[GROOVY-10156] Unreachable bytecode in switch statement

• Type: Improvement
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 3.0.8
• Component/s: Compiler
• Labels: bytecode enum
• Environment: JVM: 11.0.11 (Oracle Corporation 11.0.11+9-LTS-194) ; OS: Mac OS X 10.16 x86_64
• Assignee: Unassigned
• Reporter: Damir Murat
• Created: 29/Jun/21 12:20
• Updated: 04/Nov/22 13:36

Description

When switching over an enum, and for two or more case branches, groovy generates unreachable bytecode for each case branch except the last one.

For example

package mypackage
import groovy.transform.CompileStatic

@CompileStatic
class SwitchDemo {
  void mySwitchMethod(MyEnum myEnum) {
    switch (myEnum) {
      case MyEnum.ONE:
        println "ONE"
        break
      case MyEnum.TWO:
        println "TWO"
        break
      case MyEnum.THREE:
        println "TWO"
        break
      default:
        println "default"
        break
    }
  }
}

@CompileStatic
enum MyEnum {
  ONE, TWO, THREE
}

generates bytecode sequences like

...
   L8
   L9
    LINENUMBER 12 L9
    GOTO L10
   L3
   L11
    LINENUMBER 12 L11
    ALOAD 2
    GETSTATIC mypackage/MyEnum.TWO : Lmypackage/MyEnum;
    INVOKESTATIC org/codehaus/groovy/runtime/ScriptBytecodeAdapter.isCase (Ljava/lang/Object;Ljava/lang/Object;)Z
    IFEQ L12
   L10
...

It looks like the first four lines above are not necessary. This results in

nop, nop, athrow

sequences in generated java bytecode.

It would be great if this can be cleaned up and fixed. My scenario is related to JaCoCo reports which are not accurate and a bit confusing for switch statements.

Tnx