Hi Joseph,

Thanks for reporting this issue and the reproducer program. I just confirmed this is a bug in MPICH request-based RMA routine. I will fix it soon and add your reproducer into our test suite.

You can check the status through the ticket: https://github.com/pmodels/mpich/issues/2880

Thanks,
Min

On 11/28/17 5:48 AM, Joseph Schuchart wrote:

All,

I am running into a problem with MPI_Rget in combination with MPI_Testall, where the call to MPI_Testall fails with the error hinting at more information in the status' MPI_ERROR fields. However, these fields only signal success:

[0] MPI_Testall failed:
    See the MPI_ERROR field in MPI_Status for the error code (17)
[0]    Request error: No MPI error (0)
[0]    Request error: No MPI error (0)

I am attaching a small reproducer. I should note that the application succeeds if I am using MPI_Test on the individual requests instead of MPI_Testall.

Please let me know if my usage of MPI_Rget and MPI_Testall is erroneous for some reason. I can reproduce this error with both MPICH 3.2 and 3.2.1 as well as Intel MPI 2017 even when running with a single process.

Many thanks in advance,
Joseph
_______________________________________________
discuss mailing list     discuss@mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss