Groovy 2.5.5 unreachable line in byte code

[admxiii]

JaCoCo version: 0.8.4

Starting with groovy 2.5.5 through 2.5.7, an additional line number was added to the byte code to fix some issues. This causes jacoco to report lines that cannot ever be covered. This new line number seems to cause problems anytime it is after a ARETURN statement on the closing brace of the method. It would be great if these unreachable lines could be filtered out somehow.

int nonVoidMethod() {
1 * 1
} <-- shows red

Example of new byte code

ARETURN
L1
LINENUMBER 16 L1 <-- new line number that cant be hit
FRAME FULL [] [java/lang/Throwable]
NOP
ATHROW
LOCALVARIABLE this Lcom/company/Form; L0 L1 0
MAXSTACK = 2
MAXLOCALS = 1

groovy 2.5.4 byte code

ARETURN
L1
FRAME FULL [] [java/lang/Throwable]
NOP
ATHROW
LOCALVARIABLE this Lcom/company/Form; L0 L1 0
MAXSTACK = 2
MAXLOCALS = 1

related to fixed issues
https://issues.apache.org/jira/browse/GROOVY-7647
https://issues.apache.org/jira/browse/GROOVY-8742

[Godin]

Even in Groovy 2.5.4, there are unreachable instructions, which shows up in coverage reports:

Exactly the same instructions shows up with Groovy 2.5.7 just with different line number:

And this was already reported to developers of Groovy compiler some time ago - http://groovy.329449.n5.nabble.com/Groovy-2-5-4-generates-dead-code-td5755188.html
With conclusion that this should be fixed in compiler - see reply from @danielsun1106

So @admxiii please report this to developers of Groovy to remind them about this.

[admxiii]

ok thanks, I will open a ticket with Groovy

[Godin]

@admxiii thank you! 👍 ❤️

[Godin]

@admxiii I'm wondering if you managed to create ticket on Groovy side, so that we can track it too?

@aalmiray FYI if you need something more interesting than just writing tests at Hackergarten 😉

[admxiii]

Here's the ticket I created - https://issues.apache.org/jira/browse/GROOVY-9126

[daniellansun]

@admxiii The issue will be fixed in Groovy 3.0.0-rc-2 and Groovy 2.5.9.

[Godin]

@danielsun1106 that's cool! and thank you for the notification 👍 ❤️

I wanted to test 2.5.9-SNAPSHOT on JaCoCo validation tests, however could not find it in https://oss.jfrog.org/artifactory/oss-snapshot-local/org/codehaus/groovy/ , which is according to https://groovy-lang.org/install.html supposed to contain snapshot builds. Do you know why it is not there?

[daniellansun]

Groovy core team's CI server is down, so no SNAPSHOT is available, but you could clone the project ( https://github.com/apache/groovy.git ) and switch to 2_5_X branch, run gradlew installGroovy to get SNAPSHOT distribution ;-)

[Godin]

you could clone the project ( https://github.com/apache/groovy.git ) and switch to 2_5_X branch, run gradlew installGroovy to get SNAPSHOT distribution ;-)

@danielsun1106 I know, just wanted a bit of laziness during vacation 😆 Will do given that CI server is down. Thank you.

[daniellansun]

Enjoy your vacation :-)

[daniellansun]

Tested groovy code coverage with IntelliJ IDEA. As we can see, no red mark at the tailing of the method:

[Godin]

@danielsun1106 I made a test of 2.5.9-SNAPSHOT built from GROOVY_2_5_X branch at apache/groovy@c37ec7e :

According to this test unreachable instructions do not have line numbers anymore,
however instructions are still there as was reported in http://groovy.329449.n5.nabble.com/Groovy-2-5-4-generates-dead-code-td5755188.html , which is mentioned in #884 (comment) here.

AFAIK IntelliJ computes and shows line coverage, i.e. line is considered as covered when at least one instruction belonging to this line is covered. While JaCoCo also computes instructions coverage and shows lines as partially covered when not all instructions are covered. And so while change in Groovy helps for line coverage, instructions coverage is still affected by presence of dead untestable code.

See example.zip for which execution of mvn verify produces following JaCoCo report

and here is output of javap -v -p target/classes/Example.class for method non_void_method

  public static int non_void_method();
    descriptor: ()I
    flags: ACC_PUBLIC, ACC_STATIC
    Code:
      stack=3, locals=1, args_size=0
         0: invokestatic  #20                 // Method $getCallSiteArray:()[Lorg/codehaus/groovy/runtime/callsite/CallSite;
         3: astore_0
         4: aload_0
         5: ldc           #33                 // int 0
         7: aaload
         8: ldc           #2                  // class Example
        10: ldc           #35                 // String Hello, World!
        12: invokeinterface #41,  3           // InterfaceMethod org/codehaus/groovy/runtime/callsite/CallSite.callStatic:(Ljava/lang/Class;Ljava/lang/Object;)Ljava/lang/Object;
        17: pop
        18: bipush        42
        20: ireturn
        21: nop
        22: nop
        23: athrow
      StackMapTable: number_of_entries = 1
        frame_type = 85 /* same_locals_1_stack_item */
          stack = [ class java/lang/Throwable ]
      LineNumberTable:
        line 3: 4
        line 4: 18

sequence of nops followed by athrow at offsets 21 - 23 can not be reached.

[daniellansun]

@Godin It looks that we have more work to complete ;-)
/cc @paulk-asert