[mpich2-dev] nesting level warning with MPI_File_preallocate(fh, 0)

[Lisandro Dalcin]

Thanks, Rajeev. I'm not using 1.1 yet for production, just wanted to
point the issue.

BTW, here you have another problem I'm experiencing. I get the
following warning (see last line). Sorry again for not diving into the
sources and providing the patch. I'm really busy with the next
PETSc/SLEPc releases, and my personal mpi&python-related projects.


$ cat testfile.c
#include <mpi.h>
int main( int argc, char ** argv ) {
  MPI_File fh;
  int amode = MPI_MODE_RDWR | MPI_MODE_CREATE | MPI_MODE_DELETE_ON_CLOSE;
  MPI_Init(&argc, &argv);
  MPI_File_open(MPI_COMM_WORLD, "/tmp/datafile",
		amode, MPI_INFO_NULL, &fh);
  MPI_File_close(&fh);
  MPI_Finalize();
  return 0;
}

$ mpicc testfile.c

$ ./a.out
In direct memory block for handle type ATTRIBUTE KEY, 2 handles are
still allocated





On Tue, Dec 9, 2008 at 6:43 PM, Rajeev Thakur <thakur at mcs.anl.gov> wrote:
> In src/mpi/romio/mpi-io/prealloc.c, change line 77
>      if (size == 0) return MPI_SUCCESS;
> to
>      if (size == 0) goto fn_exit;
>
> That should fix it.
>
> Rajeev
>
>
>> -----Original Message-----
>> From: mpich2-dev-bounces at mcs.anl.gov
>> [mailto:mpich2-dev-bounces at mcs.anl.gov] On Behalf Of Lisandro Dalcin
>> Sent: Tuesday, December 09, 2008 11:28 AM
>> To: mpich2-dev at mcs.anl.gov
>> Subject: [mpich2-dev] nesting level warning with
>> MPI_File_preallocate(fh, 0)
>>
>> Consider the following Python snipet (sorry, really busy to write
>> C/C++, I used mpi4py)
>>
>> from mpi4py import MPI
>>
>> amode = MPI.MODE_RDWR | MPI.MODE_CREATE | MPI.MODE_DELETE_ON_CLOSE
>>
>> fh = MPI.File.Open(MPI.COMM_WORLD,
>>                    '/tmp/datafile', amode,
>>                    MPI.INFO_NULL)
>>
>> N = 0
>> fh.Preallocate(N)
>> size = fh.Get_size()
>> assert size == N
>>
>> fh.Close()
>> print 'Bye!!!'
>>
>>
>> Then, when I run the code (1.0.8 and --enable-g=all), I get
>> the following output
>>
>> Bye!!!
>> Unexpected value for nesting level = 1
>> Nest stack is:
>>       [0] :0
>>
>> So the warning is likely emitted at MPI_Finalize().
>>
>> If I try to preallocate sizes larger than 0, then the warning
>> does not appear.
>>
>>
>>
>>
>> --
>> Lisandro Dalcín
>> ---------------
>> Centro Internacional de Métodos Computacionales en Ingeniería (CIMEC)
>> Instituto de Desarrollo Tecnológico para la Industria Química (INTEC)
>> Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)
>> PTLC - Güemes 3450, (3000) Santa Fe, Argentina
>> Tel/Fax: +54-(0)342-451.1594
>>
>
>



-- 
Lisandro Dalcín
---------------
Centro Internacional de Métodos Computacionales en Ingeniería (CIMEC)
Instituto de Desarrollo Tecnológico para la Industria Química (INTEC)
Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)
PTLC - Güemes 3450, (3000) Santa Fe, Argentina
Tel/Fax: +54-(0)342-451.1594