Re: 961002: problem with ncdump (fwd)

>To: address@hidden
>From: Patrick Guio <address@hidden>
>Subject: problem with ncdump (fwd)
>Organization: .
>Keywords: 199610020843.AA13593

Patrick,

The files you sent, dump2.3 and dump2.43, both have attribute
declarations such as the following in them:

                n_e:_FillValue = INFf ;
                n_e:missing_value = INFf ;

According to ncgen (either the 2.3 version of the 2.4.3 version), this
is a syntax error.  I'm not sure what value you originally specified for
can't recreate the original binary netCDF file using "ncgen -b".  

Would it be possible for you to send or make available via FTP a binary
netCDF file that demonstrates the problem?

That would help me to figure out why ncdump is displaying "_" for all
those values.  I'm speculating that the value you are giving for the
_FillValue is an IEEE NaN (not a number), and those have peculiar
properties in comparisons and arithmetic that may break the test we are
currently using for equality or proximity.  For example, if x is a NaN,
then the expression (x==x) is false.

It may also be a platform-specific bug, since no one else has reported
this problem, so I'd also like to know on what SGI platform and what
version of IRIX you are seeing the problem.  Thanks!

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu