On 31 Aug 2010, at 14:54, Marc Escalier wrote:
>> Can someone direct as to how to locate/fix the problem?
>
> Hi,
>
> i would suggest that you try to find the elementary instruction that crashes, using a dichotomical printing for example (or gdb if you know how to use it)
>
I will give a try to gdb for the moment.
> often, it is due to an instruction of something like :
>
> myobject->myparameter
> for which myobject is 0, so that it is forbidden to do it (something like an object with no memory allocation before)
>
> the fact that when you remove events, the crash occurs to another place is strange anyway, and would point to an another direction, as for example forgetting to free memory
>
yes, it appears it is the case..
> good luck
>
Thanx! I think I will need a bit of it.
> ===========
> On Tue, 31 Aug 2010, Alexandros Attikis wrote:
>
>> Hi ROOTers,
>> I have a used "makeClass" to access an ntuple and manipulate it using C++ code and ROOT. I compile
>> the code with g++ and
>> have run on multiple samples including millions of events.
>> Now, when trying to run on a specific QCD ntuple, I get a segmentation fault after ~140k events,
>> always at the same point if a start from zero.
>> I have tried to start 10 events before it and and runs smoothly up to some other event later on where
>> it crashes with the same error. Obviously it is not
>> the specific event that is causing the seg. fault. I have also had a look at the ROOT files (2) and
>> they seem okay.
>> I have used cout to pinpoint the error to right before I access the specific entry of the TChain:
>> nb = fChain->GetEntry(jentry);
>> I have tried to use gdb (beginner using this) but with no success of finding something useful. The
>> Crash when running gdb gives:
>> Program received signal SIGSEV, Segmentation fault.
>> 0x00000033c627ac4a in memset () from /lib64/libc.so.6
>> Back-tracing hasn't been very useful, but as I said I am not experienced in using gdb.
>> Can someone direct as to how to locate/fix the problem?
>> Is it the aforementioned piece of code or is it possible that the problem lies somewhere else?
>> Cheers,
>> Alexandros
>>
Received on Tue Aug 31 2010 - 15:20:45 CEST