Could not retrieve add offset

[Don Hooper]

Diane,

This is something else that mystifies me.  I've heard many times of this
segfaulting problem with templated sdfopen data.  Yet, we do this here every
day without problem.  I'd begun to wonder if Solaris did something right
with memory management that everyone else got wrong, as unlikely as that
seems.  However, now I have to wonder if you aren't on to something with
your "correct metadata" notion.  I've been known to be obsessive about
internal metadata in NetCDF files.  %}  It helps in so many ways, *if*
folks know about it and know how to access it, either via the NetCDF
API or via the ncdump command from Unidata's NetCDF suite.  Memory
management problems, however, are notoriously difficult to debug.  >SIGH<

-Hoop
> From owner-gradsusr at LIST.CINECA.IT  Fri May 19 11:02:08 2006
> From: Diane Stokes <Diane.Stokes at NOAA.GOV>
>
> Hoop,
>
> Your response had me intrigued because I have been grateful for the new
> 'dtype netcdf' DDF's in version 1.9.  Prior to that, it was rare that I
> could display a time series from templated netcdf files without
> segfaulting.  Using 'dtype necdf' solved this problem for me.  In fact,
> I was wondering why in so many of your response you claimed you hadn't
> yet needed to try it.
>
> >From your msg to Paul, I gather that my problem all this time may have
> been with insufficient or incorrect metadata in the netcdf files I was
> reading.  So, I was curious to see if I would get the usual segfault
> when reading some of your data.  The result:  I was able to read a 5
> year timeseries (5 yearly files of daily data) of your netcdf version of
> the Reanalysis slps (from
> ftp://ftp.cdc.noaa.gov/Datasets/ncep.reanalysis.dailyavgs/surface)
> without problems.
>
> So, if it is the metadata that makes the difference, thanks very much
> for taking the care to accurately describe the data you provide and for
> your work to help GrADS read these data.  It's good to know we don't
> need to worry about writing a DDF for netcdf data from the CDC server.
>
> Thanks also to the other GrADs developers that have given us users the
> ability to read netcdf files from other sources.  Sure, it would be
> ideal if all netcdf data followed the same standards, but in the
> meantime we can still make use of the data we find.
>
>     Cheers!
>        Diane
>
> Don Hooper wrote:
> > I am absolutely flummoxed.  I go to all kinds of effort to create
> > NetCDF files with sufficient metadata, and to teach GrADS how to
> > *automatically* (I.E., without a Data Descriptor File (DDF),
> > 'dtype netcdf' or otherwise), via the sdfopen command.  And yet
> > people keep trying to write DDFs that require version 1.9 to
> > read files that could be read without any back in version 1.6beta4.
> > The mind boggles.  And experts at COLA abet this foolish enterprise.
> > I just don't this *insistence* on doing things the hard way.  I
> > really don't.  Just say:
> > sdfopen /home/pfwillis/2006/Grads/circulation/Climate/data/slp_daily.1999.nc /home/pfwillis/2006/Grads/circulation/Climate/data/slp_daily.%y4.nc 731
> >
> > and go to work with the data.  No fuss, no muss.  Don't work so hard,
> > folks; I did the hard work for you years ago.  Reap the benefits of
> > my sweat, will you?  Please.
> >
> > -Hoop
> >
>