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

[netCDF #AWC-509858]: H5Fflush not found



Good morning James,

We are getting closer, although they behavior you observe does not suggest
that. I’ve looked at the config.log you provided, and whilst you are still
failing to “find libhdf5”, it is now for a different reason:

configure:16765: /opt/gcc-4.9.2/bin/gcc -m64 -D__LINUX__
-D__AMD64_LINUX2_64_NA__ -D__AMD64_LINUX_NA__ -o conftest -DNDEBUG
-fPIC -D_FILE_OFFSET_BITS=64 -D_LARGE_FILES=1 -D_XOPEN_SOURCE=1 -O3
-DEXA_CURRENT_YEAR=2016   -I.
-I/fa/sw/hdf5/1.8.16-01/amd64_linux_pic/include
-L/fa/sw/hdf5/1.8.16-01/amd64_linux_pic/lib -fPIC
-D_FILE_OFFSET_BITS=64 -D_LARGE_FILES=1 -D_XOPEN_SOURCE=1 -O3
conftest.c -lhdf5  -ldl -lm -lz  >&5
$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5
/opt/binutils-2.25/bin/ld: warning: libsz.so.2, needed by
/fa/sw/hdf5/1.8.16-01/amd64_linux_pic/lib/libhdf5.so, not found (try
using -rpath or -rpath-link)
/fa/sw/hdf5/1.8.16-01/amd64_linux_pic/lib/libhdf5.so: undefined
reference to `SZ_BufftoBuffDecompress'
/fa/sw/hdf5/1.8.16-01/amd64_linux_pic/lib/libhdf5.so: undefined
reference to `SZ_encoder_enabled'
/fa/sw/hdf5/1.8.16-01/amd64_linux_pic/lib/libhdf5.so: undefined
reference to `SZ_BufftoBuffCompress'

It appears that you are using a libhdf5 which was built with szlib support,
but the system is not attempting to link against libsz. Try setting the
environmental variable LIBS=-lsz. On some systems, it would be LIBS=-lszip,
so try that if the former doesn’t work. If this fails to work, it may be
the case that libsz is not on the system path/is in a custom location. I’m
guessing it is in the same location as libhdf5, however, and we just need
to instruct the build system to link against it.

Let me know if this gets us any further,

-Ward

address@hidden> wrote:

