a bug ? ( repost )

Wei Ku wku1@utk.edu
Thu Jul 9 01:13:00 GMT 1998


Hello, all :

I repost the message since I found no reply to the original post.
Any reply is appreciated very much.
Is this a bug ? The following two expressions produced different result at
runtime when compiled successfully with egcs-1.0.3:
[
    OOMPI_Port master = OOMPI_COMM_WORLD[0];
    master >> offset >> rows
           >> OOMPI_Message( a[ 0], rows * NCA)
           >> OOMPI_Message( b[ 0], NRB * NCB);
] ( bogus )

[
    OOMPI_Port master = OOMPI_COMM_WORLD[0];
    master >> offset >> rows;
    master >> OOMPI_Message( a[ 0], rows * NCA);
    master >> OOMPI_Message( b[ 0], NRB * NCB);
] ( correct )

where the OOMPI_Port is declaired in "Port.h" :
[
class OOMPI_Port {
friend OOMPI_Comm;
friend OOMPI_Intra_comm;

public:
    .
    .
    OOMPI_Port &operator>>(OOMPI_Message buf);
    .
    .
}
]

and defined in "Port.cc" :
[
OOMPI_Port &
OOMPI_Port::operator>>(OOMPI_Message buf)
{
  MPI_Status mpi_status;
  MPI_Recv(buf.Get_top(), buf.Get_count(), buf.Get_type(), my_rank,
           buf.Get_tag(), comm_wrapper->Get(), &mpi_status);

  return *this;
}
]

Thank you for the help.

Wei Ku

***************************************
Department of Physics and Astronomy
The University of Tennessee
wku1@utk.edu
---------------------------------------
Solid State Division
Oak Ridge National Laboratory
weiku@solid.ssd.ornl.gov
***************************************

    




More information about the Gcc-bugs mailing list