ocean: error on ierr variable with intel 17

[mark-petersen]

The ierr variable causes a build fail on CORI with intel 17 at this location.

mpas_ocn_core_interface.f90
read(unitNumber, AM_pointwiseStats, iostat=ierr)

The solution for a local fix was to simply rename ierr to a different variable name in that call.

[mark-petersen]

Here is the full text from Noel Keen, who is testing ACME for us on CORI phase II (KNL). The fix is very easy, I'm making an issue so I don't forget!

With Intevl v17 (both versions installed on Cori), I ran into an issue with building
a certain file in mpas-o.  Looking closer, it looks like the code is fine.
 
The error you will see looks like:
 
mpas_ocn_core_interface.f90(85608): error #6405: The same named entity from different modules and/or program units cannot be referenced.   [IERR]
        read(unitNumber, AM_pointwiseStats, iostat=ierr)
----------------------------------------------------^
compilation aborted for mpas_ocn_core_interface.f90 (code 1)
 
The compiler seems to be saying the variable ierr is defined in one of the modules,
but it doesn't seem like it is to me.  Also, several routine above and below do the
same thing and the compiler doesn't complain about those. 
 
Only difference is that I change the variable name!  Seems to work.
I also showed this to Intel staff (martyn.j.corden@intel.com) and he agreed that
it looked like somethign was wrong and asked if he could have a copy of that one
file (mpas_ocn_core_interface.f90).  If someone working on mpas-o has a
better work-around, please let me know.

I'll post Noel's work-around when I get it.

[mark-petersen]

The temporary solution on the KNL was to modify src/tools/registry/gen_inc.c. That may be a quick fix, I would need to look more closely to see if there is an error in the way we call the routine.

edison08> pwd
/global/u2/m/mpeterse/repos/ACME/v1.0.0-beta_161202_altered_winds/components/mpas-o/model
edison08> diff src/tools/registry/gen_inc.c /global/project/projectdirs/acme/ndk/knl-specific-issues/gen_inc.c
600c600
< 		fortprintf(fd, "      integer :: ierr\n");
---
> 		fortprintf(fd, "      integer :: ierrz\n");
650c650
< 		fortprintf(fd, "         read(unitNumber, %s, iostat=ierr)\n", nmlrecname);
---
> 		fortprintf(fd, "         read(unitNumber, %s, iostat=ierrz)\n", nmlrecname);
653,654c653,654
< 		// Broadcast ierr, to check if a broadcast should happen for the options (if namelist was read in)
< 		fortprintf(fd, "      call mpas_dmpar_bcast_int(dminfo, ierr)\n");
---
> 		// Broadcast ierrz, to check if a broadcast should happen for the options (if namelist was read in)
> 		fortprintf(fd, "      call mpas_dmpar_bcast_int(dminfo, ierrz)\n");
658c658
< 		fortprintf(fd, "      if (ierr <= 0) then\n");
---
> 		fortprintf(fd, "      if (ierrz <= 0) then\n");
673c673
< 		fortprintf(fd, "         if (ierr < 0) then\n");
---
> 		fortprintf(fd, "         if (ierrz < 0) then\n");
689c689
< 		fortprintf(fd, "      else if (ierr > 0) then\n");
---
> 		fortprintf(fd, "      else if (ierrz > 0) then\n");

[jonbob]

@mark-petersen - I think you can close this one, since the ACME PR #1257 should have taken care of it.

[mark-petersen]

Yes, solved by MPAS PR #1208 which is in ACME at:
E3SM-Project/E3SM#1257