i.e change this line

statementGenerateCode(currentScope, codeStream, statement);

into

statement.generateCode(currentScope, codeStream);

we have a failure in org.eclipse.jdt.core.tests.compiler.regression.InstanceofPrimaryPatternTest.test577415_7()

I am wondering if this change in scope is somehow masking a problem elsewhere.

Does anyone understand what exactly this method does ? Why won't we call statement.generateCode(currentScope, codeStream)

BTW, its override in SwitchExpressions was causing #3276 and was removed by #3272

(2) The diagnostic call

this.scope.problemReporter().operandStackSizeInappropriate(this) in org.eclipse.jdt.internal.compiler.ast.AbstractMethodDeclaration.generateCode(ClassFile) has been commented out. Why ??
This sanity check has found so many problems so far - I don't know of a case where it issues an inappropriate message.

If it indeed does, then let us discuss the scenario.

[srikanth-sankaran]

@mpalat @stephan-herrmann - any insights are welcome

[srikanth-sankaran]

BTW, this may not be tightly coupled with patterns.

This program which doesn't use patterns also triggers a compiler assert:

		public class X {
			static void x() {}

			public static void main(String[] args) {

				switch (args[0]) {
				case "Hello":
					Object o = new Object();
					if (o != null) {
						x();
					} else if (o != null)
						x();
				}
				System.out.println("All well");
			}
		}

[srikanth-sankaran]

Does anyone understand what exactly this method does ? Why won't we call statement.generateCode(currentScope, codeStream)

OK. I have understood the problem, the switch in scope inside the method compensates for the call site not being corrected to use this.scope. See that the switch block constituent statements are resolved in and analyzed in this.scope not currentScope so code generation should also use this.scope. I have corrected this for 4.35 here: #3279

[srikanth-sankaran]

The diagnostic call

this.scope.problemReporter().operandStackSizeInappropriate(this) in org.eclipse.jdt.internal.compiler.ast.AbstractMethodDeclaration.generateCode(ClassFile) has been commented out. Why ?? This sanity check has found so many problems so far - I don't know of a case where it issues an inappropriate message.

If it indeed does, then let us discuss the scenario.

Manoj confirmed in another thread that this was an inadvertent slip up. Restoring here: #3280