Subj : Re: Rhino1_5R41 class name creation
To   : Tomita Tomohiro <tomi@osaka.niss.co.jp>
From : Igor Bukanov <igor@fastmail.fm>
Date : Thu May 29 2003 03:12 pm

Tomita Tomohiro wrote:
> Hi There,
> 
> I use Rhino1_5R41 and encountered exception as below:
> -----------------------------------------------------------------------
> java.lang.RuntimeException: java.lang.ClassFormatError:
> org/mozilla/javascript/gen/c-2147483648 (Illegal Class name
> "org/mozilla/javascript/gen/c-2147483648")
>  at org.mozilla.javascript.optimizer.Codegen.compile(Codegen.java:113)
>  at org.mozilla.javascript.Context.compile(Context.java:2018)
>  at org.mozilla.javascript.Context.compile(Context.java:1948)
>  at org.mozilla.javascript.Context.compileReader(Context.java:899)
>  at Rhino.compile(Test.java:65)
>  at Test.main(Test.java:27)
> -----------------------------------------------------------------------
> org.mozilla.javascript.optimizer.OptClassNameHelper#getJavaScriptClassName()
> is used to name the Java class compiled from javaScript.
> In the OptClassNameHelper#getJavaScriptClassName(), the globalSerial
> value is used, except function, to name the class to guarantee
> uniqueness of the class name.
> The globalSerial value type is int, and static defined, so if you
> continue to compile JavaScript and create JavaClass, you will reach the
> value of Integer.MAX_VALUE(2147483647) eventually.
> If you goes on, the globalSerial value turnes into Integer.MIN_VALUE
> (-2147483648) and compile would failed because of Java class name must
> not contain the '-' character.
> I think this is bug.
> Is it right?

You are right about it but I am currious how long it took to trigger the 
overflow?

> And I suggest that globalSerial value would be reset to 1 before
> over-flow.
> Does this measure cause any problem?

It should not.

Regards, Igor

.