migrating from CentOS7 to Rocky9 / rank mismatch / more actual than formal arguments

[gsal]

So, we lost a developer who wrote a program using CGNS 3.3.1, while in CentOS 7; now, we are moving to Rocky 9, I want to re-compile (and deploy) in Rocky9 both CGNS and the program.

Attempt 1:
Compile CGNS 3.3.1 in Rocky 9...unsuccesful; I get a couple of the following errors:

gfortran -Wno-c-binding-type  -I.  -DBUILD_HDF5 -o lib/cgns_f.o -c cgns_f.F90
cgns_f.F90:4964:54:

 4964 |     CALL cgio_get_dimensions_f_c(cgio_num, id, ndims, dims, ier)
      |                                                      1
......
 4976 |     CALL cgio_get_dimensions_f_c(cgio_num, id, ndims, dims, ier)
      |                                                      2
Error: Rank mismatch between actual argument at (1) and actual argument at (2) (rank-1 and scalar)

Attempt 2:
Compile CGNS 3.4.2 in Rocky 9 ...succesful
Compile program...unsuccesful, I get about a dozen of these error messages:

m2_cgns.f90:280:71:

  280 |   CALL cg_goto_f(file_index,base_index,ier,'FlowEquationSet_t',1,'end')
      |                                                                       1
Error: More actual than formal arguments in procedure call at (1)
m2_cgns.f90:284:86:

  284 |   CALL cg_goto_f(file_index,base_index,ier,'FlowEquationSet_t',1,'GasModel_t',1,'end')
      |                                                                                      1
Error: More actual than formal arguments in procedure call at (1)
m2_cgns.f90:291:71:

  291 |   CALL cg_goto_f(file_index,base_index,ier,'FlowEquationSet_t',1,'end')
      |                                                                       1
Error: More actual than formal arguments in procedure call at (1)

Any hints?

[brtnfld]

This has been fixed in the new release of CGNS, and you need to add the Fortran compiler flag -fallow-argument-mismatch On Oct 1, 2023, at 7:37 PM, gsal ***@***.***> wrote: So, we lost a developer who wrote a program using CGNS 3.3.1, while in CentOS 7; now, we are moving to Rocky 9, I want to re-compile (and deploy) in Rocky9 both CGNS and the program. Attempt 1: Compile CGNS 3.3.1 in Rocky 9...unsuccesful; I get a couple of the following errors: gfortran -Wno-c-binding-type -I. -DBUILD_HDF5 -o lib/cgns_f.o -c cgns_f.F90 cgns_f.F90:4964:54: 4964 | CALL cgio_get_dimensions_f_c(cgio_num, id, ndims, dims, ier) | 1 ...... 4976 | CALL cgio_get_dimensions_f_c(cgio_num, id, ndims, dims, ier) | 2 Error: Rank mismatch between actual argument at (1) and actual argument at (2) (rank-1 and scalar) Attempt 2: Compile CGNS 3.4.2 in Rocky 9 ...succesful Compile program...unsuccesful, I get about a dozen of these error messages: m2_cgns.f90:280:71: 280 | CALL cg_goto_f(file_index,base_index,ier,'FlowEquationSet_t',1,'end') | 1 Error: More actual than formal arguments in procedure call at (1) m2_cgns.f90:284:86: 284 | CALL cg_goto_f(file_index,base_index,ier,'FlowEquationSet_t',1,'GasModel_t',1,'end') | 1 Error: More actual than formal arguments in procedure call at (1) m2_cgns.f90:291:71: 291 | CALL cg_goto_f(file_index,base_index,ier,'FlowEquationSet_t',1,'end') | 1 Error: More actual than formal arguments in procedure call at (1) Any hints? — Reply to this email directly, view it on GitHub<#735>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ACB3LB6V4LKWOETAU2GND2LX5IEEBANCNFSM6AAAAAA5OWMIH4>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[gsal]

fixed in the new release? As in 4.3.0?
I fear that if I jump from using 3.3.1 to 4.3.0, there will for sure be some compatibility issues that I may not become aware of and/or if I do, won't know how to fix.....that's why I was trying to stay within versions 3.y.z

Anyway, will try the fortran flag and see how far I get. I can also try the latest release.

[gsal]

When attempting to buil CGNS 3.3.1 in rocky9, that fortran flag helped get pass the rank mismatch problem; but, then, I run into a couple of other problems, one being mattherr that seems went obsolete in "recent" GCC versions...and rocky9 comes with GCC 11.

Any ideas on how to get passed the

Error: More actual than formal arguments in procedure call at (1)

I am getting this when attempting to compile the one program, even when linking against CGNS 4.4.0

I can see in cgns_f.F90 that cg_goto_f() is suposed to handle up to 20 UserDataNames; but, somehow, when my program call the subroutine, it can't take even 1 !

[brtnfld]

There should not be an issue with upgrading to 4.3.0, which has numerous benefits over 3.3.1. I compile CGNS with gcc-12 with no problems. I'm not sure what mattherr is.

What line is it complaining about? Do you know if the CGNS test all passed on your installed version? Do you have a simple reproducer?

[gsal]

Typo...it is not mattherr, it's just matherr. I run into the issued referred to here and here.

Anyway, I fixed the problem in the program when linking CGNS 4.4.0.

I described the problem with the in-house program, above...every time cg_goto_f was called, I would get a compile error saying "too many arguments". In comparing the sources between CGNS 3.3.1 and 4.4.0, I notice a difference in some explicit interface for cg_goto_f, so, I ventured into the in-house program sources and added USE CGNS cg_goto_f ....this solved my problem.

Thanks for your assistance, I helped to hear that I can safely jumped to CGNS 4.y