Re: Soot API: Instrumenting EJB

[Feng QIAN <fqian@sable.mcgill.ca>]

Hi,

#168 is the JSR instruction. It was supposed to be inlined before calling
generateJimple. It seems that the JSR-inliner does not handle this case
corectly. You can read CFG.eliminateJsrRets, it assumes well-behaved
JSR/RET pairs. You may help to fix it.

Can you mail me the class file on which the Soot failed?

Cheers,
====================
Feng Qian					fqian@sable.mcgill.ca

On Tue, 25 Sep 2001, nathan gulley wrote:

> Hello,
>
> I am using Soot 1.2.2 in an attempt to instrument a method on a Java EJB
> at ClassLoadtime using a custom ClassLoader. However, A RunTimeException
> is thrown by the soot.coffi.CFG.generateJimple(CFG.java:4517) method when
> I try to i retrieve the active body of one of the EJB's method's before
> instrumenting it:
>
> JimpleBody jimpBody = (JimpleBody)method.retrieveActiveBody();
>
> java.lang.RuntimeException: Unrecognized bytecode instruction: 168
>         at soot.coffi.CFG.generateJimple(CFG.java:4517)
>         at soot.coffi.CFG.jimplify(CFG.java:1257)
>         at soot.coffi.CFG.jimplify(CFG.java:920)
>         at
> soot.coffi.CoffiMethodSource.getBody(CoffiMethodSource.java:98)
>         at soot.SootMethod.getBodyFromMethodSource(SootMethod.java:83)
>         at soot.SootMethod.retrieveActiveBody(SootMethod.java:276)
>         at
> com.rtscinc.util.classloaders.XClassLoaderII.getXClassData(XClassLoad
> erII.java:404)
>         at
> com.rtscinc.util.classloaders.XClassLoaderII.getClassData(XClassLoade
> rII.java:326)
>         at
> com.rtscinc.util.classloaders.XClassLoaderII.findClass(XClassLoaderII
> .java:282)
>         at
> com.rtscinc.util.classloaders.XClassLoaderII.loadClass(XClassLoaderII
> .java:256)
>         at test.TestClient.test15(TestClient.java:1010)
>         at java.lang.reflect.Method.invoke(Native Method)
>         at test.TestClient.main(TestClient.java:46)
>
> I appears that the Soot 1.2.2 Framework is unable to handle byte code
> instruction # 168 so I examined the CFG.java source code to make sure. In
> fact there wasn't any code to handle this particular instruction so
> that's why a RunTimeException was being thrown by
> soot.coffi.CFG.generateJimple() method.
>
> Is anyone aware of a workaround for this problem? it looks like someone
> was working on this because there are some commented out lines in
> CFG.java and Jimple.java  that had code to handle this instruction but it
> hasn't been fully implemented.
>
> I'd appreciate any advice or help with this.
>
> Nathan
>
>
>