Fortran warning at compile time, read error at execution

Question

I am trying to build a meteorological model in a Linux cluster (Ubuntu 12.04) using Intel fortran compiler version 14.0.1. But when trying to compile fdgrib2 module I get this warning message:

/opt/intel/bin/ifort -c  -I. /usr/local/rams60/src/pre/fdgrib2/grib_get.f90
/usr/local/rams60/src/pre/fdgrib2/grib_get.f90(209): warning #8042: This argument must be of type character of length 1.   [TOKENS]
   print*,'5--'//trim(tokens(1)), len_trim(tokens(1)),ichar(tokens(1))
------------------------------------------------------------^
ar rs fdgrib2.a grib_get.o
rm -f grib_get.o

tokens declaration

character(len=20) :: tokens(100)

Despite this warning the fdgrib2 executable is built but when trying to run stops with another error message:

/usr/local/rams60/build/fdgrib2$ ./fdgrib2 -f DGRIB_IN2
 num args:           2
  ---------------------------------------------------------------
 GRIB to GDF converter
 ---------------------------------------------------------------
 Namelist file name: DGRIB_IN2
 GRIB file name    : ./gfs.t00z.pgrbf06.grib2
 WGRIB exec        :
       Assumimg GRIB 2   : /usr/local/grib2/wgrib2/wgrib2
 Data date/time    : 99999999  
 Forecast hour     : 99999999  
 3D variables: UGRD      VGRD      TMP       HGT       RH        
 2D variables:


forrtl: severe (59): list-directed I/O syntax error, unit -5, file Internal List-Directed Read
Image              PC        Routine            Line        Source             
fdgrib2            08076240  Unknown               Unknown  Unknown
fdgrib2            08074C94  Unknown               Unknown  Unknown
fdgrib2            08057847  Unknown               Unknown  Unknown
fdgrib2            0804BFA2  Unknown               Unknown  Unknown
fdgrib2            0804A827  Unknown               Unknown  Unknown
libc.so.6          B72E94D3  Unknown               Unknown  Unknown

fdgrib2 process finishes at commands on grib_get.f90

  [fortran] call tokenize_ws(lines(1),tokens,ntok)
   nb = index(tokens(1),'=')+1
   ne = len_trim(tokens(1))
   read(tokens(1)(nb:ne),*) longdate [/fortran]

I am not an expert fortran programmer so I am not sure how to deal with this. Any help would be greatly appreciated.

Thanks in advance

EDIT 1

Added print statement to locate when the error comes up

   call tokenize_ws(lines(1),tokens,ntok)
   nb = index(tokens(1),'=')+1
   ne = len_trim(tokens(1))

print*,'P3'
print *,lines(1)
print *,tokens(1)
print *,tokens(1)(nb:ne)

read(tokens(1)(nb:ne),*) longdate

print*,'P4'

and the output (pass through P3 but does not arrive to P4). 2014012000 is the right date and time of the input data file

 P3
 1.1:0:d=2014012000
 1.1:0:d=2014012000
 2014012000

forrtl: severe (59): list-directed I/O syntax error, unit -5, file Internal List-Directed Read
Image              PC        Routine            Line        Source             
fdgrib2            080761C0  Unknown               Unknown  Unknown

EDIT 2

Code of subroutine tokenize_ws in tokenize.f90

subroutine tokenize_ws(str,tokens,ntok)

implicit none

integer :: ntok
character(len=*) :: str,tokens(*)
character(len=1) :: sep,tab

integer :: ntokmax=100,npt,nch,nc,ntbeg,ntend,n

! this routine "parses" character string str into different pieces
! or tokens by looking for  possible token separators (toks
! str contains nch characters.  the number of tokens identified is nto
! the character string tokens are stored in tokens.

