960117: NETCDF fortran errors with linux

Darin,

> Subject: netCDF fortran tests fail
> To:       address@hidden
> From: Darin Ernst <address@hidden>
> Organization: .
> Keywords: 199601171614.AA16237

In the above message you wrote:
 
> No matter what settings I use for the fortran parts, I always seem
> to see the following tests failed upon making test in netCDF.
> All the C tests are successful, and there are no compile errors.
> 
> I'm using Linux with the Promula Fortran compiler, which is a robust 
> compiler that supports VAX and Sun extensions.  I have used this
> compiler successfully on several thousand fortran modules, mixing 
> fortran and C, so generally, I know it is working.  I made up a linux.m4
> file and used the manual for the compiler to pick settings.  I'm
> appending that below the error report.  I'm using 
> 
> libc-4.6.27
> a.out format
> gcc-2.6.3
> Promula FORTRAN pfc (fortran->c translator; in script to behave as compiler)
> 
> The system successfully made and tested the HDF distribution, except for
> the netCDF parts, which give the same errors as below.  This was from
> the netCDF distribution 2.4 beta 6:
> 
> - --------------------------------------------------------/ cd fortran; make 
> test
> 
> making `test' in directory /scsi/usr/src/netcdf-2.4-beta6/src/fortran
> 
> make[1]: Entering directory `/scsi/usr/src/netcdf-2.4-beta6/src/fortran'
> ar rcuv ../libsrc/libnetcdf.a jackets.o
> ranlib ../libsrc/libnetcdf.a
> pf77 -o ftest -O2 -extend -stdio -I/scsi/usr/src/netcdf-2.4-beta6/src/fortran 
> ftest.o
> - -L../libsrc -lnetcdf
> ./ftest
> *** Testing nccre ...
> *** Testing ncddef ...
> ncdimdef: dimension "(null)" in use with index 0
> ncdimdef: dimension "(null)" in use with index 0
> ncdimdef: dimension "(null)" in use with index 0
> ncdimdef: dimension "(null)" in use with index 0
> *** Testing ncvdef ...
> ncvardef: Bad dimension id -2
> ncvardef: Invalid number of dimensions 4 > 1
> ncvardef: Invalid number of dimensions 4 > 1
> ncvardef: Invalid number of dimensions 2 > 1
> *** Testing ncapt, ncaptc ...
> ncattput: -2 is not a valid variable id
> ncattput: -2 is not a valid variable id
> ncattput: -2 is not a valid variable id
> ncattput: -2 is not a valid variable id
> ncattput: -2 is not a valid variable id
> *** Testing ncclos ...
> *** Testing ncvpt1 ...
> ncvarinq: -2 is not a valid variable id
> ncvarinq: -2 is not a valid variable id
> ncvarinq: -2 is not a valid variable id
> *** Testing ncvgt1 ...
> ncvarinq: -2 is not a valid variable id
>  error in ncvgt1 long:           0 .ne.        1000
> ncvarinq: -2 is not a valid variable id
>  error in ncvgt 1 float:  0.0000000     not close to   3.141590    
> ncvarinq: -2 is not a valid variable id
>  error in ncvgt1 double:  0.0000000000000000     not close to
>   0.3333333333000000    
> *** Testing ncvpt ...
> ncvarinq: -2 is not a valid variable id
> ncvarinq: -2 is not a valid variable id
> ncvarinq: -2 is not a valid variable id
> ncvarinq: -2 is not a valid variable id
> *** Testing ncopn, ncinq, ncdinq, ncvinq, ncanam, ncainq ...
>  error in ncinq or ncddef
>  error in ncinq or ncvdef
> make[1]: *** [test] Segmentation fault
> make[1]: Leaving directory `/scsi/usr/src/netcdf-2.4-beta6/src/fortran'
> make: *** [fortran/test] Error 1
> 
> 
> - ------------------------------------------------/ linux.m4
> divert(-1)
> 
> # Name of system platform (for use in comments)
> define(`M4__SYSTEM', 'linux PROMULA.FORTRAN')
> # Using commercial PROMULA f2c-like translator with 
> # custom pf77 script by D. Ernst
> 
> # transformation from fortran name to name of C module
> define(`NAMEF',`$1_')   # for Sun and most unixes, just append an underscore
> 
> # transformation from string name to corresponding argument name
> define(`STRINGF',`$1')
> 
> # extra arguments, if any, for string length
> define(`STRINGX',`, $1len')  # one extra stringlen parameter
> 
> # declaration to be used for argument name descriptor
> define(`STRINGD',`
>     char        *$1;    `$2'
>     int         $1`'`len';') # declare argument string with extra stringlen 
> parameter
> 
> # declarations and initializations of canonical local variables
> define(`STRINGL',`')
> 
> # C integral type equivalent to a FORTRAN INTEGER
> define(`F_INTEGER',`int') 
> 
> # FORTRAN declaration for a long integer (e.g. integer*4 for Microsoft)
> define(`LONG_INT',`integer*4')
> 
> # FORTRAN declaration for a short integer (e.g. integer*2)
> define(`SHORT_INT',`integer*2')
> 
> # FORTRAN declaration for an integer byte (e.g. integer*1 or byte)
> define(`BYTE_INT',`integer*1')
> 
> # FORTRAN declaration for single precision
> define(`SINGLE_PRECISION',`real')
> 
> # FORTRAN declaration for double precision (e.g. real for a Cray)
> define(`DOUBLE_PRECISION',`double precision')
> 
> divert(0)dnl
> 
> - ----------------------------------------------------\

What you did looks correct (though I'll need more information).  We've
never used a FORTRAN-to-C translator on the ftest.f program, so I'm
currently leary of that.

Would you please do the following:

    1.  Go to the fortran/ subdirectory.

    2.  Remove the following files: ftest, ftest.o, ftest.f.

    3.  Execute the command `make ftest.f'.  Trap the output and send
        it to me.

    4.  Send me the file ftest.f.

    5.  Execute the command `make ftest.obj'.  Trap the output and send
        it to me.

--------
Steve Emmerson   <address@hidden>