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

Re: Problems with NetCDF on Solaris

> Organization: NOAA/PMEL
> Keywords: 199404222007.AA21825

Hi Steve,

> This is just an FYI report on some peculiarities we have found with NetCDF on
> Solaris.  (Kevin and I have both decided we HATE Solaris - even redirection
> doesn't work right.)
> 
> Earlier I reported that the NetCDF WRITE speeds under the HDF-bundled version
> of NetCDF were MUCH slower than your Unidata version (on Solaris).  We have
> confirmed that this is dramatically true in our application (FERRET).
> 
> The new news is that the Unidata version of NetCDF gets intermittent READ
> errors on Solaris (the worst kind to track down!).  We have switched back to
> the HDF-bundled version to get around these errors.

We have never seen any read errors such as you report under Solaris 2.3.
Occasionally the amd automounter we are running causes problems, but they
appear to be in amd, which is not from Sun.  I would be very interested in a
dataset and program that we could use to try to reproduce this problem.

> Here are the particulars such as they are ... I'll probe deeper if you have
> specific questions:
> 
> o NetCDF version 2.3.2
> o Solaris 5.3
> o SparcServer 1000
> o call to NCVGT1 (FORTRAN jacket) is returning zero intermittently where
> there
>       is in fact some non-zero data

When we installed Solaris 2.3 (SunOS 5.3) here, we applied a large number of
patches that Sun provided before we did anything with the machines.  Our
system administrator has continued to faithfully install patches as they are
made available.  I use Solaris 2.3 now as my main desktop development
platform and have not encountered the difficulties you report.  In fact I'm
convinced it has some significant advantages over SunOS 4.1.3.

I'd like to try Purify on a program that demonstrates the problem, to see if
there are any memory access bugs masquerading as the problems you report.

__________________________________________________________________________
                      
Russ Rew                                              UCAR Unidata Program
address@hidden                                        P.O. Box 3000
(303)497-8645                                 Boulder, Colorado 80307-3000