Subj : Re: Is rhino's Codegen.generateCode() method thread safe?
To   : mozilla-jseng@mozilla.org
From : tomi@osaka.niss.co.jp (Tomita Tomohiro)
Date : Thu Mar 13 2003 05:14 am


On Wed, 12 Mar 2003 14:38:40 +0100
Igor Bukanov <igor@icesoft.com> wrote:

> Tomita Tomohiro wrote:
> > Hi,
> > I have some question about rhino.
> 
> What version of Rhino do you use and what is your JVM name/version/platform?
I use rhino version 1_5R3.
and JVM is J9 in WSDD5.0 for MontaVistaLinux3.0/XScale.
It is not a FCS version,so it is supposed that the jvm have some bugs as
you guess.
> 
> > 
> > I'm using rhino in multi thread.
> > Is the Codegen.generateCode() method thread safe?
> 
> Codegen is not thread safe but its usage is safe since its instances are 
> accessed only by the single thread that creates them.
I see.
> 
> > #I have no time to trace or debug rhino source,so I ask this ML :-)
> > 
> > in my enviroment(j9VM) the error occures very often as follows:
> > java.lang.StringIndexOutOfBoundsException
> > Stack trace:
> >    java/lang/Throwable.<init>()V
> >    java/lang/String.charAt(I)C
> >    org/mozilla/classfile/ClassFileWriter.sizeOfParameters(Ljava/lang/String;)I
> >    org/mozilla/classfile/ClassFileWriter.add(BLjava/lang/String;Ljava/lang/String;Ljava/lang/String;Ljava/lang/String;)V
> >    org/mozilla/javascript/optimizer/Codegen.addStaticInvoke(Ljava/lang/String;Ljava/lang/String;Ljava/lang/String;Ljava/lang/String;)V
> >    org/mozilla/javascript/optimizer/Codegen.visitRegularCall(Lorg/mozilla/javascript/Node;ILorg/mozilla/javascript/Node;Z)V
> >    org/mozilla/javascript/optimizer/Codegen.visitCall(Lorg/mozilla/javascript/Node;ILorg/mozilla/javascript/Node;)V
> >    org/mozilla/javascript/optimizer/Codegen.generateCodeFromNode(Lorg/mozilla/javascript/Node;Lorg/mozilla/javascript/Node;II)V
> >    org/mozilla/javascript/optimizer/Codegen.generateCodeFromNode(Lorg/mozilla/javascript/Node;Lorg/mozilla/javascript/Node;II)V
> >    org/mozilla/javascript/optimizer/Codegen.generateCodeFromNode(Lorg/mozilla/javascript/Node;Lorg/mozilla/javascript/Node;II)V
> >    org/mozilla/javascript/optimizer/Codegen.generateCode(Lorg/mozilla/javascript/Node;Ljava/util/Vector;Ljava/util/Vector;Lorg/mozilla/javascript/ClassNameHelper;)Ljava/lang/String;
> >    org/mozilla/javascript/optimizer/Codegen.compile(Lorg/mozilla/javascript/Context;Lorg/mozilla/javascript/Scriptable;Lorg/mozilla/javascript/Node;Ljava/lang/Object;Lorg/mozilla/javascript/SecuritySupport;Lorg/mozilla/javascript/ClassNameHelper;)Ljava/lang/Object;
> >    org/mozilla/javascript/Context.compile(Lorg/mozilla/javascript/Scriptable;Lorg/mozilla/javascript/TokenStream;Ljava/lang/Object;Ljava/io/Reader;Z)Ljava/lang/Object;
> >    org/mozilla/javascript/Context.compile(Lorg/mozilla/javascript/Scriptable;Ljava/io/Reader;Ljava/lang/String;ILjava/lang/Object;Z)Ljava/lang/Object;
> >    org/mozilla/javascript/Context.compileReader(Lorg/mozilla/javascript/Scriptable;Ljava/io/Reader;Ljava/lang/String;ILjava/lang/Object;)Lorg/mozilla/javascript/Script;
> >    tomi.run()V
> >    java/lang/Thread.run()V
> > 
> > if compiling JavaScript in single thread, then it works good.
> 
> The fact that org/mozilla/classfile/ClassFileWriter.sizeOfParameters() 
> accesses a character outside string is a bug in sizeOfParameters since 
> the function is supposed to check for bad method signatures and throw 
> exlicit exception about it. But its usage in 
> org/mozilla/javascript/optimizer/Codegen.java is correct, so I suspect a 
> bug in JVM.
> 
> Regards, Igor
> 
Thanks
Tomohiro

.