NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
Hi,
At about line 30 in file netcdf4_constants.f90, the following code:
integer (kind = EightByteInt), parameter, public :: &
nf90_fill_uint = 4294967295_8
should be changed to read:
integer (kind = EightByteInt), parameter, public :: &
nf90_fill_uint = 4294967295_EightByteInt
This is needed because the NAG compiler does not, by default, use the
typical practice of using byte counts for kinds. An 8-byte integer
actually has a kind of 4. (Kinds are 1, 2, 3, and 4 for one-byte,
two-byte, four-byte, and eight-byte integers. Kind=8 is not accepted.)
NAG does have a -kind=byte compilation option. But use of this option introduces further problems. Namely that all Fortran code that USEs a module compiled in this mode must also be compiled the same way. Which is a pain for large programs that use libraries from multiple sources.
I checked the netcdf-fortran-4.4-beta1 tarball, and the bug is not fixed. The same code appears in file netcdf_constants.f90 at around line 152.
Walter
netcdfgroup archives: