[Soot-list] NullPointerException in soot.SootMethod.getBodyFromMethodSource

[Marc-Andre Laverdiere-Papineau]

Please ignore that last email.

I somehow had still in mind Patrick's suggestion of exiting the VM.

I've implemented what Eric suggested, and I have a memory explosion, and 
I'm running with 2GB given to the VM process. I'll try giving a bit 
more, but my poor system doesn't have too much RAM to spare :(

On 07/04/2012 08:28 AM, Marc-Andre Laverdiere-Papineau wrote:
> Hello,
>
> I can get going along those lines. I only have one big big piece missing
> in the puzzle.
>
> How do I set the entry points in the second phase?
> Even if I store them in a text file, I need to have the SootMethod
> objects. To get those, I need to have the classpaths and process dirs
> set, and to use loadNecessaryClasses(), etc. Won't that give me problems
> later on?
> Unless there is a special command-line option I missed.
>
> As an alternative implementation, is there an easy-ish way of extending
> the entry point detection algorithm for my use case?
>
> On 07/04/2012 03:59 AM, Eric Bodden wrote:
>> Hello.
>>
>> Ok, I see your use case now. However, you have to keep in mind the following:
>> At the time you search for entry points, you have not yet conducted a
>> whole-program analysis. Hence, at this point it is unclear what your
>> whole program actually is. At this time, it does not make much sense
>> to ask the hierarchy for subclasses because the hierarchy does not yet
>> know the universe of all classes to be considered. This is kind of a
>> chicken/egg problem. Here is a possible other solution that I can see:
>>
>> Use process-dir to process all classes in your jar(s). At this time,
>> scann all classes, one by one, checking if the class is a subclass of
>> javax.servlet.http.HttpServlet and contains one of the methods you are
>> looking for. Do not use the hierarchy. Store a table of those methods.
>>
>> Then reset Soot, and now run it in whole-program mode. As entry
>> points, choose the methods computed above.
>>
>> I think this should work. If not, let us know.
>>
>> Eric
>>
>> On 3 July 2012 18:11, Marc-Andre Laverdiere-Papineau
>> <marc-andre.laverdiere-papineau at polymtl.ca> wrote:
>>> Hello,
>>>
>>> I work with servlet entry points. I find all classes that are subclasses
>>> of HttpServlet and then the doGet, etc. methods
>>>
>>> Code is like this:
>>>
>>> final SootClass servletClass =
>>> Scene.v().loadClassAndSupport("javax.servlet.http.HttpServlet");
>>>
>>> FastHierarchy fh = Scene.v().getOrMakeFastHierarchy();
>>> final String[] servletMethods = new String[]{
>>> //HTTP Servlet
>>> "doGet", "doPost", "doPut", "doDelete", "init", "destroy",
>>> "getServletInfo",
>>> //From JspC
>>> "_jspInit", "_jspDestroy", "_jspService"
>>> };
>>> for (SootClass sc : Scene.v().getApplicationClasses()){
>>>           if (!sc.isInterface() && fh.isSubclass(sc, servletClass)){
>>>       logger.debug("Servlet detected: {}", sc.toString());
>>>           for(String method : servletMethods){
>>>           if (sc.declaresMethodByName(method))
>>>                   entryPoints.add(sc.getMethodByName(method));
>>> }
>>> }
>>> }
>>>
>>> On 2012-07-03 11:51, Eric Bodden wrote:
>>>> Hi all.
>>>>
>>>> G.reset() resets all static state. But that's not Marc-Andre's
>>>> problem. His problem is that Soot also releases all the method bodies
>>>> after they were written out. Calling G.reset() does not undo this
>>>> "garbage collection".
>>>>
>>>> Marc: How do you compute entry points, i.e., based on which conditions?
>>>>
>>>> Eric
>>>>
>>>> On 3 July 2012 16:45, Patrick Lam <p.lam at ece.uwaterloo.ca> wrote:
>>>>> On 07/03/2012 10:20 AM, Marc-Andre Laverdiere-Papineau wrote:
>>>>>> So what would be the proper way of picking my entry points then?
>>>>>
>>>>> G.reset() ought to work, but maybe it doesn't. You could run Soot once,
>>>>> save the results, quit the JVM, and run Soot again, loading the results.
>>>>> At least that will tell you if it works or not.
>>>>>
>>>>> pat
>>>>> _______________________________________________
>>>>> Soot-list mailing list
>>>>> Soot-list at sable.mcgill.ca
>>>>> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>>>>
>>>>
>>>>
>>>
>>>
>>> --
>>> Marc-André Laverdière-Papineau
>>> Étudiant au doctorat - PhD Student
>>>
>>>
>>> _______________________________________________
>>> Soot-list mailing list
>>> Soot-list at sable.mcgill.ca
>>> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>>
>>
>>
>
>


-- 
Marc-André Laverdière-Papineau
Étudiant au doctorat - PhD Student