New Client Reply: H5Fflush not found
>
> I am perplexed....
>
> Indeed, my HDF5 is located in an unusual location
> /fa/sw/hdf5/1.8.16-01/amd64_linux_pic
>
> It's not a local build - they're pre-built binaries from the hdf group -
> an extract of hdf5-1.8.16-linux-centos6-x86_64-gcc447-shared.tar.gz
>
>
> noreaster:/fa/sw/netcdf/4.3.3.1-01-jrm-02% ls
> /fa/sw/hdf5/1.8.16-01/amd64_linux_pic/lib
> libhdf5.a              libhdf5_fortran.so libhdf5_hl_cpp.so.11
> libhdf5_hl.so libhdf5.so.10.1.0  libz.so.1.2.5
> libhdf5_cpp.a          libhdf5_fortran.so.10
> libhdf5_hl_cpp.so.11.0.0     libhdf5_hl.so.10      libsz.so
> libhdf5_cpp.so         libhdf5_fortran.so.10.0.2
> libhdf5hl_fortran.a          libhdf5_hl.so.10.0.2  libsz.so.2
> libhdf5_cpp.so.11      libhdf5_hl.a libhdf5hl_fortran.so
> libhdf5.settings libsz.so.2.0.0
> libhdf5_cpp.so.11.0.0  libhdf5_hl_cpp.a libhdf5hl_fortran.so.10
> libhdf5.so            libz.so
> libhdf5_fortran.a      libhdf5_hl_cpp.so libhdf5hl_fortran.so.10.0.2
> libhdf5.so.10         libz.so.1
>
> noreaster:/fa/sw/netcdf/4.3.3.1-01-jrm-02% ls
> /fa/sw/hdf5/1.8.16-01/amd64_linux_pic/include
> H5AbstractDs.h         H5DcreatProp.h H5FDdirect.h
> H5IdComponent.h        h5o.mod H5Spublic.h
> H5ACpublic.h           h5d.mod H5FDfamily.h
> h5im.mod               h5o_provisional.mod H5StrType.h
> h5a.mod                H5DOpublic.h H5FDlog.h
> h5i.mod                H5Opublic.h h5tb.mod
> H5api_adpt.h           h5d_provisional.mod H5FDmpi.h
> H5IMpublic.h           H5overflow.h H5TBpublic.h
> h5a_provisional.mod    H5Dpublic.h H5FDmpio.h
> H5Include.h            H5PacketTable.h h5test_kind_sizeof_mod.mod
> H5Apublic.h            h5ds.mod H5FDmulti.h
> H5IntType.h            H5PLextern.h h5t.mod
> H5ArrayType.h          H5DSpublic.h H5FDpublic.h
> H5Ipublic.h            H5PLpublic.h h5t_provisional.mod
> H5AtomType.h           H5DxferProp.h H5FDsec2.h
> h5lib.mod              h5p.mod H5Tpublic.h
> H5Attribute.h          h5e.mod H5FDstdio.h
> h5lib_provisional.mod  h5p_provisional.mod H5VarLenType.h
> H5Classes.h            H5EnumType.h H5File.h
> H5Library.h            H5Ppublic.h H5version.h
> H5CommonFG.h           h5e_provisional.mod H5FloatType.h
> h5l.mod                H5PredType.h h5z.mod
> H5CompType.h           H5Epubgen.h h5f.mod
> H5Location.h           H5PropList.h H5Zpublic.h
> H5CppDoc.h             H5Epublic.h h5fortran_types.mod
> h5l_provisional.mod    H5PTpublic.h hdf5.h
> H5Cpp.h                H5Exception.h h5f_provisional.mod
> H5Lpublic.h            H5pubconf.h hdf5_hl.h
> H5Cpublic.h            H5f90i_gen.h H5Fpublic.h
> h5lt.mod               H5public.h hdf5.mod
> H5DataSet.h            H5f90i.h h5global.mod
> H5LTpublic.h           h5r.mod
> H5DataSpace.h          H5FaccProp.h h5g.mod
> H5MMpublic.h           h5r_provisional.mod
> H5DataType.h           H5FcreatProp.h H5Gpublic.h
> H5Object.h             H5Rpublic.h
> h5_dble_interface.mod  H5FDcore.h H5Group.h
> H5OcreatProp.h         h5s.mod
>
>
> I've also gone back and revisited my CFLAGS and LDFLAGS environment
> variables.  They now seem to be correct.
> echo "Starting target configuration" ; \
>                    CC="/opt/gcc-4.9.2/bin/gcc -m64 -D__LINUX__
> -D__AMD64_LINUX2_64_NA__ -D__AMD64_LINUX_NA__"; export CC; \
>                    CXX="/opt/gcc-4.9.2/bin/g++ -m64 -D__LINUX__
> -D__AMD64_LINUX2_64_NA__ -D__AMD64_LINUX_NA__"; export CXX; \
>                    AR="ar"; export AR; \
>                    CFLAGS="-DNDEBUG -fPIC -D_FILE_OFFSET_BITS=64
> -D_LARGE_FILES=1 -D_XOPEN_SOURCE=1 -O3 -DEXA_CURRENT_YEAR=2016   -I .
> -I/fa/sw/hdf5/1.8.16-01/amd64_linux_pic/include"; export CFLAGS; \
>                    RANLIB="echo"; export RANLIB; \
> LDFLAGS="-L/fa/sw/hdf5/1.8.16-01/amd64_linux_pic/lib -fPIC
> -D_FILE_OFFSET_BITS=64 -D_LARGE_FILES=1 -D_XOPEN_SOURCE=1 -O3 "; export
> LDFLAGS; \
>                    cd src; \
>                    sh -vx ./configure  --disable-rpath --disable-dap
> --with-zlib=/fa/sw/zlib/1.2.8-06/
