[Soot-list] Paddle analysis crashes in a buddy native frame

[Benoit Claudel]

I'll check the hypothesis of an out of memory error and you're right.
However, when I am running the example provided with the Survivor
guide I have 2.5GB of memory available for buddy allocations.
It seems that the analysis scans a lot of classes from the rt.jar not  
used by
the Test1 application. Is there a mean to limit the scope of analyzed  
classes?

Thanks so much,

Benoit.

Le 11 nov. 09 à 19:39, Saswat Anand a écrit :

> This usually happens when Buddy (the BDD library) runs out of memory.
> The out of memory error from Buddy is not properly acted upon in Jedd,
> and leads to crash.
>
>
> Saswat
>
> On Wed, Nov 11, 2009 at 10:01 AM, Benoit Claudel
> <benoit.claudel at inrialpes.fr> wrote:
>> Hello,
>>
>> I am playing with the PointsToAnalysis provided with the Survivor  
>> guide.
>>
>> I am using Paddle with the "buddy" backend. When I am lauching a 1cfa
>> analysis with "context-heap" option set to false, everything works  
>> fine.
>> However, if I enable "context-heap" option or use kcfa analysis (with
>> k > 1),
>> the virtual machine crashes in a native frame.
>> I already unsuccessfully tried to increase thread stack size.
>>
>> I have enclosed a log with more details on my configuration and the
>> crash.
>>
>> Does someone already had the same problem?
>>
>> Thanks so much.
>> Benoit.
>> PhD Candidate at INRIA (France).
>>
>>
>> _______________________________________________
>> Soot-list mailing list
>> Soot-list at sable.mcgill.ca
>> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>>