[mpich-discuss] building LAMMPS against MPICH 2.0-1.4.1.p1 on intel mac 10.6.8

[Dave Goodell]

This is basically just numerology at this point, but the address it is complaining about is suspiciously similar to MPICH2's constant value for MPI_COMM_WORLD.  MPI_COMM_WORLD==0x44000000, and your value is that plus 0xb0 (176 decimal).

Take a look at the "LAMMPS_NS::Universe::Universe(LAMMPS_NS::LAMMPS*, int)" function and see if there's any obvious mishandling of MPI_COMM_WORLD or a communicator handle value anywhere nearby, or possibly higher up the call stack.  The +176 offset is probably an attempt to access a member of a structure or class but the communicator handle value has erroneously been used as a pointer value for that struct/class.

-Dave

On Sep 12, 2011, at 11:25 PM CDT, Bruce Allen wrote:

> 
> 
> I configured, make and installed  MPICH 2.0-1.4.1p1 from sources on an Intel mac 10.6.8. 629 of the 630 tests for "make testing"
> are successful. The bcastlength test fails. The attached make file built based on the mac_mpi. The executable lmp_mi_mpi
> is created with no unresolved references. 
> 
> The following stack dump is generated when the executable is executed.
> 
> [Profmembrane:10900] *** Process received signal ***
> [Profmembrane:10900] Signal: Segmentation fault (11)
> [Profmembrane:10900] Signal code: Address not mapped (1)
> [Profmembrane:10900] Failing at address: 0x440000b0
> [Profmembrane:10900] [ 0] 2   libSystem.B.dylib                   0x00007fff8284d1ba _sigtramp + 26
> [Profmembrane:10900] [ 1] 3   ???                                 0x00007fff5fbff388 0x0 + 140734799803272
> [Profmembrane:10900] [ 2] 4   lmp_mi_mpi                          0x00000001001fac9a _ZN9LAMMPS_NS8UniverseC2EPNS_6LAMMPSEi + 218
> [Profmembrane:10900] [ 3] 5   lmp_mi_mpi                          0x00000001001fad19 _ZN9LAMMPS_NS8UniverseC1EPNS_6LAMMPSEi + 9
> [Profmembrane:10900] [ 4] 6   lmp_mi_mpi                          0x000000010011b8af _ZN9LAMMPS_NS6LAMMPSC2EiPPci + 173
> [Profmembrane:10900] [ 5] 7   lmp_mi_mpi                          0x000000010011c805 _ZN9LAMMPS_NS6LAMMPSC1EiPPci + 9
> [Profmembrane:10900] [ 6] 8   lmp_mi_mpi                          0x000000010011f3da main + 64
> [Profmembrane:10900] [ 7] 9   lmp_mi_mpi                          0x0000000100001344 start + 52
> [Profmembrane:10900] [ 8] 10  ???                                 0x0000000000000001 0x0 + 1
> [Profmembrane:10900] *** End of error message ***
> Segmentation fault
> Profmembrane:src claudevms$ cd make
> Profmembrane:make claudevms$ make mi_mpi
> make: *** No rule to make target `mi_mpi'.  Stop.
> Profmembrane:make claudevms$ xedit Makefile.mi_mpi &
> [4] 10966
> 
> Anyone who can point me to additional information on building MPICH 2.0 and LAMMPS I would appreciate it.
> 
> Thanks in advance.
>  
> 
> _______________________________________________
> mpich-discuss mailing list     mpich-discuss at mcs.anl.gov
> To manage subscription options or unsubscribe:
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss