citgm-smoker-nobuild fails to start on Node 6 and 8

[targos]

See the runs from https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker-pipeline/46/console

15:28:19 Started by upstream project "citgm-smoker-pipeline" build number 46
15:28:19 originally caused by:
15:28:19  Started by user Michaël Zasso
15:28:19 [EnvInject] - Loading node environment variables.
15:28:19 Building remotely on test-packetnet-ubuntu1604-x64-1 (Ubuntu Ubuntu-16.04 amd64-Ubuntu jenkins-workspace amd64-Ubuntu-16.04 16.04 amd64) in workspace /home/iojs/build/workspace/citgm-smoker-nobuild
15:28:19 using NODE_VERSION: v8
15:28:19 sliced: 8
15:28:19 FATAL: String index out of range: 2
15:28:19 java.lang.StringIndexOutOfBoundsException: String index out of range: 2
15:28:19 	at java.lang.String.substring(String.java:1963)
15:28:19 	at org.codehaus.groovy.runtime.StringGroovyMethods.getAt(StringGroovyMethods.java:1480)
15:28:19 	at org.codehaus.groovy.runtime.StringGroovyMethods.getAt(StringGroovyMethods.java:1467)
15:28:19 	at org.codehaus.groovy.runtime.dgm$1086.invoke(Unknown Source)
15:28:19 	at org.codehaus.groovy.runtime.callsite.PojoMetaMethodSite$PojoMetaMethodSiteNoUnwrapNoCoerce.invoke(PojoMetaMethodSite.java:274)
15:28:19 	at org.codehaus.groovy.runtime.callsite.PojoMetaMethodSite.call(PojoMetaMethodSite.java:56)
15:28:19 	at org.codehaus.groovy.runtime.callsite.CallSiteArray.defaultCall(CallSiteArray.java:48)
15:28:19 	at org.codehaus.groovy.runtime.callsite.AbstractCallSite.call(AbstractCallSite.java:113)
15:28:19 	at org.codehaus.groovy.runtime.callsite.AbstractCallSite.call(AbstractCallSite.java:125)
15:28:19 	at Script1.run(Script1.groovy:9)
15:28:19 	at groovy.lang.GroovyShell.evaluate(GroovyShell.java:585)
15:28:19 	at groovy.lang.GroovyShell.evaluate(GroovyShell.java:623)
15:28:19 	at groovy.lang.GroovyShell.evaluate(GroovyShell.java:594)
15:28:19 	at org.jenkinsci.plugins.scriptsecurity.sandbox.groovy.SecureGroovyScript.evaluate(SecureGroovyScript.java:343)
15:28:19 	at org.jenkinsci.plugins.scriptsecurity.sandbox.groovy.SecureGroovyScript$evaluate.call(Unknown Source)
15:28:19 	at org.jenkinsci.plugins.ScriptRunner.run(ScriptRunner.groovy:71)
15:28:19 	at org.jenkinsci.plugins.ScriptRunner$run.call(Unknown Source)
15:28:19 	at org.jenkinsci.plugins.GroovyScriptMES.decideOrder(GroovyScriptMES.groovy:62)
15:28:19 	at org.jenkinsci.plugins.GroovyScriptMES$decideOrder.callCurrent(Unknown Source)
15:28:19 	at org.jenkinsci.plugins.BaseMES.run(BaseMES.groovy:51)
15:28:19 	at hudson.matrix.MatrixBuild$MatrixBuildExecution.doRun(MatrixBuild.java:375)
15:28:19 	at hudson.model.AbstractBuild$AbstractBuildExecution.run(AbstractBuild.java:504)
15:28:19 	at hudson.model.Run.execute(Run.java:1819)
15:28:19 	at hudson.matrix.MatrixBuild.run(MatrixBuild.java:323)
15:28:19 	at hudson.model.ResourceController.execute(ResourceController.java:97)
15:28:19 	at hudson.model.Executor.run(Executor.java:429)
15:28:19 All downstream projects complete!
15:28:19 Notifying upstream projects of job completion
15:28:19 Join notifier cannot find upstream project: citgm-smoker-pipeline
15:28:19 Finished: FAILURE

/cc @richardlau

[richardlau]

There is a bug with the groovy script in the job. I tried to fix it as part of #1577 but groovy script edits in Jenkins jobs require admin approval (being a member of citgm-admins is not sufficient).

I can't access the CI at the moment since it's locked down for the security releases.

[richardlau]

It may be possible to work around the bug by specifying a more exact version (e.g. v8.13.0 instead of v8).

[refack]

So in Groovy there's a difference between '.' and ".", and .split needs to former. NM.

Well this works:

  def verString = parameters['NODE_VERSION'].toString()
  println "using NODE_VERSION: " + verString
  def sliced = verString.substring(1)
  println "sliced: $sliced"
  def spaced = spaced.replace(".", " ")
  def split = sliced.split()
  println "split : $split"
  nodeMajorVersion = split[0].toInteger()

[refack]

And BTW ws fails :(

[refack]

BTW2 I restored fetching the groovy script from GitHub which I hacked in #1509 (comment)

[richardlau]

Instead of the slicing/replacing, we could always use a regexp to extract the major version. Something like:

def nodeMajorVersion = (verString =~ /v?(\d+)/)[0][1]

[refack]

def nodeMajorVersion = (verString =~ /v?(\d+)/)[0][1]

I think I don't like Groovy 😢

IMO what we should do is rename the job Parameters like I did in https://ci.nodejs.org/job/refack-node-test-node-addon-api/, and use numbers instead of strings (with no v), and then update citgm-smoker-pipeline.

[richardlau]

def nodeMajorVersion = (verString =~ /v?(\d+)/)[0][1]

I think I don't like Groovy 😢

IMO what we should do is rename the job Parameters like I did in https://ci.nodejs.org/job/refack-node-test-node-addon-api/, and use numbers instead of strings (with no v), and then update citgm-smoker-pipeline.

That sounds more or less what I did with the node-report job: #1421

[rvagg]

@refack can I get you to hold off on the citgm job testing please? I need to get all these security jobs through and your testing is blocking that. I'd really like a pause on all non-security-release-related CI activity to be honest. There's a ton to do including fixing small bugs across all release lines and getting to citgm when things are stable.

[refack]

Abort them. The pipe line just schedules a ton of job.

[rvagg]

I feel bad cancelling people's stuff, it's always hard to know how critical something is or how much I'm going to ruin their day. Thanks for cancelling, it turns out that I have a small, but extremely hairy yak to shave and it's taking a lot of jenkins time to get it done.

[lpinca]

And BTW ws fails :(

@refack we are testing it in AppVeyor, Node v6, v8, v10, v11 both x86 and x64. Why it fails on CITGM? What tests are failing?

[richardlau]

BTW2 I restored fetching the groovy script from GitHub which I hacked in #1509 (comment)

Had to fix up the matrix axis name (MACHINE->nodes) to match what the script was expecting: #1632 (comment)

Anyway I think the reported issue has now been fixed?

[targos]

Seems fixed indeed