[ET Trac] [Einstein Toolkit] #1910: ExternalLibraries/HDF5 does not set HDF5_INC_DIRS

[Einstein Toolkit]

#1910: ExternalLibraries/HDF5 does not set HDF5_INC_DIRS
----------------------------+-----------------------------------------------
  Reporter:  barry.wardell  |       Owner:  eschnett           
      Type:  defect         |      Status:  assigned           
  Priority:  unset          |   Milestone:                     
 Component:  Other          |     Version:  development version
Resolution:                 |    Keywords:                     
----------------------------+-----------------------------------------------

Comment (by barry.wardell):

 Replying to [comment:10 eschnett]:
 > These are the lines that are missing:
 > {{{
 > # Set options
 > if [ "${FFTW3_DIR}" != 'NO_BUILD' ]; then
 >     : ${FFTW3_INC_DIRS="${FFTW3_DIR}/include"}
 >     : ${FFTW3_LIB_DIRS="${FFTW3_DIR}/lib"}
 > fi
 > }}}
 > (taken from FFTW3; obviously the library name need to be changed.)
 >
 > It is left as exercise for the reader to find out where these lines
 should go.

 The attached patch implements this suggestion. I have tested that it works
 and fixes the original problem.

 Replying to [comment:11 knarf]:
 > I would also suggest to add (assuming testing shows that's fine):
 > {{{
 > FFTW3_INC_DIRS="$(${CCTK_HOME}/lib/sbin/strip-incdirs.sh
 ${FFTW3_INC_DIRS})"
 > FFTW3_LIB_DIRS="$(${CCTK_HOME}/lib/sbin/strip-libdirs.sh
 ${FFTW3_LIB_DIRS})"
 > }}}
 >
 > ... with the same obvious replacements.

 Where should these go? I'm guessing right before the script finishes?

 > > For the release, I suggest setting HDF5_INC_DIRS manually, together
 with HDF5_DIR, as well as HDF5_LIB_DIRS and HDF5_LIBS.
 >
 > So, you suggest not to add those lines for the release?

 This affects all option lists that set HDF5_DIR but not HDF5_INC_DIRS or
 HDF5_LIB_DIRS. A quick check suggests that most simfactory optionlists
 fall into this category. Is it better to change all of these optionlists,
 or to fix the problem with detect.sh?

 > Testing this automatically isn't trivial, as the testsuites wouldn't
 test for the utilities to be build, nor (I think) does Jenkins. As far as
 I read from the ticket, the actual configuration dos work, it is "only"
 the utilities that fail to build in that case.

 Yes, it is "only" the utilities that are affected, but this means that the
 last thing you see when running ```sim build``` is a failed compilation.
 This can be confusing and seems to me like something that should be fixed
 before the release.

-- 
Ticket URL: <https://trac.einsteintoolkit.org/ticket/1910#comment:12>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit