On Mon, May 30, 2016 at 10:32 AM, Ian Hinder <ian.hinder@aei.mpg.de> wrote:

On 30 May 2016, at 14:33, Roland Haas <rhaas@aei.mpg.de> wrote:

Hello all,

I guess this would be one of the few times where a change in HDF5
breaks binary compatibility between different minor HDF5 versions.

I believe there is some code that casts a hid_t (or some other hdf5
type) to a CCTK_INT (that I wrote) though I also think that that code
contains an assert that the actual value is small enough to fit into a
CCTK_INT.

One option would be to use the hdf5-18 package in MacPorts for this release, assuming that this fixes the problem.

For me, using Homebrew on OS X all tests pass apart from CT_MultiLevel.boostedpuncture/2procs. Homebrew currently provides HDF5 1.8.16, so it sounds likely that HDF5 1.8.10 is to blame for your test failures other than the CT_MultiLevel one.