Re: [Jruby-devel] JRuby 0.8.3 Released

[Charles O Nutter]

I think I see what's going on. You get the ruby instance in one thread, and before you call anything in that thread you call against it in another thread. The issue here is that although that non-main thread gets adopted, it's missing one piece of the puzzle needed to execute: self. The main execution path (started up by init(), which is called when creating a new Ruby instance) handles this correctly. The adopted path does not.

The fix was to also set self on the adopted thread's frame and evaluator. I would LOVE to commit this fix right now, but SF CVS is still down. So here's the changed code. With this fix, your test case runs successfully.

in ThreadContext.java:
    public void preAdoptThread() {
        pushIter(Iter.ITER_NOT);
        pushFrame();
        getCurrentFrame().newScope(null);
        pushRubyClass(runtime.getObject());
        pushCRef(runtime.getObject());
        getCurrentFrame().setSelf(runtime.getTopSelf());
        getCurrentFrame().getEvalState().setSelf( runtime.getTopSelf());
    }

I'll make an additional test case out of your code, and commit both when I am able. Thank you for finding this!

On 3/30/06, Werner Schuster (murphee) <[EMAIL PROTECTED]> wrote:

Charles O Nutter wrote:

>Yes, that AOOB was the one I saw breaking. I'll see if I can find a way to
>reproduce your other issue.
>
>

OK, I created a simple test that reproduces the NPE. If you run this code,
you'll get an NPE when calling the defined "foo" method from a new
thread; it works if
called from the Main Thread where Ruby was instantiated.

This will give an exception in a different place:
java.lang.NullPointerException
    at
org.jruby.evaluator.EvaluateVisitor$VCallNodeVisitor.execute(EvaluateVisitor.java:1923)
    at
org.jruby.evaluator.EvaluationState.executeNext(EvaluationState.java:211)
    at org.jruby.evaluator.EvaluationState.begin(EvaluationState.java:291)
    at org.jruby.Ruby.eval(Ruby.java:188)
    at org.jruby.Ruby.evalScript(Ruby.java:181)
    at Test$1.run(Test.java:17)
    at java.lang.Thread.run (Thread.java:595)


Having debugged the issue in the EclipseShell code, it seems like the
InstructionBundles
EvaluationState returns null for it's "self" field.

Test code attached.


murphee
--
Blog @ http://jroller.com/page/murphee
Maintainer of EclipseShell @ http://eclipse-shell.sourceforge.net/

--
Charles Oliver Nutter @ headius.blogspot.com
JRuby Developer @ jruby.sourceforge.net
Application Architect @ www.ventera.com