[Soot-list] How to modify soot method to generate this kind of IR code?

[Eric Bodden]

Hi Ding.

I am not sure what you are asking. The conversion from Java to Jimple
does exactly the transformation that you describe: it breaks such
compound expressions down to single statements. Have you actually
looked at the Jimple code?

Eric

--
Dr. Eric Bodden
Software Technology Group, Technische Universität Darmstadt, Germany
Tel: +49 6151 16-5478    Fax: +49 6151 16-5410
Mailing Address: S2|02 A209, Hochschulstraße 10, 64289 Darmstadt



On 23 June 2010 14:05, DING SUN <xyz031702 at hotmail.com> wrote:
> Dear All
>     I'm using soot (I'm using Jimple format) for inter-procedure analysis
> (forward sequence). I met a case:
> void main(){
>   int p1=2;
>   int p2=3;
>   int result= f1( f2(p1), f3(p2));
>
> }
>
> f1(int x, int y), f2(int x), f3(int x) are all normal functions.
>
> If I want to analysis f1(), I must known its parameters' actual values,
> which are return values from f2() and f3() respectively.
>
> So I think the proper call sequence shall be :  main-->
> f2()-->main-->f3() --> main-->f1()-->main()
> In another word, is it possible to transform the above piece of code into
> the following IR code ?:
>
> void main(){
>   int p1=2;
>   int p2=3;
>   int f1p1=f2(p1);
>   int f1p2=f3(p2);
>   int result= f1( f1p1, f1p2);
> }
>
> Would you suggest me a good way to solve this?
>
> Best Regards,
> Ding Sun
> --------------------------------------------------
> From: "Eric Bodden" <bodden at st.informatik.tu-darmstadt.de>
> Sent: Wednesday, June 23, 2010 5:18 PM
> To: "Jochen Wuttke" <wuttkej at usi.ch>
> Cc: <soot-list at sable.mcgill.ca>
> Subject: Re: [Soot-list] How to get the type of locals in a FlowAnalysis?
>
>> Hi Jochen.
>>
>> The problem is that in bytecode different primitive type are mangled
>> together: byte, short, boolean and int all become int in bytecode.
>> Soot therefore includes a special range analysis that attempts to
>> re-construct the best, i.e. most narrow type possible. Sometimes this
>> analysis may be too good and give you a type that is more narrow than
>> the one that was in the original source code.
>>
>> Would it be possible for you to analyze the program's source code,
>> instead of the bytecode? I am reasonably sure that in this case Soot
>> will just use the exact types that it finds in the source.
>>
>> Eric
>>
>> P.S. Sorry but I know nothing about the AST metrics.
>>
>> --
>> Dr. Eric Bodden
>> Software Technology Group, Technische Universität Darmstadt, Germany
>> Tel: +49 6151 16-5478    Fax: +49 6151 16-5410
>> Mailing Address: S2|02 A209, Hochschulstraße 10, 64289 Darmstadt
>>
>>
>>
>> On 23 June 2010 10:03, Jochen Wuttke <wuttkej at usi.ch> wrote:
>>> Hi,
>>>
>>> I'm trying to build a very simple analysis that effectively counts the
>>> number and type of local variables occurring in a method.
>>> I tried to use the normal ExceptionalUnitGraph to extract this
>>> information. However, when I run the analysis on the following method,
>>> the variable 'x' is reported to have type 'byte', not 'int'. Guessing
>>> from the documentation of some of the phase options to Soot, it looks
>>> like the standard type assigner does not care about the declared type
>>> of locals, but tries to infer the minimum size the actually values fit
>>> in. So in my case, byte rather than int, because I only assign a small
>>> numeric value.
>>>
>>> After digging a little in the API of Soot I saw that there is a
>>> package soot.tootlkits.astmetrics that seems to contain a bunch of
>>> things working on the AST provided by Polyglot. Is there a way that I
>>> can hook up my own AST metric and run it together with the others? I
>>> guess that could solve my problem?
>>>
>>> Thanks for any advice,
>>> Jochen
>>>
>>>
>>>        public void hasOnlyPrimitives() {
>>>                int x = 9;
>>>                long y = 0;
>>>                x++;
>>>                y += x;
>>>        }
>>>
>>> _______________________________________________
>>> Soot-list mailing list
>>> Soot-list at sable.mcgill.ca
>>> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>>>
>> _______________________________________________
>> Soot-list mailing list
>> Soot-list at sable.mcgill.ca
>> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>>
> _______________________________________________
> Soot-list mailing list
> Soot-list at sable.mcgill.ca
> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>