HTMLUnit 4.19.0 and 4.20.0 report 'caller is not a function' in the Jenkins test harness

[MarkEWaite]

HTMLUnit 4.19.0 and 4.20.0 report that the caller is not a function for the sortable-drag-drop.js file.

A sample stack trace looks like this:

EcmaError: lineNumber=[115] column=[0] lineSource=[<no source>] name=[TypeError] sourceName=[https://localhost:44557/jenkins/static/c3b2c270/jsbundles/sortable-drag-drop.js] message=[TypeError: caller is not a function, it is object. (https://localhost:44557/jenkins/static/c3b2c270/jsbundles/sortable-drag-drop.js#115)]
org.htmlunit.ScriptException: TypeError: caller is not a function, it is object. (https://localhost:44557/jenkins/static/c3b2c270/jsbundles/sortable-drag-drop.js#115)
        at org.htmlunit.javascript.JavaScriptEngine$HtmlUnitContextAction.run(JavaScriptEngine.java:938)
        at org.htmlunit.corejs.javascript.Context.call(Context.java:636)
        at org.htmlunit.corejs.javascript.ContextFactory.call(ContextFactory.java:462)
        at org.htmlunit.javascript.HtmlUnitContextFactory.callSecured(HtmlUnitContextFactory.java:325)
        at org.htmlunit.javascript.JavaScriptEngine.execute(JavaScriptEngine.java:785)
        at org.htmlunit.javascript.JavaScriptEngine.execute(JavaScriptEngine.java:105)
        at org.htmlunit.html.HtmlPage.loadExternalJavaScriptFile(HtmlPage.java:1030)
        at org.htmlunit.html.ScriptElementSupport.executeScriptIfNeeded(ScriptElementSupport.java:185)

HTMLUnit 4.18.0 passes tests in the Jenkins test harness and is used widely in Jenkins.

The sortable-drop-drop.js source file does not have 115 lines. Line 115 of jsbundles/sortable-drag-drop.js in my Jenkins installation looks like this:

/******/                                        if ((priority & 1 === 0 || notFulfilled >= priority) && Object.keys(__webpack_require__.O).every(function(key) { return __webpack_require__.O[key](chunkIds[j]); })) {

I assume that is something that is added by webpack. Any hints to suggest what we should investigate so that we can upgrade to HTMLUnit 4.20.0?

Related pull requests

• Bump org.htmlunit:htmlunit from 4.18.0 to 4.20.0 jenkinsci/jenkins-test-harness-htmlunit#202
• Bump org.htmlunit:htmlunit from 4.18.0 to 4.19.0 jenkinsci/jenkins-test-harness-htmlunit#201

[rbri]

@MarkEWaite sorry for that, core-js got a bunch of updates for this two releases improving the js support a lot. At the moment i have no idea what is wrong here but will have a look. Give me some seconds...

[rbri]

@MarkEWaite any chance to 'download' the sortable-drag-drop.js file while running one of the failing tests? Or maybe manually do parts of the test in the browser to get the content of that file?

[MarkEWaite]

@MarkEWaite any chance to 'download' the sortable-drag-drop.js file while running one of the failing tests? Or maybe manually do parts of the test in the browser to get the content of that file?

Here is a tar.gz of the jsbundles directory from my running Jenkins controller.

jsbundles.tar.gz

[rbri]

had a first look and wrote some test for the 'caller' property of functions but found nothing special. Will have another look tomorrow

[rbri]

@MarkEWaite sorryn no real progress so far, will be away from my pc for two days. Will try to locally reproduce this on friday and inform you about the progress.

[MarkEWaite]

@MarkEWaite sorryn no real progress so far, will be away from my pc for two days. Will try to locally reproduce this on friday and inform you about the progress.

Thanks so much for helping with the investigation. I've also made no progress. I'll keep you informed if I make any further progress as well. I likely won't look at this in more detail until I have some weekend time to spend on it.

[rbri]

my current theory

Object.keys(__webpack_require__.O).every(function(key) { return __webpack_require__.O[key](chunkIds[j]); })

Objects.keys returns for unknown reason the 'caller' property - every then tries to retrieve and execute the value and the value is not a function.

[rbri]

Have a first idea about the root cause of this. Working already on a rhino PR but this might requere some time.

[rbri]

@MarkEWaite have spend the last days working on this - in fact I have changed the way the 'arguments' and 'caller' properties of functions are implemented. This is now much, much closer to the spec and what browsers are doing.
Hopefully this solves our problem...

Will make a new snapshot build available soon, hopefully you can test that.

[rbri]

@MarkEWaite the 4.21.0-SNAPSHOT is available (requires a 4.21.0-SNAPSHOT of core-js)

[MarkEWaite]

@MarkEWaite the 4.21.0-SNAPSHOT is available (requires a 4.21.0-SNAPSHOT of core-js)

I tried to build a snapshot of core-js but it seems to require a snapshot of rhino-fork. I was unable to build rhino-fork. I received various errors from Gradle while trying to build rhino-fork. Is there another way that I could obtain the snapshot builds for testing?

[rbri]

@MarkEWaite oh sorry my fault - there is a snapshot available. You simply have to use it ;-)
https://github.com/HtmlUnit/htmlunit?tab=readme-ov-file#last-ci-build

The snapshot already points to the (latest) core-js snapshot, in theory everything should work out of the box. Simply change the pom of the htmlunit test harness