[JENKINS-26481] Mishandling of binary methods accepting Closure

[jenkins-infra-bot]

At least some closures are executed only once inside of Groovy CPS DSL scripts managed by the workflow plugin.

Steps to reproduce:
1. Create a new workflow with the following script:
node {
[1, 2, 3].each

{ println it }

println "abc".replaceAll(/[a-z]/)

{ it.toUpperCase() }

}
2. Build the workflow

Actual output:
Started by user anonymous
Running: Allocate node : Start
Running on master in /var/lib/jenkins/jobs/testflow/workspace
Running: Allocate node : Body : Start
Running: Print Message
1
Running: Print Message
A
Running: Allocate node : Body : End
Running: Allocate node : End
Running: End of Workflow
Finished: SUCCESS

Expected output:
Started by user anonymous
Running: Allocate node : Start
Running on master in /var/lib/jenkins/jobs/testflow/workspace
Running: Allocate node : Body : Start
Running: Print Message
1
Running: Print Message
2
Running: Print Message
3
Running: Print Message
ABC
Running: Allocate node : Body : End
Running: Allocate node : End
Running: End of Workflow
Finished: SUCCESS

---

Originally reported by dtschan, imported from: Mishandling of binary methods accepting Closure

• assignee: jglick
• status: Resolved
• priority: Critical
• component(s): workflow-cps-plugin
• label(s): groovy, kohsuke-plane-project, pipeline-hangs
• resolution: Fixed
• resolved: 2017-05-30T23:06:35+00:00
• votes: 108
• watchers: 137
• imported: 2025-12-07

Raw content of original issue

At least some closures are executed only once inside of Groovy CPS DSL scripts managed by the workflow plugin.

Steps to reproduce:
1. Create a new workflow with the following script:
   node {
     [1, 2, 3].each 
{ println it }

     println "abc".replaceAll(/[a-z]/) 
{ it.toUpperCase() }
   }
2. Build the workflow

Actual output:
Started by user anonymous
Running: Allocate node : Start
Running on master in /var/lib/jenkins/jobs/testflow/workspace
Running: Allocate node : Body : Start
Running: Print Message
1
Running: Print Message
A
Running: Allocate node : Body : End
Running: Allocate node : End
Running: End of Workflow
Finished: SUCCESS

Expected output:
Started by user anonymous
Running: Allocate node : Start
Running on master in /var/lib/jenkins/jobs/testflow/workspace
Running: Allocate node : Body : Start
Running: Print Message
1
Running: Print Message
2
Running: Print Message
3
Running: Print Message
ABC
Running: Allocate node : Body : End
Running: Allocate node : End
Running: End of Workflow
Finished: SUCCESS

• environment: Jenkins LTS 1.580.2 and workflow-plugin 1.1. Both are the latest versions at the time of this writing. Tested on RHEL 6 with OpenJDK 1.7 and Fedora 21 with OpenJDK 1.8.

[jenkins-infra-bot]

• Remote links associated with this issue:

  • groovy-cps #​8
  • groovy-cps #​9
  • groovy-cps #​7
  • PR 372
  • workflow-cps PR 15
  • groovy-cps PR 24
  • In-progress DGM patching PR (groovy-cps #​52)
  • workflow-cps PR 124
  • CloudBees Internal OSS-1410

[jenkins-infra-bot]

jglick:

• Original comment link
• Raw content of original comment:

  Sounds like a bad implementation of something in the groovy-cps library.
  

Sounds like a bad implementation of something in the groovy-cps library.

[jenkins-infra-bot]

jglick:

• Original comment link
• Raw content of original comment:

  JENKINS-26300 sounds similar, perhaps with the same root cause.
  

JENKINS-26300 sounds similar, perhaps with the same root cause.

[jenkins-infra-bot]

jglick:

• Original comment link
• Raw content of original comment:

  groovy-cps #8 covers the particular case reported here: there is an implementation of each for CPS-transformed code, but for reasons still TBD it is not getting used when called on an array (it is used for example on sequences).
  

groovy-cps #8 covers the particular case reported here: there is an implementation of each for CPS-transformed code, but for reasons still TBD it is not getting used when called on an array (it is used for example on sequences).

[jenkins-infra-bot]

jglick:

• Original comment link
• Raw content of original comment:

  groovy-cps #9 would allow closures to run synchronously, permitting routine functional-style programs to be used so long as the closures were not trying to do long-running tasks like calling shell steps.
  

groovy-cps #9 would allow closures to run synchronously, permitting routine functional-style programs to be used so long as the closures were not trying to do long-running tasks like calling shell steps.

[jenkins-infra-bot]

jglick:

• Original comment link
• Raw content of original comment:

  groovy-cps #7 would allow us to offer CPS-transformed versions of all standard Groovy closure-based methods like find without having to reimplement each one by hand.
  

groovy-cps #7 would allow us to offer CPS-transformed versions of all standard Groovy closure-based methods like find without having to reimplement each one by hand.

