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

[netCDF #KMY-638599]: Problem building netcdf-4 on SuSE 10.3 / Intel



> Hello,
>
> OK, actually that subject line isn't quite honest.  I'm actually having
> trouble building HDF 1.8.1 as a preparatory step to making netcdf-4. The
> package builds OK, but then fails the 'make check' step with many hundreds
> of lines, the first few of which are:
>
> Testing  -- Dataspaces (h5s)
> Testing  -- Dataspace coordinates (coords)
> Testing  -- Shared Object Header Messages (sohm)
> *** UNEXPECTED RETURN from H5Dread is -1 line 2612 in tsohm.c
> HDF5-DIAG: Error detected in HDF5 (1.8.1) thread 0:
> #000: H5Dio.c line 175 in H5Dread(): can't read data
> major: Dataset
> minor: Read failed
> #001: H5Dio.c line 407 in H5D_read(): can't read data
> major: Dataset
> minor: Read failed
> #002: H5Dchunk.c line 1246 in H5D_chunk_read(): unable to read raw data
> chunk
> major: Low-level I/O
> minor: Read failed
> #003: H5Distore.c line 1893 in H5D_istore_lock(): data pipeline read failed
> major: Data filters
> minor: Filter operation failed
> #004: H5Z.c line 1012 in H5Z_pipeline(): filter returned failure during
> read
> major: Data filters
> minor: Read failed
> #005: H5Zdeflate.c line 103 in H5Z_filter_deflate(): inflateInit() failed
> major: Data filters
> minor: Unable to initialize object
> *** UNEXPECTED VALUE from H5Dread should be 100, but is 0 at line 2613 in
> tsohm.c
> HDF5-DIAG: Error detected in HDF5 (1.8.1) thread 0:

Wow, looks bad! :-(

Are you sure it's not something like running out of disk space during the
build?

>
> I tried looking for information on the HDF5 web site, but it's pretty
> sparse. The HDF5 FAQ refers questions to sci.data.formats (!) which was a
> great group years ago, but has exactly 8 posts in 2008, none of which is
> about HDF.  They have some precompiled HDF5 1.8.1 binaries, but the ones
> for 64-bit linux use the PGI compilers, which is not an option for me.

They have a very active support group, and they will get your build working:
http://hdfgroup.org/about/contact.html

>
> I did notice, though, that you have precompiled binaries for Linux 2.6 x64
> with the gnu compilers.  So my question is, can you let me know (or do you
> have on your web site somewhere) what flags you used to compile?  Or, for
> that matter, if you have any insights into the HDF errors I'm getting, any
> info would be appreciated.
>
> Thanks,

All the output for the 4.0 test builds can be found here:
http://www.unidata.ucar.edu/software/netcdf/builds/4_0/index.jsp

It a lot of output. Find the build of interest and then search for CFLAGS, etc.
The same flags would have been used to compile HDF5 1.8.1 on my system.

As you seem to already know, without HDF5 1.8.1, netCDF-4 is not going to work
properly. Are you sure that you want to turn on netCDF-4 in your release? You
only need to use it if you are planning on creating netCDF-4/HDF5 data sets.
Since this format is new, there are no existing datasets to support with it.

If you build netCDF without the --enable-netcdf-4 option, you will get a
library that supports the classic netCDF formats, and does not require HDF5.

Please let me know if you need more help.

Thanks,

Ed

Ticket Details
===================
Ticket ID: KMY-638599
Department: Support netCDF
Priority: Normal
Status: Closed