Mailing-List: contact bcel-user-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "BCEL Users List" <bcel-user@jakarta.apache.org>
Received-SPF: pass (hermes.apache.org: domain of
 rajendra.kumar.reddy.koduru@sap.com designates 155.56.68.140 as permitted
 sender)
content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="US-ASCII"
Content-Transfer-Encoding: quoted-printable
Subject: RE: Query regarding index of local variables
Date: Wed, 9 Feb 2005 12:03:24 +0100
Message-ID: <73DFF891F8024B48AD56B594C4C5038A6BDAD4@dewdfe23.wdf.sap.corp>
Thread-Topic: Query regarding index of local variables
Thread-Index: AcUN/LbpQcZl08i2S6y9oOoHRiOVlwAmMj0A
From: "Koduru, Rajendra Kumar Reddy" <rajendra.kumar.reddy.koduru@sap.com>
To: "BCEL Users List" <bcel-user@jakarta.apache.org>

Hi Andrew,=20

  I have gone thru this concept of assigning a slot to local variables
by observing the local variable tables of  the class files, yep each
slot is being taken by many local var. so I used getMaxLocals() and it
worked fine (removed my own way of fidning max index, which sometimes is
causing an error due to incomplete lv table).

Please find a comment inline regarding the comment made by you "stack
height mismatch verification error".


=09
=20

>Koduru, Rajendra Kumar Reddy wrote:
>> Hi Andrew
>>=20
>>    Thank you for recommending me the VM spec...
>
>Reading the VM spec is not as stupid an idea as it might=20
>sound.  I'm not=20
>aware of any more concise or helpful reference for bytecode=20
>manipulators.  Can anyone suggest a better one?
>
>>=20
>>   But my prob is a little different,
>
>It seems that your problem can be summarized as, "How do i=20
>choose a free=20
>slot?"  That's the question i answered in my previous email. =20
>see below=20
>for further explanation.
>
>>=20
>> 	as you have said I can add local variables using=20
>>=20
>> 	LocalVar var =3D localvargen.addLocalvar()
>> 	int index =3D var.getIndex();
>> 	llcalvargen.setstart( ilist.append ( new ASTORE ( index ) ) )
>> 	.
>> 	.
>> 	.=20
>>  =20
>> 	as many as required
>>=20
>> 	and at the calling setMaxLocals();
>>=20
>> 	That is OK.
>
>none of this is actually required.  the local variable table=20
>is just for=20
>debugging purposes, and as i said before, may be inaccurate,=20
>incomplete,=20
>obfuscated, or missing.  feel free to ignore LocalVar objects=20
>completely=20
>and just choose some slots yourself and use them.
>
>The usual way of choosing slots, as far as I'm aware, is to call=20
>getMaxLocals:
>
>MethodGen mg =3D ...
>int freeSlot =3D mg.getMaxLocals();
>int freeSlot2 =3D freeSlot + 1;
>mg.getInstructionList().append(new ASTORE(freeSlot));
>...
>
>>=20
>> 	but now consider a scenario like this
>> =09
>> 	public int getI(){
>> 		return 2;
>> 	}
>>=20
>> 	Now I am trying to insert try-finally to this method by
>> instrumenting in bytecode=20
>>=20
>> 	which should look like =20
>>=20
>> 	public int getI(){
>> 	try{
>> 		return 2;
>> 	}finally {=20
>> 			System.out.println("sample");
>> 	}
>> =09
>> 	}=09
>>=20
>> 	To create this, I should create a empty catch block which throws
>> the exception that is not handled and redirect all the calls=20
>to finally
>> before the=20
>> 	return statement
>>=20
>> 	now comes the problem with local variables.
>>=20
>>=20
>> 	I will get the instructionlist
>> 	I will find the instance of return, and insert a JSR instr
>> before this return and redirect all the branches
>
>btw, when you get past your other problems, it seems like you=20
>might have=20
>a stack height mismatch verification error at your subroutine.  ask=20
>about that if it happens.  if i'm right, it'll happen in any non-void=20
>method you instrument.


	Andrew, I did not get that type of error, however could you
please let me know , why were you expecting this??
	in non void methods, I was using=20

	1.	new ASTORE(x) 	// storing the result
	2.	jsr call		// calling jsr
	3.	new ALOAD(x)	// loading back the result
	4.	return		// returning the result

	and all the branches were reidrected to 1 (i.e store statement)
in case of non void methods

	so could you please let me know , why were you expecting an
error?? So that I could rectify it
	am I right in doing like this???

	However just to mention, I was also using methodgen.setMaxStack
before returning the isntrumented method


>
>> 	Next step is creation of catch block, where I have a problem
>> 			here I should store the exception, make a jsr
>> call, load the exception which I stored before and rethrow it
>>=20
>> 	so I use ilist.append ( new ASTORE ( ? ) )
>> 		   ilist.append ( new JSR() );
>> 		   ilist.load ( new ALOAD ( ? ) )
>> =09
>> 	so what should I use in the place of "?"  (2,3,4......)
>
>i suggest mg.getMaxLocals()
>
>>=20
>> 	same while creating finally block, I should store the return
>> address
>> =09
>> 		   ilist.append ( new ASTORE ( ?? ) )
>> 		   // performs the instrumentation of system.out.....
>> 		   ret ??=20
>> =09
>> 	so what should I use in place of "?"  (2,3,4......)
>
>i suggest mg.getMaxLocals() + 1
>
>>=20
>> 	so what I am doing is=20
>> =09
>> 	getting the max_index (maximum index) available in the
>> localvargen=20
>> 	and assigning max_index+1 to the first ?
>> 	and assigning max_index+2 to the second ??
>>=20
>> 	am I wrong in doing like these?????
>
>If the local variable table were complete and accurate, this method=20
>would be correct, but you should not assume that. =20
>getMaxLocals() always=20
>returns a free slot (assuming you call setMaxLocals()=20
>appropriately) and=20
>frees you from writing your own code to find max_index.
>
>>=20
>> 	is there some other efficient way for doing this???
>
>getMaxLocals()
>
>>=20
>> 	Please help me in this regard.
>>=20
>> Thank you
>> Reddy
>> 	=09
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: bcel-user-unsubscribe@jakarta.apache.org
>For additional commands, e-mail: bcel-user-help@jakarta.apache.org
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: bcel-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: bcel-user-help@jakarta.apache.org