[jenkins-infra-bot]

rawlingsj:

• Original comment link
• Raw content of original comment:

  I think I've just hit this same problem, using the workflow plugin I'm reading a multiline file but unable to loop over each line, I only get one line printed in this example..
  
  
  def file = readFile('MultiLineFile.txt')
  file.eachLine { line ->
    println line
  }
  
  
  

I think I've just hit this same problem, using the workflow plugin I'm reading a multiline file but unable to loop over each line, I only get one line printed in this example..

def file = readFile('MultiLineFile.txt')
file.eachLine { line ->
  println line
}

[jenkins-infra-bot]

jglick:

• Original comment link
• Raw content of original comment:

  That is the same. You need to split and use a C-style loop.
  

That is the same. You need to split and use a C-style loop.

[jenkins-infra-bot]

rawlingsj:

• Original comment link
• Raw content of original comment:

  Yeah figured it was worth mentioning and thanks for the original comments as it helped me move on.  For completeness and to complement above answer here's what I used..
  
  
  def lines = new String(multiLineFile).split( '\n' )
  for(int i = 0; i < lines.size(); i++){
    println lines[i]
  }
  
  
  

Yeah figured it was worth mentioning and thanks for the original comments as it helped me move on. For completeness and to complement above answer here's what I used..

def lines = new String(multiLineFile).split( '\n' )
for(int i = 0; i < lines.size(); i++){
  println lines[i]
}

[jenkins-infra-bot]

jglick:

• Original comment link
• Raw content of original comment:

  That works, though note that new String(multiLineFile) is redundant; you can just use multiLineFile directly.
  

That works, though note that new String(multiLineFile) is redundant; you can just use multiLineFile directly.

[jenkins-infra-bot]

rawlingsj:

• Original comment link
• Raw content of original comment:

  Awesome - thanks!
  

Awesome - thanks!

[jenkins-infra-bot]

fchuong:

• Original comment link
• Raw content of original comment:

  workflow-plugin 1.10, using @com.cloudbees.groovy.cps.NonCPS does produce the expected output (when combined with synchronous steps such as org.jenkinsci.plugins.workflow.steps.EchoStep.Execution) as a workaround to groovy-cps #9.
  
  
  @com.cloudbees.groovy.cps.NonCPS
  def foo1() {
    [1, 2, 3].each {
      println it
    }
  }
  
  @com.cloudbees.groovy.cps.NonCPS
  def foo2() {
    println "abc".replaceAll(/[a-z]/) {
      it.toUpperCase()
    }
  }
  
  node {
    foo1()
    foo2()
  }
  
  
  
  Started by user anonymous
  Running: Allocate node : Start
  Running on master in xxx
  Running: Allocate node : Body : Start
  Running: Print Message
  1
  Running: Print Message
  2
  Running: Print Message
  3
  Running: Print Message
  ABC
  Running: Allocate node : Body : End
  Running: Allocate node : End
  Running: End of Workflow
  Finished: SUCCESS
  
  
  

workflow-plugin 1.10, using @com.cloudbees.groovy.cps.NonCPS does produce the expected output (when combined with synchronous steps such as org.jenkinsci.plugins.workflow.steps.EchoStep.Execution) as a workaround to groovy-cps #9.

@​com.cloudbees.groovy.cps.NonCPS
def foo1() {
  [1, 2, 3].each {
    println it
  }
}

@​com.cloudbees.groovy.cps.NonCPS
def foo2() {
  println "abc".replaceAll(/[a-z]/) {
    it.toUpperCase()
  }
}

node {
  foo1()
  foo2()
}

Started by user anonymous
Running: Allocate node : Start
Running on master in xxx
Running: Allocate node : Body : Start
Running: Print Message
1
Running: Print Message
2
Running: Print Message
3
Running: Print Message
ABC
Running: Allocate node : Body : End
Running: Allocate node : End
Running: End of Workflow
Finished: SUCCESS

[jenkins-infra-bot]

jglick:

• Original comment link
• Raw content of original comment:

  fchuong while it may work to run certain steps inside a @NonCPS method (generally only those using AbstractSynchronousStepExecution: echo, error, pwd), in general it will fail horribly, so my recommendation is to never call any step from inside such a method: just use it to do some computation and then pass the result back to the flow. Ideally the engine would automatically detect the attempt and reject it.
  

fchuong while it may work to run certain steps inside a @​NonCPS method (generally only those using AbstractSynchronousStepExecution: echo, error, pwd), in general it will fail horribly, so my recommendation is to never call any step from inside such a method: just use it to do some computation and then pass the result back to the flow. Ideally the engine would automatically detect the attempt and reject it.