! Modified 12/11/07: this version tokenizes by white space (spaces and tabs)
sep = char(32)
tab = char(9)

ntok=0
npt=1
nch=len_trim(str)
nc=1
do ntok=1,ntokmax
   do n=nc,nch
      if(str(n:n) /= sep .and. str(n:n) /= tab) then
         ntbeg=n
         goto 21
      endif
   enddo
   21 continue
   do n=ntbeg,nch
      if(str(n:n) == sep .or. str(n:n) == tab) then
         ntend=n-1
         goto 22
      endif
      if(n == nch) then
         ntend=n
         goto 22
      endif
   enddo
   22 continue
   tokens(ntok)=str(ntbeg:ntend)
   nc=ntend+1
   if(nc.ge.nch) goto 25
enddo

25 continue

return
end

EDIT 3

Added source code and input files (linux tar files)

http://ubuntuone.com/5tL2IO8OV5dBxhs18kt1WI

http://ubuntuone.com/1meYXbBuBt1Rxcx4ODMzVp

EDIT 4

Output of compile options -traceback -warn -check

log file

and output of model run

2014012000

       9          19 forrtl: severe (59): list-directed I/O syntax error, unit -5, file Internal List-Directed Read Image              PC

Routine Line Source fdgrib2
080942C0 Unknown Unknown Unknown fdgrib2
08092D14 Unknown Unknown Unknown fdgrib2
0806F822 grib_get_mod_mp_g 291 grib_get.f90 fdgrib2
08051CD0 MAIN__ 130 dgrib.f90 fdgrib2
0804A757 Unknown Unknown Unknown libc.so.6
B72CF4D3 Unknown Unknown Unknown

EDIT 5

After changing read(tokens(1)(nb:ne),*) longdate (chars 9 to 19) to read(tokens(1)(nb:ne-1),*) (chars 9 to 18) longdate it seems to properly read longdate, expected value was 2014012000. But then another error appears, as I post this info I'll continue looking for a solution. fdgrib2 is a "relatively old" read code for grib2 files but it worked and should still work, my fault. Thanks for your help.

print*,'P3'
print *,lines(1)
print *,tokens(1)
print *,tokens(1)(nb:ne)
print *,nb,ne
read(tokens(1)(nb:ne-1),*) longdate
print*,'P4'
print*,'ddd:', longdate

P3
 1.1:0:d=2014012000
 1.1:0:d=2014012000
 2014012000
          9          19
 P4
 ddd:  2014012000
 1--1.1:0:scan=0 input=WE:NS output=WE:SN:grid_template=0:winds(N/S):
--
 ppp0:           0         360         181 lat-lon             
 ppp0:  0.0000000E+00  0.0000000E+00  0.0000000E+00  0.0000000E+00
forrtl: severe (59): list-directed I/O syntax error, unit -5, file Internal List-Directed Read
Image              PC        Routine            Line        Source             
fdgrib2            08094350  Unknown               Unknown  Unknown
fdgrib2            08092DA4  Unknown               Unknown  Unknown
fdgrib2            0806D724  grib_get_mod_mp_g         219  grib_get.f90
fdgrib2            08051D20  MAIN__                    132  dgrib.f90
fdgrib2            0804A757  Unknown               Unknown  Unknown
libc.so.6          B72984D3  Unknown               Unknown  Unknown

Answer 1

Thanks to everyone who tried to help. Finally I could succeed.

My problem came from an old piece of software that was not optimised for Intel compiler, hence giving such error messages when reading grib2 files (GRIB2 is a format used in meteorology). Actually, a new version of the software has been uploaded and I could succesfully compile and run.

Thanks again

Answer 2

If you haven't done it already, then fix the cause of the warning raised by the compiler. The snippet

ichar(tokens(1))

passes a character variable of length 20 to ichar which expects a variable of length 1.

I don't see how this could cause your later problems but it's generally a good idea to grasp at every straw when trying to fix mysterious bugs.

Answer 3

Check the input files, and specifically the location the code is currently processing at the time of crash. There is some problem when reading longdate from the tokens(1)(nb:ne), that are made from the lines(1). You can add some print statement to make it clearer:

    print *,lines(1)
    print *,tokens(1)
    print *,tokens(1)(nb:ne)

Also, don't worry about the warning. It is not a good piece of code, but it is in a print statement, so it just prints something not that well defined to the standard output.