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

[netCDF #QGA-640676]: Fwd: 64-bit offset flag ignored by netcdf 4.3.3.1/4.4.0rc3 for pnetcdf?

Hi Ted,

I saw your post on the mailing list and have been trying to find out why it is 
like this; I agree that it *looks* like an oversight, but since I didn't write 
that code I can't say for certain.  It appears that it is new code coming in 
since the last release, and I've queried Dennis Heimbigner re: the context.  
I'll follow up with you when I know the reasoning or, hopefully, it can be 
removed so that we don't clobber the flag.

Thanks!

-Ward

> I thought I would forward this to the support email, too. It looks like an 
> error/oversight to me, but maybe there was a reason for this change? The "#if 
> 0" makes me think that there's a higher level of code with a preprocessed 
> token?
> 
> Cheers,
> 
> -- Ted
> 
> Begin forwarded message:
> 
> > From: Ted Mansell <address@hidden>
> > Subject: 64-bit offset flag ignored by netcdf 4.3.3.1 for pnetcdf?
> > Date: October 9, 2015 at 5:43:58 PM CDT
> > To: "netcdfgroup (address@hidden)" <address@hidden>
> >
> > Howdy,
> >
> > I have been using netcdf-c-4.3.1.1 (with a pnetcdf bug fix) for my weather 
> > model production runs, and just tried using 4.3.3.1 on a new system. I use 
> > the netcdf-4 interface to create a pnetcdf file, for which I pass the 
> > 64-bit offset bits:
> >
> >  cmode  = ior(ior(NF90_PNETCDF,NF90_MPIIO),nf90_64bit_offset)
> >  status = nf90_create(filename, cmode, ncid, comm = my_comm, info = my_info)
> >
> > I found that the files were being created as just "classic", however.
> >
> > Wei-keng Liao found the problem in libsrc5/nc5dispatch.c, which had this:
> >
> > 75 #if 0
> > 76     cmode &= (NC_WRITE | NC_NOCLOBBER | NC_LOCK | NC_SHARE | 
> > NC_64BIT_OFFSET);
> > 77 #else
> > 78     cmode &= (NC_WRITE | NC_NOCLOBBER | NC_LOCK | NC_SHARE );
> > 79 #endif
> >
> > which for whatever reason was clobbering the NC_64BIT_OFFSET bits. Is there 
> > any reason not to always use the first line, i.e., change this to "#if 1"? 
> > This is what I did, and it fixed the problem.
> >
> > 4.4.0-rc3 has the same issue in libsrcp/ncpdispatch.c (same line numbers). 
> > Shouldn't the first line (with "| NC_64BIT_OFFSET") be the default?
> >
> > Cheers,
> >
> > -- Ted
> >
> > __________________________________________________________
> > | Edward Mansell <address@hidden>
> > | National Severe Storms Laboratory
> > |--------------------------------------------------------------
> > | "The contents of this message are mine personally and
> > | do not reflect any position of the U.S. Government or NOAA."
> > |--------------------------------------------------------------
> >
> 
> 


Ticket Details
===================
Ticket ID: QGA-640676
Department: Support netCDF
Priority: Normal
Status: Closed