[jenkins-infra-bot]

fchuong:

• Original comment link
• Raw content of original comment:

  Yes, which is why synchronous steps were mentioned when using @NonCPS as the current implementation already fails with basic Groovy:
  
  @com.cloudbees.groovy.cps.NonCPS
  def doAssertNoCPS() {
    // no error when CPS disabled
    assert [1, 2, 3].collect { it } == [1, 2, 3]
  }
  def doAssertCPS() {
    // fails because CPS transformation results in '4'
    assert [4, 5, 6].collect { it } == [4, 5, 6]
  }
  
  doAssertNoCPS()
  doAssertCPS()
  
  
  To be more relevant with current issue, it prevents such direct Groovy approach:
  
  // won't retrieve parameter value as result of 'find' is inconsistent
  currentBuild.rawBuild.getAction(hudson.model.ParametersAction).parameters.find { it.name == 'FOO' }.value
  
  
  

Yes, which is why synchronous steps were mentioned when using @​NonCPS as the current implementation already fails with basic Groovy:

@​com.cloudbees.groovy.cps.NonCPS
def doAssertNoCPS() {
  // no error when CPS disabled
  assert [1, 2, 3].collect { it } == [1, 2, 3]
}
def doAssertCPS() {
  // fails because CPS transformation results in '4'
  assert [4, 5, 6].collect { it } == [4, 5, 6]
}

doAssertNoCPS()
doAssertCPS()

To be more relevant with current issue, it prevents such direct Groovy approach:

// won't retrieve parameter value as result of 'find' is inconsistent
currentBuild.rawBuild.getAction(hudson.model.ParametersAction).parameters.find { it.name == 'FOO' }.value

[jenkins-infra-bot]

eholzwarth:

• Original comment link
• Raw content of original comment:

  I'm not understanding the workaround, apparently. I tried adding the NonCPS annotation as follows:
  
  
  		@com.cloudbees.groovy.cps.NonCPS 
  		def targetsNotYetDeployed = config.deployTargets.findAll{ !it.hasProperty("deployed") }
  		def done = (targetsNotYetDeployed.size() == 0)
  
  
  But this still throws:
  
  groovy.lang.MissingMethodException: No signature of method: java.lang.Boolean.size() is applicable for argument types: () values: []
  Possible solutions: is(java.lang.Object), find(), find(groovy.lang.Closure), with(groovy.lang.Closure), sleep(long), use([Ljava.lang.Object;)
  	at org.codehaus.groovy.runtime.ScriptBytecodeAdapter.unwrap(ScriptBytecodeAdapter.java:55)
  	at org.codehaus.groovy.runtime.callsite.PojoMetaClassSite.call(PojoMetaClassSite.java:46)
  	at org.codehaus.groovy.runtime.callsite.CallSiteArray.defaultCall(CallSiteArray.java:42)
  	at org.codehaus.groovy.runtime.callsite.AbstractCallSite.call(AbstractCallSite.java:108)
  	at com.cloudbees.groovy.cps.sandbox.DefaultInvoker.methodCall(DefaultInvoker.java:15)
  	at WorkflowScript.flow(WorkflowScript:52)
  	at WorkflowScript.run(WorkflowScript:91)
  	at ___cps.transform___(Native Method)
  	...
  
  
  How can I work around this?
  Thanks
  

I'm not understanding the workaround, apparently. I tried adding the NonCPS annotation as follows:

		@​com.cloudbees.groovy.cps.NonCPS 
		def targetsNotYetDeployed = config.deployTargets.findAll{ !it.hasProperty("deployed") }
		def done = (targetsNotYetDeployed.size() == 0)

But this still throws:

groovy.lang.MissingMethodException: No signature of method: java.lang.Boolean.size() is applicable for argument types: () values: []
Possible solutions: is(java.lang.Object), find(), find(groovy.lang.Closure), with(groovy.lang.Closure), sleep(long), use([Ljava.lang.Object;)
	at org.codehaus.groovy.runtime.ScriptBytecodeAdapter.unwrap(ScriptBytecodeAdapter.java:55)
	at org.codehaus.groovy.runtime.callsite.PojoMetaClassSite.call(PojoMetaClassSite.java:46)
	at org.codehaus.groovy.runtime.callsite.CallSiteArray.defaultCall(CallSiteArray.java:42)
	at org.codehaus.groovy.runtime.callsite.AbstractCallSite.call(AbstractCallSite.java:108)
	at com.cloudbees.groovy.cps.sandbox.DefaultInvoker.methodCall(DefaultInvoker.java:15)
	at WorkflowScript.flow(WorkflowScript:52)
	at WorkflowScript.run(WorkflowScript:91)
	at ___cps.transform___(Native Method)
	...

How can I work around this?

Thanks