[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 950110: netCDF possible bug

>Organization: NOAA
>Keywords: 199501101741.AA11198 netCDF

Hi Cathy,

> Hi- I came across a behavior of netcdf that I'm not sure is a bug.  When I
> use the fortran call call ncopn(file,writetype,icode) and look at icode to
> see if a file is netcdf or not, I always get a 0, even if the file is not
> a netcdf file or even if it doesn't exist. When I check the return from
> the function (netid=ncopn.....) I get a -1 for nonnetcdf files and a 0 for
> netcdf files.
> 
> Since the icode is clearly meant as an error code, I would expect to check
> that in fortran and not the value of netid?
> 
> Coudl someone tell me if my interpretaion of icode is correct?

I was skeptical that such a bug could lurk undiscovered in the netCDF
sources for so long, but it looks it is a bug in both the C and Fortran
interfaces.  In the C interface, when ncopen() is called on a nonexistent
file, -1 is returned, but the global error variable `ncerr' is set to 0
instead of NC_SYSERR.  The Fortran interface uses `ncerr' to set its return
code, so it returns 0 in this case.  Hence if we fix the C behavior, the
Fortran will work OK.

However, things seem to work OK in the case of trying to open an existing
file that is not a netCDF file, as shown by the results I get in running the
test program below:

      program opbug
C  This program will open bogus files and check error returns
      include 'netcdf.inc'
C  Turn off fatal, verbose errors, so we can check return code
      call ncpopt(0)
C  Attempt to open a nonexistent file.
      ncid = ncopn("tooth_fairy.nc", NCNOWRIT, icode)
      print *,"icode=",icode, ", ncid=", ncid
C  Attempt to open non-netCDF file
      print *, "icode=",icode, ", ncid=", ncid
      end

Running this on my system produces the following output:

 icode=  0, ncid=  -1
 icode=  19, ncid=  -1

which shows the bug for the first case, but seems to verify the second case
is working OK.

I'll file this bug with a few others we have to fix in the next release.

Thanks for reporting it!

______________________________________________________________________________

Russ Rew                                                UCAR Unidata Program
address@hidden                                          P.O. Box 3000
http://www.unidata.ucar.edu/                          Boulder, CO 80307-3000
______________________________________________________________________________