Charm++ parallel programming system

[Ryan Mokos <ryan.mokos AT gmail.com>]

Hi Xuehan,

How did you modify the code, and what were you trying to accomplish by modifying it?  The only time you should need to write/modify code is if you're adding another network model.

It's hard to say exactly what the problem is without knowing what you changed.  The only guess I have is that you are manually deleting a message received by a poser in one of its entry methods.  This is not necessary as POSE automatically takes care of this.

When you compile a POSE program (such as BigNetSim), first, a source-to-source translation is performed on the code by a pearl script, etrans.pl.  This is what generates the*_sim.h and *_sim.C files that appear.  This translation converts POSE macros (such as POSE_invoke) to charm code.  So when you run the program, you're basically running a program built on POSE, which is built on charm, which is built on converse.  All the code written for a poser ends up in functions that are called by the POSE framework.  If something is breaking in POSE, then either there's a bug in POSE, or BigNetSim is doing something to the state of a poser that causes POSE to break.

Ryan

On Sat, Jun 25, 2011 at 8:34 AM, Xuehan Xu <xxhdx1985126 AT gmail.com> wrote:

Dear sir:
        I modified the source code of BigNetSim to do some new simulation. But I got a segmentation fault. I run the BigNetSim like this:

        ../tmp/charmrun +p1 ../tmp/bigemulator 1 0 ++remote-shell ssh

        I debugged the program, it seems that the segmentation fault occurs when the deconstructor of class Event attempt to "delete msg", which is the line 19 in event.C.

        What kind of error can this indicate in my modification?

        And does pose have any debugging facility? I'm asking this because the debugger I'm using now debugs the whole converse program, which often shows a error within the pose code and can not point out what's wrong in my own code directly.

        Please help me. Thank you very much:-)