Changes to fix FindHDF5.cmake on lxplus

The HDF5 find modules have a nasty habit of picking up the system version of HDF5, which is unfortunate given that:

• our build servers apparently don't have HDF5, but
• lxplus, and most other servers that scientists use, do have HDF5

Using lxplus as a test case, this version of the FindHDF5.cmake module seems to work.

Jenkins Build SUCCESS

Results available at: Jenkins [AE-MERGE-REQUEST #302]

Hi @akraszna, this is an attempt at fixing HDF5 so that it works on lxplus (or other clusters where it's already installed in an older version). I had to use find_package(HDF5 NO_MODULE) to get the version number, if this is going to be a problem let me know.

The other option (which, ironically, seems better supported by cmake) is to not use cmake to build the package, since building with the standard configure script will (I think) give us all the utilities we'd need to use the builtin FindHDF5.cmake module.

mentioned in merge request !209 (closed)

Let me do some tests. I'd prefer a different solution than this, but I'll have to test whether that actually works...

mentioned in merge request !226 (merged)

Hi @dguest,

I'd propose to fix this module's behaviour with !226 (merged) instead. Please have a look. I tried to run as many tests with it as I could come up with, I think it behaves correctly under all circumstances on lxplus.

Cheers, Attila

Hi,

To be honest I'm a bit confused as to how this works: if you look at the kitware github page on their FindHDF5.cmake module they say

# This module invokes the HDF5 wrapper compiler that should be installed
# alongside HDF5. Depending upon the HDF5 Configuration, the wrapper
# compiler is called either h5cc or h5pcc.  

Which is odd because when you build with cmake it doesn't give you h5cc. That's why I created !209 (closed). I'll give it a shot and see if I can break it.

Let me close this one, since !226 (merged) is now in the repository.

closed