Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ptp-user] parallel debug STL Container

The debugger has code that tries to parse the type information about a variable and convert it into an internal format (called AIF). This format is the used to display the variable (and type information) in the debugger. AIF has support for C++ classes, but it may be there is something in an STL container that is not getting parsed correctly and is causing the segmentation fault. There is not any easy way to fix this, other than patch the debugger code that is generating the format information. If you can send a small test program that demonstrates the problem, I may be able to take a look at it.

Regards,
Greg

On Feb 19, 2014, at 1:59 AM, Marco Cisternino <marco.cisternino@xxxxxxxxxx> wrote:

Hi to all,
I'm new on the list and I'm a PTP newbie!!
I would like to debug my parallel c++ code in PTP. The matter is: STL Container!!
I hope someone can help me, I really don't know what to do.
I'm able to inspect STL Container in serial codes using eclipse and gdb.
But I'm going crazy with them in parallel.
My parallel debug configuration:
Open MPI-Generic-Interactive
Connection Type: local (?????)
Number of processes: 1
Debugger: SDM
Debugger backend: gdb-mi
Use built-in SDM ....
All the rest is set at default values.

.gdbinit looks like
python
import sys
sys.path.insert(0, '/home/marco/gdb_printers/python')
from libstdcxx.v6.printers import register_libstdcxx_printers
register_libstdcxx_printers (None)
end


If I run the parallel debug session everything works fine, until I select Variables view and stl container are in it. If so, things get ruined:
- stl container have no value
- not all stl container are in Variables view
- and waiting for some seconds you get a python error like this and a Segmentation Fault stopping mpirun, like this:
  close failed in file object destructor:
  sys.excepthook is missing
  lost sys.stderr
  -------------------------------------------------------------------------- 
  mpirun noticed that process rank 0 with PID 21677 on node lello exited on signal 11 (Segmentation fault).
- from this point on a weird error is continuously produced:
  Error on tasks: 0 - Partial _expression_ evaluation error!
  Reason: Time out for this request.
- if I close Parallel Debugger Perspective, Eclipse exit unexpectedly. 

I think it is obvious that not only a C code works perfectly in parallel debug with the same configuration, but I can use command line (parallel with xterm) gbd with no problem.
Is there a way to debug c++ parallel code in eclipse, inspecting STL container?

Thanks a lot for any suggestion.
Best regards,

Marco Cisternino


_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ptp-user


Back to the top