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

Re: 19980630: Error status on Cray computers

This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.

  • Subject: Re: 19980630: Error status on Cray computers
  • Date: Tue, 30 Jun 1998 12:37:54 -0600
Olaf,

> To: address@hidden,
> From: Olaf Heudecker <address@hidden>
> Subject: Error status on Cray computers
> Organization: Alfred-Wegener-Instutute for Polar and Marine Reseach
> Keywords: 199806301210.GAA13833

In the above message, you wrote:

> one of our users experienced a strange behavior of a netcdf
> application: it ran silently when called interactively but failed to
> run in batch (there was no error-string from NF_STRERROR, but
> status=5010).
> 
> After some investigation it turned out that the memory limits in the
> different modes where different and status=5010 means errno=5010 and
> this again means 
> (looking up in /opt/ctl/craylibs/3.0.2.0/include/liberrno.h):
> #define FDC_ERR_NOMEM   5010    /* no memory! */
> 
> It turned out, that the ffio-errnos on Cray Computers aren't
> recognized by strerror in libsrc/error.c (i.e. in libc). This means
> that ffio errors can't be interpreted by standard means. It results in 
> long debugging and code developping times. In fact there have been
> several encounters  at our institute concerning this problem within a
> few months. 
> 
> I'm no expert in unix system programming but it appears to me that
> there is no simple way to solve this problem (except for calling Cray
> for help). 
> 
> A pragmatic help could be: There are a few errors that might be of
> major interest. These could be cared for with a quick-and-dirty
> solution like the one that follows:
> 
> ------------ in nc_strerror before calling strerror -----
> /* Cray ffio error codes in /opt/ctl/craylibs/3.0.2.0/include/liberrno.h */
> 
> switch ( err ) {
>   case 5010:
>       return "ffio: Out of memory (buffer allocation?)!\n" ;
>   case 5008:
>       return "ffio: Maximum block size exceeded!\n";
>   case 5044:
>       return "ffio: Buffer size is negative or too large\n";
> }
> 
> if ( (err > 5000) && (err < 5066) ) return "ffio error (see liberrno.h in 
> craylibs directory)\n"
> ----------- end suggestion ---------------------
> 
> I will try to include this into netcdf-3.4 on a Cray J90 and supply a
> patch if you like.
> 
> Besides I will send an email to Cray/SGI and keep you informed about
> the status of this query.
> 
> Sorry for keeping you busy and thanks in advance 
> 
> Olaf R. Heudecker
> Alfred-Wegener-Instutute for Polar and Marine Reseach
> Bremerhaven, Germany

Thanks for this message.

Unfortunately, the person in charge of netCDF development is on vacation
at the moment and I don't know when he'll return.  I'll see that he
receives it, however.

--------
Steve Emmerson   <http://www.unidata.ucar.edu>