Possible problem in persistent requests

[gropp]

We routinely fail one of the Fortran tests in the Intel suite when using the Intel compiler - the fortran/persist_request/functional/MPI_Send_init_overtake test. There is a C test with the same name, but it turns out that the Fortran test creates and starts a bunch of requests and then uses MPI_Waitall while the C test creates a request, starts it, and then waits on it.

Now, there is a character variable that is too short in the Fortran version, which I'll fix. It would be good, however, to have a C version of this Fortran test that used the Waitall.

[Rajeev Thakur]

I believe the two tests fail with the current version of the Intel
compiler, but not the previous version. We have been ignoring the
failure thinking it is a bug in the test or the compiler.

Rajeev


>  We routinely fail one of the Fortran tests in the Intel
> suite when using  the Intel compiler - the
> fortran/persist_request/functional/MPI_Send_init_overtake
> test.  There is  a C test with the same name, but it turns
> out that the Fortran test  creates and starts a bunch of
> requests and then uses MPI_Waitall while the  C test creates
> a request, starts it, and then waits on it.
>
>  Now, there is a character variable that is too short in the
> Fortran  version, which I'll fix.  It would be good, however,
> to have a C version  of this Fortran test that used the Waitall.

[thakur]

Bill fixed the Intel tests.