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

[netCDF #FIL-651276]: netcdf-c-4.3.1-rc2 ncdump test failures



Dom,

> Thank you for your and your colleagues efforts on this topic. Unfortunately I 
> cannot grant access to the system myself and I believe it is a little tricky 
> to get this done officially. Those HPC centers are bureaucratic and require 
> lots of paperwork. I would guess that the problem lies in the hdf5 library 
> that was compiled by the sysadmins or the compilers, but maybe I used 
> different configuration options than your colleague. If he could send me his 
> settings, I would give it another try.

Here are the settings used:

Nothing too special.  Here's the HDF5 & netCDF-4 configuration lines:

HDF5:
env CC=mpicc FC=mpif90 ../hdf5_1_8/configure --disable-shared --enable-debug 
--disable-production --enable-parallel
--enable-fortran --enable-build-all --prefix=/test/build_hdf5_1_8_parallel/hdf5

netCDF-4:
env CC=mpicc CFLAGS="-g -O0" 
CPPFLAGS=-I/test/build_hdf5_1_8_parallel/hdf5/include 
LDFLAGS=-L/test/build_hdf5_1_8_parallel/hdf5/lib ../netcdf-c/configure 
--enable-dependency-tracking --enable-netcdf-4
--enable-extra-tests --enable-extra-example-tests 
--prefix=/test/install_parallel --disable-shared
--enable-logging --enable-parallel-tests

--Russ

> On 9/09/2013, at 5:57 PM, Unidata netCDF Support <address@hidden> wrote:
> 
> > Hi Dom,
> >
> > We had someone build and run your test program, and he reported:
> >
> >  I spent some time today looking into this, and unfortunately, I
> >  can't duplicate the failure.  I've tried on a POWER7 system running
> >  Linux that I have access to, but that passes all the tests properly.
> >  The source & test code looks good to me also, so I don't have any
> >  good options for helping Dominikus out, unless he'd like to give me
> >  access to the system where the failure is occurring.
> >
> > Is it possible to provide an external user access to your IBM Power6
> > system, with enough resources to build, run, and debug this problem?
> >
> > --Russ
> >
> >>> I built netcdf-c-4.3.1 from the latest release candidate 2 against 
> >>> hdf5-1.8.11, 64bit, compiled by the sysadmins of the HPC I am using (IBM 
> >>> Power6). Unfortunately, make check fails in directory ncdump for 5 tests, 
> >>> and I think they are all related to scalar variables/dataspaces in 
> >>> NetCDF4 files.
> >>
> >> I see you are enabling many options for configure:
> >>
> >>> $ ./configure --prefix=/pf/k/k203080/3rdparty --enable-netcdf4 
> >>> --enable-extra-example-tests --enable-dependency-tracking 
> >>> --enable-dap-long-tests --enable-extra-tests --enable-large-file-tests 
> >>> --enable-shared --enable-static
> >>
> >> We don't test releases with all combinations of configure options
> >> enabled.  The default configure will run the tests we use for
> >> development.  It will also build a library with netCDF-4 enabled (if
> >> an HDF5 library is found installed where CPPFLAGS and LDFLAGS), with
> >> DAP remote access enabled (if libcurl is installed), and with shared
> >> and static libraries.  The --enable-dap-long-tests and
> >> --enable-extra-example-tests are intended for developers, and may
> >> sometimes include tests that don't yet pass on all platforms,
> >> especially in release candidates.
> >>
> >>> I read about a problem with scalar data in NetCDF4 files with the earlier 
> >>> version 4.3.0 (for which I got three failures in ncdump) and that this 
> >>> was fixed in rc1. Presumably my assumption that the bug was related to 
> >>> the failures was wrong?
> >>
> >> We don't know yet.  All the ncdump tests pass on our test platforms, 
> >> including Linux,
> >> Solaris, OSX, and Windows, but we don't have access to an up-to-date AIX 
> >> platform on
> >> which we can duplicate the problems you are seeing.
> >>
> >>> I am attaching my config.log that contains all the necessary information 
> >>> for my build and the ncdump/test-suite.log.
> >>>
> >>> Note that I can use the NetCDF library without problems with WRF, so I am 
> >>> not sure if it is a real problem - but I wanted to report it as requested 
> >>> in the make check output.
> >>
> >> Thanks for reporting the problems!  If we can reproduce them, we'll put 
> >> them in our
> >> bug-tracking system for fixing.
> >>
> >> --Russ
> >>
> >>> Regards
> >>>
> >>> Dom
> >>>
> >>> --
> >>> Dr. Dominikus Heinzeller
> >>>
> >>> Karlsruhe Institute of Technology (KIT)
> >>> Institute for Meteorology and Climate Research (IMK-IFU)
> >>> Kreuzeckbahnstr. 19
> >>> 82467 Garmisch-Partenkirchen, Germany
> >>>
> >>> Phone +49-8821-183259
> >>> Email address@hidden
> >>>
> >> Russ Rew                                         UCAR Unidata Program
> >> address@hidden                      http://www.unidata.ucar.edu
> >>
> >>
> >
> > Russ Rew                                         UCAR Unidata Program
> > address@hidden                      http://www.unidata.ucar.edu
> >
> >
> >
> > Ticket Details
> > ===================
> > Ticket ID: FIL-651276
> > Department: Support netCDF
> > Priority: Normal
> > Status: Closed
> >
> 
> 
Russ Rew                                         UCAR Unidata Program
address@hidden                      http://www.unidata.ucar.edu



Ticket Details
===================
Ticket ID: FIL-651276
Department: Support netCDF
Priority: Normal
Status: Closed