> --prefix=/fa/sw/netcdf/4.3.3.1-01-jrm-02/amd64_linux_pic 2> configure.log ;
>
> Yet - I'm still failing....!?
>
> Latest config.log attached....
>
> -jrm
>
>
> On 04/11/2016 05:57 PM, Unidata netCDF Support wrote:
> > Hi James,
> >
> > Thanks! I’m a bit surprised; often in these cases, the can't find hdf5
> > error is masking some other problem. A missing upstream dependency, an
> > incorrect LD_LIBRARY_PATH value, etc. However, in this case it appears
> that
> > configure simply can’t find the libhdf5 file. I base this on the
> following:
> >
> > configure:16765: /opt/gcc-4.9.2/bin/gcc -m64 -D__LINUX__
> > -D__AMD64_LINUX2_64_NA__ -D__AMD64_LINUX_NA__ -o conftest -DNDEBUG
> > -fPIC -D_FILE_OFFSET_BITS=64 -D_LARGE_FILES=1 -D_XOPEN_SOURCE=1 -O3
> > -DEXA_CURRENT_YEAR=2016   -I.     -fPIC -D_FILE_OFFSET_BITS=64
> > -D_LARGE_FILES=1 -D_XOPEN_SOURCE=1 -O3   conftest.c -lhdf5  -ldl -lm
> > -lz  >&5
> > $CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext
> $LIBS >&5
> > /opt/binutils-2.25/bin/ld: cannot find -lhdf5
> >
> > When you installed libhdf5, where did you install it? I see in config.log
> > that the CFLAGS and LDFLAGS aren’t set. If you installed libhdf5 in a
> > non-default location, you would need to set these variables as follows
> (my
> > apologies for telling you something you probably already know):
> >
> > $ CFLAGS="-I/custom/location/include" LDFLAGS="-L/custom/location/lib"
> > ./configure [your configure options]
> >
> > Speaking of configuration options, I notice you’re passing
> --disable-fortran
> > to configure. That’s no longer necessary, as the fortran interface is in
> a
> > completely separate package now. No harm is being done by invoking it, it
> > just doesn’t do anything anymore.
> >
> > In any event, it looks like we need to make the compiler aware of where
> > libhdf5 was installed. If you can tell me where it was installed, I can
> > give you a specific command to run, or you can determine those values
> > yourself and give ‘configure’ another shot.
> >
> > Thanks!
> >
> > -Ward
> >
> > address@hidden> wrote:
> >
> > New Client Reply: H5Fflush not found
> >> Sure - I've got that file - here you go....
> >>
> >> -jrm
> >>
> >> On 04/11/2016 05:29 PM, Unidata netCDF Support wrote:
> >>> Hi,
> >>>
> >>> Thanks for confirming the presence of libdl; that helps dial things in.
> >>> Looking at the file you attached, it still looks like it might be the
> >>> `configure` script?  When configure is run, it generates a file
> >>> `config.log` (note that it's config, not configure).  You shouldn't
> need
> >> to
> >>> capture any output or anything.  The main thing `config.log` gives us
> is
> >>> that it's very readable, particularly compared to the `configure`
> script,
> >>> and contains more information than the output sent by the configure
> >> script
> >>> to standard output.
> >>>
> >>> I've attached a sample config.log file so you can see what I'm talking
> >>> about.  If you can locate and send off the config.log from your build
> >>> attempt, I will take a look and see if I can sort this out!
> >>>
> >>> Have a great day,
> >>>
> >>> -Ward
> >>>
> >>>
> >>>
> >>> address@hidden> wrote:
> >>>
> >>>> New Client Reply: H5Fflush not found
> >>>>
> >>>> Thanks for the very fast turn around!
> >>>>
> >>>> I attach the correct version of configure.log - or at least - one that
> >>>> finishes with a non-zero exit...
> >>>>
> >>>> I would be kind of surprised if libdl wasn't present.  Indeed - on
> /lib
> >>>> I find:
> >>>>
> >>>>       -rwxr-xr-x 1 root root 20668 May 30  2013 libdl-2.5.so
> >>>>       lrwxrwxrwx 1 root root    12 Dec 29  2012 libdl.so.2 ->
> >> libdl-2.5.so
> >>>> -jrm
> >>>>
> >>>>
> >>>> On 04/11/2016 03:35 PM, Unidata netCDF Support wrote:
> >>>>> Hello James,
> >>>>>
> >>>>> Thanks for providing all of the information; as you've deduced, the
> >> HDF5
> >>>> library 1.8.8 is too old to use with modern netCDF.  1.8.9 is the
> >> minimum
> >>>> version required, and 1.8.16 should work just fine.  It appears that
