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

Re: perl interface problem, plus ncdump oddity

>To: address@hidden
>From:  Jason Thaxter <address@hidden>
>Subject: Re: 20030715: perl interface problem, plus ncdump oddity 
>Organization: GoMOOS
>Keywords: 200309161520.h8GFKSLd005087, ncdump, FreeBSD, gcc

Jason,

> Likewise, sorry for the delay in responding, but this isn't an old
> freebsd system... if you test on older systems you may find no
> problem at all.  Nevertheless, I do have easy access to both current
> FreeBSD branchesn- the 4.x branch, which is stabler but destined for
> obsolescence, and the 5.x branch, which is the future, but still
> less stable than other FreeBSD (and is still pretty darn stable).
> 
> The only relevant differences I can see will be libc versions and
> gcc versions...

So from that, it sounds as if the problem is a bug in the older libc
or gcc versions on the FreeBSD system, rather than a bug in ncdump or
the netCDF library.  Nevertheless, I just build netCDF 3.5.0 on a
FreeBSD 4.7 system:

 % uname -a
 FreeBSD emo.unidata.ucar.edu 4.7-STABLE FreeBSD 4.7-STABLE #0: Fri Jan 24 
15:53:42 MST 2003     address@hidden:/usr/obj/usr/src/sys/MYKERNEL  i386

and tried the resulting ncdump on your netCDF file.  It appears to
work fine, with the correct
    
 time = 2452281, 2452281.04166667, 2452281.08333333, ...

rather than the bug you were seeing with five other values between
each correct time value:

 time = 2452281, 1345.53613472733, 4.16077581271464e-317, 
    4.16077581271464e-317, 6.4131214821086e+17, -0.0556587280591438, 
    2452281.04166667, 214.527954340916, 36700.1563112745, 1720.32031441483, 
    4.70626367892367e+16, 2.11248055210126e+25, 2452281.08333333, ...

But I saw other differences in the netcdf-ncdump-questionable output
you sent that might indicate the ncdump you are using on that system
is not from netCDF 3.5.0, but from a considerably older version,
perhaps netCDF 2.4.3.  For example, in netcdf-ncdump-questionable, the
following attributes values are printed using an older syntax:

                air_temperature_qc:_FillValue = '\200' ;
                air_temperature_qc:valid_range = '\0', '\177' ;
                air_temperature_qc:quality_good = '\0' ;
                air_temperature_qc:quality_bad = '\1' ;

whereas in the ncdump output I see from netCDF 3.5.0 on FreeBSD 4.7
and all other platforms, these appear instead as:

                air_temperature_qc:_FillValue = -128b ;
                air_temperature_qc:valid_range = 0b, 127b ;
                air_temperature_qc:quality_good = 0b ;
                air_temperature_qc:quality_bad = 1b ;

using the "b" suffix for byte constants that has been used in all
netCDF 3.x versions of ncdump, as far as I know.  Also in
nettcdf-ncdump-questionable, there are commas before the constants in
some data lists:

 air_temperature =
  4.657 ,
  3.669 ,
  4.018 ,

which was a bug fixed a long time ago, so that ncdump output in 3.5.0
and subsequent versions is:

 air_temperature =
  4.657,
  3.669,
  4.018,

So could you please check where you are getting the ncdump used on
your FreeBSD system, by using "which ncdump", for example.  If you
built but inadvertently didn't install the ncdump from version 3.5.0,
you may be using a much older version that had bugs that have long
since been fixed.

If the output you sent me was actually from a 3.5.0 version of ncdump,
I will be very surprised.

--Russ