fortran:fix integer kind=8 problem

[hppritcha]

user reported problem on the ompi-users mail list.

https://www.mail-archive.com/[email protected]//msg35382.html

[devreal]

LGTM (although my understand of the problem is limited)

[jsquyres]

@hppritcha Giles noted in 32db65f#commitcomment-154262549 that this might not be right -- the implicit string length argument is passed as an int, not an MPI_Fint.

Can you dig a little deeper and see if this is actually correct?

(oops -- should have posted this on #13160)

[ggouaillardet]

FWIW, I would reverse that commit and fix it like that instead

diff --git a/ompi/mpi/fortran/use-mpi-f08/session_get_nth_pset_f08.F90 b/ompi/mpi/fortran/use-mpi-f08/session_get_nth_pset_f08.F90
index c34ed8d008..65dffc8852 100644
--- a/ompi/mpi/fortran/use-mpi-f08/session_get_nth_pset_f08.F90
+++ b/ompi/mpi/fortran/use-mpi-f08/session_get_nth_pset_f08.F90
@@ -14,6 +14,7 @@
 subroutine MPI_Session_get_nth_pset_f08(session, info, n, pset_len, pset_name, ierror)
    use :: mpi_f08_types, only : MPI_Session, MPI_Info, MPI_INFO_NULL
    use :: ompi_mpifh_bindings, only : ompi_session_get_nth_pset_f
+   use, intrinsic :: ISO_C_BINDING, only : C_INT
    implicit none
    TYPE(MPI_Session), INTENT(IN) :: session
    TYPE(MPI_Info), INTENT(IN) :: info
@@ -21,10 +22,12 @@ subroutine MPI_Session_get_nth_pset_f08(session, info, n, pset_len, pset_name, i
    INTEGER, OPTIONAL, INTENT(INOUT) :: pset_len
    CHARACTER(LEN=*), INTENT(OUT) :: pset_name
    INTEGER, OPTIONAL, INTENT(OUT) :: ierror
+   INTEGER(KIND=C_INT) :: pset_name_len
    integer :: c_ierror
 
+   pset_name_len = len(pset_name)
    call ompi_session_get_nth_pset_f(session%MPI_VAL, MPI_INFO_NULL%MPI_VAL, n,     &
-                                    pset_len, pset_name, c_ierror, len(pset_name))
+                                    pset_len, pset_name, c_ierror, pset_name_len)
    if (present(ierror)) ierror = c_ierror
 
 end subroutine MPI_Session_get_nth_pset_f08

[hppritcha]

actually there a number of *.F90 files that do what the original commit did. check how ompi_comm_set_name_f and similar are used in the use-mpi-f08 folder. Not saying they're correct but that explains why the user is able to build with default KIND=8 for older releases but compilation fails. Anyway I'll use a different approach which is nicer (no new variable introduction).
I suppose there are a number of files which really need to be updated. I'll include them in a new PR.

[hppritcha]

according to these discussions these files are incorrect along with the prototypes in bindings/mpi-f-interfaces-bind.h

add_error_string_f08.F90:   call ompi_add_error_string_f(errorcode, string, c_ierror, len(string))
close_port_f08.F90:   call ompi_close_port_f(port_name,c_ierror,len(port_name))
comm_accept_f08.F90:                           c_ierror,len(port_name))
comm_connect_f08.F90:                            c_ierror,len(port_name))
comm_create_from_group_f08.F90:                                      newcomm%MPI_VAL, c_ierror, len(stringtag))
comm_get_name_f08.F90:                             len(comm_name))
comm_set_name_f08.F90:   call ompi_comm_set_name_f(comm%MPI_VAL,comm_name,c_ierror,len(comm_name))
comm_spawn_f08.F90:                          len(command), len(argv))
comm_spawn_multiple_f08.F90:                                   len(array_of_commands), len(array_of_argv))
error_string_f08.F90:   call ompi_error_string_f(errorcode,string,resultlen,c_ierror,len(string))
file_delete_f08.F90:   call ompi_file_delete_f(filename,info%MPI_VAL,c_ierror,len(filename))
file_get_view_f08.F90:                             datarep,c_ierror,len(datarep))
file_open_f08.F90:                         c_ierror, len(filename))
file_set_view_f08.F90:                             datarep,info%MPI_VAL,c_ierror,len(datarep))
get_library_version_f08.F90:   call ompi_get_library_version_f(version,resultlen,c_ierror,len(version))
get_processor_name_f08.F90:   call ompi_get_processor_name_f(name,resultlen,c_ierror,len(name))
group_from_session_pset_f08.F90:   call ompi_group_from_session_pset_f(session%MPI_VAL, pset_name, newgroup%MPI_VAL, c_ierror, len(pset_name))
info_delete_f08.F90:   call ompi_info_delete_f(info%MPI_VAL,key,c_ierror,len(key))
info_get_nthkey_f08.F90:   call ompi_info_get_nthkey_f(info%MPI_VAL,n,key,c_ierror,len(key))
info_get_valuelen_f08.F90:   call PMPI_Info_get_valuelen(info%MPI_VAL,key,valuelen,flag,c_ierror)
info_set_f08.F90:   call ompi_info_set_f(info%MPI_VAL,key,value,c_ierror,len(key),len(value))
intercomm_create_from_groups_f08.F90:                                            newintercomm%MPI_VAL, c_ierror, len(stringtag))
lookup_name_f08.F90:                           len(service_name),len(port_name))
open_port_f08.F90:   call ompi_open_port_f(info%MPI_VAL,port_name,c_ierror,len(port_name))
pack_external_f08.F90:                             outsize,position,c_ierror,len(datarep))
pack_external_size_f08.F90:   call ompi_pack_external_size_f(datarep,incount,datatype%MPI_VAL,size,c_ierror,len(datarep))
publish_name_f08.F90:                            len(service_name), len(port_name))
register_datarep_f08.F90:                                fdtype_fn,extra_state,c_ierror,len(datarep))
session_get_nth_pset_f08.F90:                                    len(pset_name,KIND=C_INT))
session_get_pset_info_f08.F90:   call ompi_session_get_pset_info_f(session%MPI_VAL, pset_name, info%MPI_VAL, c_ierror, len(pset_name))
type_get_name_f08.F90:   call ompi_type_get_name_f(datatype%MPI_VAL,type_name,resultlen,c_ierror,len(type_name))
type_set_name_f08.F90:   call ompi_type_set_name_f(datatype%MPI_VAL,type_name,c_ierror,len(type_name))
unpack_external_f08.F90:                               outcount,datatype%MPI_VAL,c_ierror,len(datarep))
unpublish_name_f08.F90:                              len(service_name), len(port_name))
win_get_name_f08.F90:   call ompi_win_get_name_f(win%MPI_VAL,win_name,resultlen,c_ierror,len(win_name))
win_set_name_f08.F90:   call ompi_win_set_name_f(win%MPI_VAL,win_name,c_ierror,len(win_name))

@jsquyres

[jsquyres]

Yikes. Thanks for tracking this down, @hppritcha!