> >> you do
> >>>> not have `libdl` installed on your system.  If this is the case, you
> can
> >>>> either install it, or, if hdf5 was built without dynamic-loading
> >> support,
> >>>> you should be able to pass `--disable-dynamic-loading` (instead of
> using
> >>>> LIBS=-ldl).  If libdl is present, `configure` should find it itself,
> and
> >>>> you wouldn't need to force inclusion via `LIBS`.
> >>>>> It looks like the file you attached, ` Config_4.3.3.1.log`, is
> actually
> >>>> the `configure` script and not the generated `config.log` file
> generated
> >>>> when you ran configure.  If you provide the config.log file I can look
> >> at
> >>>> it to confirm my suspicions, or you can try my suggestions and let me
> >> know
> >>>> if they work, fail, or if the symptoms change.
> >>>>> I hope this helps,
> >>>>>
> >>>>> -Ward
> >>>>>
> >>>>>> I'm trying to build a newer version of NETCDF and am encountering
> the
> >>>>>> H5Fflush failure on configure.  Specifically:
> >>>>>>
> >>>>>> checking for library containing H5Fflush... no
> >>>>>>
> >>>>>> I'm trying to work forward from an older NETCDF build that worked,
> to
> >>>>>> something newer.
> >>>>>>
> >>>>>> * Our last working build of NETCDF was 4.2.1.1, using HDF5 1.8.8.  I
> >>>>>> was able to reproduce this.
> >>>>>> * I obtained the HDF 1.8.16, and was also able to use that to repeat
> >>>>>> the 4.2.1.1 build
> >>>>>> * Substituting in the NETCDF 4.3.3.1 archive in the procedure that
> >>>>>> built 4.2.1.1 yields the configure failure
> >>>>>> * I observed the remarks about the need for "LIBS=-ldl" in [netCDF
> >>>>>> #AWC-509858], so added that env setting to my build
> >>>>>> * Reproduced the 4.2.1.1/HDF 1.8.16 build with the "LIBS=-ldl"
> >> setting
> >>>>>> * NETCDF 4.3.3.1 build modified with "LIBS=-ldl" still yields the
> >>>>>> configure failure
> >>>>>>
> >>>>>>
> >>>>>> I am attaching four files as follows:
> >>>>>>
> >>>>>> 1. Build_4.2.1.1.log - stdout/err for configure and build of NETCDF
> >>>>>> 4.2.1.1 w/HDF5 1.8.16 (successful)
> >>>>>> 2. Config_4.2.1.1.log - configure.log associated with the build of
> >>>> 4.2.1.1
> >>>>>> 3. Build_4.3.3.1.log - stdout/err for configure of NETCDF 4.3.3.1
> >>>>>> w/HDF5 1.8.16 (failed)
> >>>>>> 4. Config_4.3.3.1.log - configure.log associated with attempted
> build
> >>>>>> 4.3.3.1
> >>>>>>
> >>>>>>
> >>>>>> Assistance of a prompt nature would be greatly appreciated....
> >>>>>>
> >>>>>> -jrm
> >>>>>>
> >>>>>> James Mason
> >>>>>> Exa Corp
> >>>>>> Burlington, MA
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>> Ticket Details
> >>>>> ===================
> >>>>> Ticket ID: AWC-509858
> >>>>> Department: Support netCDF
> >>>>> Priority: Normal
> >>>>> Status: Closed
> >>>>
> >>>>
> >>>> Ticket Details
> >>>> ===================
> >>>> Ticket ID: AWC-509858
> >>>> Department: Support netCDF
> >>>> Priority: Normal
> >>>> Status: Open
> >>>> Link:
> >>>>
> >>
> https://andy.unidata.ucar.edu/esupport/staff/index.php?_m=tickets&_a=viewticket&ticketid=22078
> >>>
> >>> Ticket Details
> >>> ===================
> >>> Ticket ID: AWC-509858
> >>> Department: Support netCDF
> >>> Priority: Normal
> >>> Status: Open
> >>
> >>
> >>
> >> Ticket Details
> >> ===================
> >> Ticket ID: AWC-509858
> >> Department: Support netCDF
> >> Priority: Normal
> >> Status: Open
> >> Link:
> >>
> https://andy.unidata.ucar.edu/esupport/staff/index.php?_m=tickets&_a=viewticket&ticketid=22078
> >>
> > ​
> >
> >
> >
> > Ticket Details
> > ===================
> > Ticket ID: AWC-509858
> > Department: Support netCDF
> > Priority: Normal
> > Status: Open
>
>
>
>
> Ticket Details
> ===================
> Ticket ID: AWC-509858
> Department: Support netCDF
> Priority: Normal
> Status: Open
> Link:
> https://andy.unidata.ucar.edu/esupport/staff/index.php?_m=tickets&_a=viewticket&ticketid=22078

​



Ticket Details
===================
Ticket ID: AWC-509858
Department: Support netCDF
Priority: Normal
Status: Open