Re: Fwd: ax_lib_hdf5 bug with library ordering, static libraries

Hello,

I've solved the problem described below by the attached change to
ax_lib_hdf5.m4 which first properly sets HDF5_FC and splits the
result of "$HDF5_FC -show" for the derivation of HDF5_FLIBS,
rather than doing the contortions in the original version.

I hope that this is helpful for others, too.

Enjoy,
Harald

On 08/30/16 23:14, Eric Blake wrote:
> On 08/30/2016 03:57 PM, Harald Anlauf wrote:
>>
>> I have a system where HDF5 installation seems to differ from what the
>> autoconf package ax_lib_hdf5 expects:
>>
> 
> Except that autoconf proper doesn't provide ax_lib_hdf5.  Given the ax_
> prefix, it probably belongs to the Autoconf Archive project,
> https://www.gnu.org/software/autoconf-archive/
> 
> I'm not sure if I have the correct bug reporting address for that
> project, so I've added the maintainer address mentioned in the manual in cc.
> 
>> % h5fc -show
>> gfortran -I/usr/include -L/usr/lib /usr/lib/libhdf5hl_fortran.a
>> /usr/lib/libhdf5_hl.a /usr/lib/libhdf5_fortran.a /usr/lib/libhdf5.a
>> -lpthread -lz -lrt -ldl -lm -Wl,-rpath -Wl,/usr/lib
>>
>> Using the latest snapshot of ax_lib_hdf5, I get:
>>
>> configure.in:
>>
>> AX_LIB_HDF5([serial])
>> echo "HDF5_LDFLAGS=$HDF5_LDFLAGS"
>> echo "HDF5_FLIBS=$HDF5_FLIBS"
>> echo "HDF5_FFLAGS=$HDF5_FFLAGS"
>>
>> Running configure:
>>
>> checking for a sed that does not truncate output... /usr/bin/sed
>> checking for gawk... gawk
>> checking for h5cc... /usr/bin/h5cc
>> checking for HDF5 type... serial
>> checking for HDF5 libraries... yes (version 1.8.12)
>> checking hdf5.h usability... yes
>> checking hdf5.h presence... yes
>> checking for hdf5.h... yes
>> checking for H5Fcreate in -lhdf5... yes
>> checking for main in -lhdf5_hl... yes
>> checking for matching HDF5 Fortran wrapper... /usr/bin/h5fc
>> HDF5_LDFLAGS=-L/usr/lib
>> HDF5_FLIBS= -lm -ldl -lrt -lz -lpthread -lhdf5_fortran -lhdf5
>> -lhdf5hl_fortran -lhdf5_hl
>> HDF5_FFLAGS=-I/usr/lib -L/usr/lib -I/usr/include
>>
>> The ordering of the libraries in HDF5_FLIB seems to get screwed up.
>> I think the logic in rewriting the FLIB part in lines 270-308 is not
>> prepared to handle my environment.
>>
>> Is somebody still working on the HDF5 support?
>>
>> Thanks,
>> Harald
>>
>>
>>
>