[mpich-discuss] Strange error

[Dave Goodell]

IIRC, the MPI_ERROR field of the status structure is not used when performing an MPI_Recv.  It is only used when performing multiple completions in the same call, such as with MPI_Waitall/Testsome/etc.

Did you change the error handler for MPI_COMM_WORLD?  If not, then it defaults to MPI_ERRORS_ARE_FATAL, which will cause the MPI job to abort if an error is encountered.  If you change it to MPI_ERRORS_RETURN then you need to capture the return value from MPI_Recv.  For example:

----8<----
int mpi_errno;
…
mpi_errno = MPI_Recv(greeting,…);
if (mpi_errno != MPI_SUCCESS) {
    /* inspect mpi_errno here */
}
----8<----

-Dave

On Oct 9, 2012, at 1:36 PM CDT, Alejandro Allievi wrote:

> Thanks for the reply.
> 
> Here goes a code snippet containing the part that generates the error
> 
>               MPI_Recv(greeting, MAX_STRING, MPI_CHAR, MPI_ANY_SOURCE, MPI_ANY_TAG , MPI_COMM_WORLD, &status);       
>          
> //These are fine
>               printf("\nSource of this message is process: %d\n", status.MPI_SOURCE);
>               printf("Tag of this message is: %d\n", status.MPI_TAG);
> 
> //This is the problem. Not quite sure if correct though.
>               MPI_Error_string( status.MPI_ERROR , mystring , &len );
>               printf("Tag of this message is: %d %s\n", status.MPI_ERROR , mystring );
> 
> //These are fine too
>               MPI_Get_count(&status , MPI_CHAR , &count);
>               printf("Size of this message is: %d MPI_CHAR", count);
> 
> Alejandro
> 
> 
> On Tue, Oct 9, 2012 at 3:38 PM, Dave Goodell <goodell at mcs.anl.gov> wrote:
> Can you post some example code showing how you are calling MPI_Recv and how you are processing the error code/class/message?  It's easy to get confused by the error handling conventions of the MPI standard, and this sounds like either user error or memory corruption, depending on what code you post.
> 
> -Dave
> 
> On Oct 9, 2012, at 11:55 AM CDT, Alejandro Allievi wrote:
> 
> > Hi,
> >
> > I'm getting the following error: "No MPI error, error stack: (unknown)(): No MPI error"
> > returned by "status.MPI_ERROR" in "status" variable  in MPI_Recv. However, there seems
> > to be no error. Does anybody know why the "error stack: (unknown)(): No MPI error"
> >
> > Thanks for any help.
> >
> > Alejandro
> > _______________________________________________
> > 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
> 
> _______________________________________________
> 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
> 
> 
> 
> -- 
> Alejandro Allievi
> http://www.ace-net.ca/wiki/Alejandro_Allievi   
>  
> "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre véritable sentiment n'est pas celui dans lequel nous n'avons jamais vacillé; mais celui auquel nous sommes le plus habituellement revenus". Denis Diderot. 
>  
> Este e-mail esta consignado sólo para el destinatario designado en el mismo y puede contener información confidencial y privilegiada. Su distribución o copiado está prohibido. Si Usted no fuera el destinatario designado, por favor notifíquenos de inmediato y destruya este e-mail en forma permanente así como todas las copias del mismo. Ce courriel peut renfermer des renseignements confidentiels et privilégiés et s'adresse au destinataire désigné seulement. La distribution ou la copie de ce courriel est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en aviser immédiatement et détruire de façon permanente ce courriel ainsi que toute copie de celui-ci. This e-mail may contain confidential information, and is intended only for the named recipient and may be privileged. Distribution or copying of this email is prohibited. If you are not the named recipient, please notify us immediately and permanently destroy this email and all copies of it.
> 
> _______________________________________________
> 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