Re: Compiling with netCDF ver.3-3 on HP 10.20

Leigh,

> Just a quick follow up.  We went ahead and downloaded netCDF version
> 3.3 so that I can get access to the file descriptor of an open
> netCDF data file using NcFile::id().  When I compile software here at
> FSL on HPUX 10.20, I get several warning messages.  Also, I get an error
> message that a put()/get() call can't resolve the long data type.  The
> compiler error message shows that there is a "nclong" type, but I don't
> see a long type.  So my questions for you are:
> 
> 1.  Is there a document that shows what compiler switches should be set
>     to compile programs using netCDF ver. 3.3 using various computer
>     architectures (ours in particular HPUX version 10.20). These compiler
>     switches may quiet down a lot of the warning messages that show up
>     under 10.20.

Yes, we will have that platform covered when we release the next minor
version, but for now, check out

  http://www.unidata.ucar.edu/packages/netcdf/other-builds.html#hpux-10

We didn't have an HPUX 10 to test on when we released the last version.

> 2.  I don't see an overloaded version of the NcVar::put() or get()
>     functions that understand the long data type.  I do, however, see that
>     there is an "nclong" data type.  If I want to "put()" a long data value
>     do I have to cast it to a "nclong" in order to put() it?  Maybe this
>     information (and more) is available in some netCDF version 3.3 document
>     somewhere.  If so, could you direct me to one?

If by "long" data value, you mean a 64-bit integer, there is no way to
do this in netCDF, because it only supports 8-, 16-, and 32-bit integers
(because it's based on XDR, and that's all XDR had).

If a "long" on your HP is a 32-bit integer (I'm almost sure it is, but
can't check right now), then I'm afraid you have to cast to "nclong", an
artifact left-over from netCDF-2 that was a typedef for "int" on most
platforms.  We haven't been working on the C++ interface lately, but
have been putting our efforts into a good Java interface instead, so the
C++ is beginning to creak a bit.  It really should have interfaces
overloaded for both int and long instead of nclong.

--Russ