[mpich-discuss] long message testing in MPICH test suite

[Kenneth Raffenetti]

Just closing the loop on this issue. Through side-channel discussions, 
we modified the patch and included it in the latest mpich/master branch.

Ken

On 10/30/2015 12:22 PM, Rusty Lusk wrote:
> In a recent exchange with the OSU folks, it was discovered that there is
> a gap in the MPICH test suite in that the code that tests for the
> correctness of handling long messages forgets to test that count is
> handled correctly.  This led to the escape of a bug in the MVAPICH code,
> since corrected.  Hari tracked this down and has sent a patch.  Could
> the MPICH team look at this?
>
> Thanks,
> Rusty
>
> Begin forwarded message:
>
>> *From: *Hari Subramoni <subramoni.1 at osu.edu <mailto:subramoni.1 at osu.edu>>
>> *Subject: **Re: [lcrc-support #279383] Re: Long messages*
>> *Date: *October 26, 2015 at 12:19:42 PM CDT
>> *To: *Rusty Lusk <lusk at mcs.anl.gov <mailto:lusk at mcs.anl.gov>>
>> *Cc: *"Low, John J." <jlow at mcs.anl.gov <mailto:jlow at mcs.anl.gov>>,
>> Steve Pieper <spieper at anl.gov <mailto:spieper at anl.gov>>, "Wiringa,
>> Robert B." <wiringa at anl.gov <mailto:wiringa at anl.gov>>, "D.K. Panda"
>> <panda at cse.ohio-state.edu <mailto:panda at cse.ohio-state.edu>>,
>> "support at lcrc.anl.gov <mailto:support at lcrc.anl.gov>"
>> <support at lcrc.anl.gov <mailto:support at lcrc.anl.gov>>, "Balaji, Pavan"
>> <balaji at anl.gov <mailto:balaji at anl.gov>>
>>
>> Hello All,
>>
>> I was investigating to see how this was missed earlier. In our MPICH
>> large-message test (test/mpi/pt2pt/large_message.c) which should've
>> caught this, the code which tests the result of MPI_Get_count was missing.
>>
>> I was checking the large_message.c test that is shipped with
>> MPICH-3.1.4 (latest stable) and MPICH-3.2rc1 (preview release)
>> available from the MPICH site. Both have the same issue. The
>> large-message-test.patch should fix this.
>>
>> If the patch looks good, it would be good if this gets applied to
>> MPICH itself so that all MPICH derivatives can benefit from it.
>>
>> Best Regards,
>> Hari.
>>
>
_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss